From: Subject: Mihai Christodorescu, Somesh Jha "Static Analysis of Executables to Detect Malicious Patterns" Date: Tue, 16 Oct 2007 02:26:23 +0800 MIME-Version: 1.0 Content-Type: multipart/related; type="text/html"; boundary="----=_NextPart_000_0000_01C80F9B.F234E740" X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.3028 This is a multi-part message in MIME format. ------=_NextPart_000_0000_01C80F9B.F234E740 Content-Type: text/html; charset="gb2312" Content-Transfer-Encoding: quoted-printable Content-Location: http://www.cs.wisc.edu/wisa/papers/security03/cj03.html Mihai Christodorescu, Somesh Jha "Static Analysis of = Executables to Detect Malicious Patterns"

Static Analysis of Executables to Detect Malicious Patterns*

Mihai Christodorescu
mihai@cs.wisc.edu = 		     Somesh Jha
mihai@cs.wisc.edu=20

Computer Sciences=20 Department
University of = Wisconsin,=20 Madison

Abstract

Malicious code detection is a crucial component of any defense=20 mechanism. In this paper, we present a unique viewpoint on = malicious code=20 detection. We regard malicious code detection as an=20 obfuscation-deobfuscation game between malicious code writers and=20 researchers working on malicious code detection. Malicious code = writers=20 attempt to obfuscate the malicious code to subvert the malicious = code=20 detectors, such as anti-virus software. We tested the resilience = of three=20 commercial virus scanners against code-obfuscation attacks. The = results=20 were surprising: the three commercial virus scanners could be = subverted by=20 very simple obfuscation transformations! We present an = architecture for=20 detecting malicious patterns in executables that is resilient to = common=20 obfuscation transformations. Experimental results demonstrate the = efficacy=20 of our prototype tool, SAFE (a static analyzer = for=20 executables).


* This work was supported in part by the Office of Naval Research = under=20 contracts N00014-01-1-0796 and N00014-01-1-0708. The U.S. = Government is=20 authorized to reproduce and distribute reprints for Governmental = purposes,=20 notwithstanding any copyright notices affixed=20 thereon.
    The views and conclusions = contained=20 herein are those of the authors, and should not be interpreted as=20 necessarily representing the official policies or endorsements, = either=20 expressed or implied, of the above government agencies or the U.S. = Government.

1 Introduction

In the interconnected world of computers, malicious code has become = an=20 omnipresent and dangerous threat. Malicious code can infiltrate hosts = using a=20 variety of methods such as attacks against known software flaws, hidden=20 functionality in regular programs, and social engineering. Given the = devastating=20 effect malicious code has on our cyber infrastructure, identifying = malicious=20 programs is an important goal. Detecting the presence of malicious code = on a=20 given host is a crucial component of any defense mechanism.

Malicious code is usually classified [16]=20 according to its propagation method and goal into the following = categories:=20

Combining two or more of these malicious code categories can lead to = powerful=20 attack tools. For example, a worm can contain a payload that installs a = back=20 door to allow remote access. When the worm replicates to a new system = (via email=20 or other means), the back door is installed on that system, thus = providing an=20 attacker with a quick and easy way to gain access to a large set of = hosts.=20 Staniford et al. have demonstrated that worms can propagate = extremely=20 quickly through a network, and thus potentially cripple the entire cyber = infrastructure [1].=20 In a recent outbreak, the Sapphire/SQL Slammer worm reached the peak = infection=20 rate in about 10 minutes since launch, doubling every 8.5 seconds [32].=20 Once the back-door tool gains a large installed base, the attacker can = use the=20 compromised hosts to launch a coordinated attack, such as a distributed=20 denial-of-service (DDoS) attack [2].=20

In this paper, we develop a methodology for detecting malicious = patterns in=20 executables. Although our method is general, we have initially focused = our=20 attention on viruses. A computer virus replicates itself by inserting a = copy of=20 its code (the viral code) into a host program. When a user = executes the=20 infected program, the virus copy runs, infects more programs, and then = the=20 original program continues to execute. To the casual user, there is no = perceived=20 difference between the clean and the infected copies of a program until = the=20 virus activates its malicious payload.

The classic virus-detection techniques look for the presence of a=20 virus-specific sequence of instructions (called a virus = signature) inside=20 the program: if the signature is found, it is highly probable that the = program=20 is infected. For example, the Chernobyl/CIH virus is detected by = checking for=20 the hexadecimal sequence [5]:=20

E800 0000 005B 8D4B 4251 5050
0F01 4C24 FE5B 83C3 = 1CFA=20 8B2B=20

This corresponds to the following IA-32 instruction sequence, which=20 constitutes part of the virus body:

E8 00000000		     
call 0h
5B      pop ebx
8D 4B 42      lea ecx, [ebx + 42h]
51      push ecx
50      push eax
50      push eax
0F01 4C 24 FE      sidt [esp - 02h]
5B      pop ebx
83 C3 1C      add ebx, 1Ch
FA      cli
8B 2B
     mov ebp, [ebx]

This classic detection approach is effective when the virus code does = not=20 change significantly over time. Detection is also easier when viruses = originate=20 from the same source code, with only minor modifications and updates. = Thus, a=20 virus signature can be common to several virus variants. For example,=20 Chernobyl/CIH versions 1.2, 1.3, and 1.4 differ mainly in the trigger = date on=20 which the malicious code becomes active and can be effectively detected = by=20 scanning for a single signature, namely the one shown above.

The virus writers and the antivirus software developers are engaged = in an=20 obfuscation-deobfuscation game. Virus writers try to obfuscate = the=20 "vanilla" virus so that signatures used by the antivirus software cannot = detect=20 these "morphed" viruses. Therefore, to detect an obfuscated virus, the = virus=20 scanners first must undo the obfuscation transformations used by the = virus=20 writers. In this game, virus writers are obfuscators and researchers = working on=20 malicious code detection are deobfuscators. A method to detect malicious = code=20 should be resistant to common obfuscation transformations. This paper = introduces=20 such a method. The main contributions of this paper include:

• The obfuscation-deobfuscation game and attacks on = commercial virus=20 scanners
We view malicious code detection as an = obfuscation-deobfuscation=20 game between the virus writers and the researchers working to detect = malicious=20 code. Background on some common obfuscation techniques used by virus = writers is=20 given in Section=20 3. We also have developed an obfuscator for executables. = Surprisingly, the=20 three commercial virus scanners we considered could be easily thwarted = by simple=20 obfuscation transformations (Section=20 4). For example, in some cases the Norton antivirus scanner could = not even=20 detect insertions of nop instructions.

• A general architecture for detecting malicious patterns in = executables
We introduce a general architecture for detecting = malicious=20 patterns in executables. An overview of the architecture and its novel = features=20 is given in Section=20 5. External predicates and uninterpreted symbols are two important = elements=20 in our architecture. External predicates are used to summarize results = of=20 various static analyses, such as points-to and live-range analysis. We = allow=20 these external predicates to be referred in the abstraction patterns = that=20 describe the malicious code. Moreover, we allow uninterpreted symbols in = patterns, which makes the method resistant to renaming, a common = obfuscation=20 transformation. Two key components of our architecture, the program=20 annotator and the malicious code detector, are described in = Sections=20 6=20 and 7=20 respectively.

• Prototype for x86 executables
We have implemented a = prototype for=20 detecting malicious patterns in x86 executables. The tool is called a=20 static analyzer for executables or=20 SAFE. We have successfully tried SAFE on multiple viruses; for = brevity we=20 report on our experience with four specific viruses. Experimental = results (Section=20 8) demonstrate the efficacy of SAFE. There are several interesting=20 directions we intend to pursue as future work, which are summarized in = Section=20 9.

• Extensibility of analysis
SAFE depends heavily on = static analysis=20 techniques. As a result, the precision of the tool directly depends on = the=20 static analysis techniques that are integrated into it. In other words, = SAFE=20 is as good as the static analysis techniques it is built upon. For = example,=20 if SAFE uses the result of points-to analysis, it will be able to track = values=20 across memory references. In the absence of a points-to analyzer, SAFE = makes the=20 conservative assumption that a memory reference can access any memory = location=20 (i.e., everything points to everything). We have designed SAFE so that = various=20 static analysis techniques can be readily integrated into it. Several = simple=20 static analysis techniques are already implemented in SAFE.

2 Related Work

2.1 Theoretical Discussion

The theoretical limits of malicious code detection (specifically of = virus=20 detection) have been the focus of many researchers. Cohen [13]=20 and Chess-White [14]=20 showed that in general the problem of virus detection is undecidable. = Similarly,=20 several important static analysis problems are undecidable or = computationally=20 hard [31,=20 30].=20

However, the problem considered in this paper is slightly different = than the=20 one considered by Cohen [13]=20 and Chess-White [14].=20 Assume that we are given a vanilla virus V = which=20 contains a malicious sequence of instructions =A6=D3. Next we=20 are given an obfuscated version O(V) of the = virus. The=20 problem is to find whether there exists a sequence of instructions =A6=D3 in O(V) which is = "semantically=20 equivalent" to =A6=D3. A recent result by = Vadhan et=20 al. [25]=20 proves that in general program obfuscation is impossible. This leads us = to=20 believe that a computationally bounded adversary will not be able to = obfuscate a=20 virus to completely hide its malicious behavior. We will further explore = these=20 theoretical issues in the future.

2.2 Other Detection Techniques

Our work is closely related to previous results on static analysis = techniques=20 for verifying security properties of software [42,=20 47,=20 44= ,=20 39,=20 40,=20 48].=20 In a larger context, our work is similar to existing research on = software=20 verification [41,=20 = 45].=20 However, there are several important differences. First, viewing = malicious code=20 detection as an obfuscation-deobfuscation game is unique. The=20 obfuscation-deobfuscation viewpoint lead us to explore obfuscation = attacks upon=20 commercial virus scanners. Second, to our knowledge, all existing work = on static=20 analysis techniques for verifying security properties analyze source = code. On=20 the other hand, our analysis technique works on executables. In certain=20 contexts, such as virus detection, source code is not available. = Finally, we=20 believe that using uninterpreted variables in the specification of the = malicious=20 code is unique (Section=20 6.2).

Currie et al. looked at the problem of automatically checking = the=20 equivalence of DSP routines in the context of verifying the correctness = of=20 optimizing transformations [33].=20 Their approach is similar to ours, but they impose a set of simplifying=20 assumptions for their simulation tool to execute with reasonable = performance.=20 Feng and Hu took this approach one step further by using a theorem = prover to=20 determine when to unroll loops [34].=20 In both cases the scope of the problem is limited to VLIW or DSP code = and there=20 is limited support for user-specified analyses. Our work is applied to = x86=20 (IA-32) assembly and can take full advantage of static analyses = available to the=20 user of our SAFE tool. Necula adopts a similar approach based on = comparing a=20 transformed code sequence against the original code sequence in the = setting of=20 verifying the correctness of the GNU C compiler [35].=20 Using knowledge of the transformations performed by the compiler, = equivalence=20 between the compiler input and the compiler output is proven using a = simulation=20 relation. In our work, we require no a priori knowledge of the=20 obfuscation transformations performed, as it would be unrealistic to = expect such=20 information in the presence of malicious code.

We plan to enhance our framework by using the ideas from existing = work on=20 type systems for assembly code. We are currently investigating Morrisett = et=20 al.'s Typed Assembly Language [27,=20 26].=20 We apply a simple type system (Section=20 6) to the binaries we analyze by manually inserting the type = annotations. We=20 are unaware of a compiler that can produce Typed Assembly Language, and = thus we=20 plan to support external type annotations to enhance the power of our = static=20 analysis.

Dynamic monitoring can also be used for malicious code detection. = Cohen [13]=20 and Chess-White [14]=20 propose a virus detection model that executes code in a sandbox. Another = approach rewrites the binary to introduce checks driven by an = enforceable=20 security policy [43]=20 (known as the inline reference monitor or the IRM = approach). We=20 believe static analysis can be used to improve the efficiency of dynamic = analysis techniques, e.g., static analysis can remove redundant checks = in the=20 IRM framework. We construct our models for executables similar to the = work done=20 in specification-based monitoring [29,=20 28],=20 and apply our detection algorithm in a context-insensitive fashion. = Other=20 research used context-sensitive analysis by employing push-down systems = (PDSs).=20 Analyses described in [39,=20 40]=20 use the model checking algorithms for pushdown systems [38]=20 to verify security properties of programs. The data structures used in=20 interprocedural slicing [36],=20 interprocedural DFA [37],=20 and Boolean programs [41]=20 are hierarchically structured graphs and can be translated to pushdown = systems.=20

2.3 Other Obfuscators

While deciding on the initial obfuscation techniques to focus on, we = were=20 influenced by several existing tools. Mistfall (by z0mbie) = is a=20 library for binary obfuscation, specifically written to blend malicious = code=20 into a host program [9].=20 It can encrypt, morph, and blend the virus code into the host program. = Our=20 binary obfuscator is very similar to Mistfall. Unfortunately, we could = not=20 successfully morph binaries using Mistfall, so we could not perform a = direct=20 comparison between our obfuscator and Mistfall. Burneye (by = TESO)=20 is a Linux binary encapsulation tool. Burneye encrypts a binary = (possibly=20 multiple times), and packages it into a new binary with an extraction = tool [10].=20 In this paper, we have not considered encryption based obfuscation = techniques.=20 In the future, we will incorporate encryption based obfuscation = techniques into=20 our tool, by incorporating or extending existing libraries.

3 Background on Obfuscating Viruses

To detect obfuscated viruses, antivirus software have become more = complex.=20 This section discusses some common obfuscation transformations used by = virus=20 writers and how antivirus software have historically dealt with = obfuscated=20 viruses.

A polymorphic virus uses multiple techniques to prevent = signature=20 matching. First, the virus code is encrypted, and only a small in-clear = routine=20 is designed to decrypt the code before running the virus. When the = polymorphic=20 virus replicates itself by infecting another program, it encrypts the = virus body=20 with a newly-generated key, and it changes the decryption routine by = generating=20 new code for it. To obfuscate the decryption routine, several = transformations=20 are applied to it. These include: nop-insertion, code = transposition=20 (changing the order of instructions and placing jump instructions to = maintain=20 the original semantics), and register reassignment (permuting the = register=20 allocation). These transformations effectively change the virus = signature (Figure=20 1), inhibiting effective signature scanning by an antivirus tool. =

Original code
      Obfuscated code
E8 00000000    call 0h E8 00000000    call 0h
E8 00000000 call 0h E8 00000000 call 0h
5B pop ebx 5B pop ebx
8D 4B 42 lea ecx, [ebx + 42h] 8D 4B 42 lea ecx, [ebx + 45h]
51 push ecx 90 nop
50 push eax 51 push ecx
50 push eax 50 push eax
0F01 4C 24 FE sidt [esp - 02h] 50 push eax
5B pop ebx 90 nop
83 C3 1C add ebx, 1Ch 0F01 4C 24 FE sidt [esp - 02h]
FA cli 5B pop ebx
8B 2B mov ebp, [ebx] 83 C3 1C add ebx, 1Ch
 90 nop
FA cli
8B 2B mov ebp, [ebx]
Signature
 New signature
E800 0000 005B 8D4B 4251 5050 E800 0000 005B 8D4B 4290 5150
0F01 4C24 FE5B 83C3 1CFA 8B2B
 5090 0F01=20 4C24 FE5B 83C3 1C90
Figure 1: Original code and obfuscated = code from=20 Chernobyl/CIH, and their corresponding signatures. Newly added = instructions are=20 highlighted.

The obfuscated code in Figure=20 1 will behave in the same manner as before since the nop=20 instruction has no effect other than incrementing the program counter1.=20 However the signature has changed. Analysis can detect simple = obfuscations, like=20 nop-insertion, by using regular expressions instead of fixed=20 signatures. To catch nop insertions, the signature should allow = for any=20 number of nops at instruction boundaries (Figure=20 2). In fact, most modern antivirus software use regular expressions = as virus=20 signatures.

E800    0000    00(90)*    5B(90)*
8D4B 42(90)* 51(90)* 50(90)*
50(90)* 0F01 4C24 FE(90)*
5B(90)* 83C3 1C(90)* FA(90)*
8B2B
Figure 2: Extended signature to catch=20 nop-insertion.

Antivirus software deals with polymorphic viruses by performing = heuristic=20 analyses of the code (such as checking only certain program locations = for virus=20 code, as most polymorphic viruses attach themselves only at the = beginning or end=20 of the executable binary [18]),=20 and even emulating the program in a sandbox to catch the virus in action = [17].=20 The emulation technique is effective because at some point during the = execution=20 of the infected program, the virus body appears decrypted in main = memory, ready=20 for execution; the detection comes down to frequently scanning the = in-memory=20 image of the program for virus signatures while the program executes. =

Metamorphic viruses attempt to evade heuristic detection = techniques by=20 using more complex obfuscations. When they replicate, these viruses = change their=20 code in a variety of ways, such as code transposition, substitution of=20 equivalent instruction sequences, and register reassignment [11,=20 7].=20 Furthermore, they can "weave" the virus code into the host program, = making=20 detection by traditional heuristics almost impossible since the virus = code is=20 mixed with program code and the virus entry point is no longer at the = beginning=20 of the program (these are designated as entry point=20 obscuring (EPO) viruses [15]).=20

As virus writers employ more complex obfuscation techniques, = heuristic=20 virus-detection techniques are bound to fail. Therefore, there is = need to=20 perform a deeper analysis of malicious code based upon more = sophisticated=20 static-analysis techniques. In other words, inspection of the code = to detect=20 malicious patterns should use structures that are closer to the = semantics of the=20 code, as purely syntactic techniques, such as regular expression = matching, are=20 no longer adequate.

3.1 The Suite of Viruses

We have analyzed multiple viruses using our tool, and discuss four of = them in=20 this paper. Descriptions of these viruses are given below.

3.1.1 Detailed Description of the Viruses

Chernobyl (CIH)
According to the Symantec Antivirus Reseach = Center=20 (SARC), Chernobyl/CIH is a virus that infects 32-bit Windows = 95/98/NT=20 executable files [22].=20 When a user executes an infected program under Windows 95/98/ME, the = virus=20 becomes resident in memory. Once the virus is resident, CIH infects = other files=20 when they are accessed. Infected files may have the same size as the = original=20 files because of CIH's unique mode of infection: the virus searches for = empty,=20 unused spaces in the file2.=20 Next it breaks itself up into smaller pieces and inserts its code into = these=20 unused spaces. Chernobyl has two different payloads: the first one = overwrites=20 the hard disk with random data, starting at the beginning of the disk = (sector 0)=20 using an infinite loop. The second payload tries to cause permanent = damage to=20 the computer by corrupting the Flash BIOS.

zombie-6.b
The z0mbie-6.b virus includes an = interesting=20 feature -- the polymorphic engine hides every piece of the virus, and = the virus=20 code is added to the infected file as a chain of differently-sized = routines,=20 making standard signature detection techniques almost useless.

f0sf0r0
The f0sf0r0 virus uses a polymorphic engine = combined=20 with an EPO technique to hide its entry point. According to Kaspersky = Labs [21],=20 when an infected file is run and the virus code gains control, it = searches for=20 portable executable files in the system directories and infects them. = While=20 infecting, the virus encrypts itself with a polymorphic loop and writes = a result=20 to the end of the file. To gain control when the infected file is run, = the virus=20 does not modify the program's start address, but instead writes a = "jmp=20 <virus_entry>" instruction into the middle of the file.

Hare
Finally, the Hare virus infects the bootloader = sectors=20 of floppy disks and hard drives, as well as executable programs. When = the=20 payload is triggered, the virus overwrites random sectors on the hard = disk,=20 making the data inaccessible. The virus spreads by polymorphically = changing its=20 decryption routine and encrypting its main body.

The Hare and Chernobyl/CIH viruses are well known in the antivirus = community,=20 with their presence in the wild peaking in 1996 and 1998, respectively. = In spite=20 of this, we discovered that current commercial virus scanners could = not=20 detect slightly obfuscated versions of these viruses.

4 Obfuscation Attacks on Commercial Virus Scanners

We tested three commercial virus scanners against several common = obfuscation=20 transformations. To test the resilience of commercial virus scanners to = common=20 obfuscation transformations, we have developed an obfuscator for = binaries. Our=20 obfuscator supports four common obfuscation transformations: dead-code=20 insertion, code transposition, register reassignment, and instruction=20 substitution. While there are other generic obfuscation techniques [12,=20 23],=20 those described here seem to be preferred by malicious code writers, = possibly=20 because implementing them is easy and they add little to the memory = footprint.=20

4.1 Common Obfuscation Transformations

4.1.1 Dead-Code Insertion

Also known as trash insertion, dead-code insertion adds code = to a=20 program without modifying its behavior. Inserting a sequence of = nop=20 instructions is the simplest example. More interesting obfuscations = involve=20 constructing challenging code sequences that modify the program state, = only to=20 restore it immediately.

Some code sequences are designed to fool antivirus software that = solely rely=20 on signature matching as their detection mechanism. Other code sequences = are=20 complicated enough to make automatic analysis very time-consuming, if = not=20 impossible. For example, passing values through memory rather than = registers or=20 the stack requires accurate pointer analysis to recover values. The = example=20 shown in Figure=20 3 should clarify this. The code marked by (*) can be easily = eliminated by automated analysis. On the other hand, the second and = third=20 insertions, marked by (**), do cancel out but the analysis is = more=20 complex. Our obfuscator supports dead-code insertion.

Not all dead-code sequence can be detected and eliminated, as this = problem=20 reduces to program equivalence (i.e., is this code sequence = equivalent to an=20 empty program?), which is undecidable. We believe that many common = dead-code=20 sequences can be detected and eliminated with acceptable performance. To = quote=20 the documentation of the RPME virus permutation engine [8],=20

[T]rash [does not make the] program more complex [...]. If = [the]=20 detecting algorithm will be written such as I think, then there is no=20 difference between NOP and more complex trash.

Our detection tool, SAFE, identifies several kinds of such dead-code=20 segments.

Original code
       Code obfuscated through
dead-code insertion
    Code obfuscated through
code transposition
       Code obfuscated through
instruction substitution
call 0h call 0h call 0h call 0h
pop ebx pop ebx pop ebx pop ebx
lea ecx, [ebx+42h] lea ecx, [ebx+42h] jmp S2 lea ecx, [ebx+42h]
push ecx nop (*) S3: push eax sub esp, 03h
push eax nop (*) push eax sidt [esp - 02h]
push eax push ecx sidt [esp - 02h] add [esp], 1Ch
sidt [esp - 02h] push eax jmp S4 mov ebx, [esp]
pop ebx inc eax (**) add ebx, 1Ch inc esp
add ebx, 1Ch push eax jmp S6 cli
cli dec [esp - 0h] (**) S2: lea ecx, [ebx+42h ] mov ebp, [ebx]
mov ebp, [ebx] dec eax (**) push ecx
sidt [esp - 02h] jmp S3
pop ebx S4: pop ebx
add ebx, 1Ch cli
cli jmp S5
mov ebp, [ebx] S5: mov ebp, [ebx]
Figure 3: Examples of obfuscation = through=20 dead-code insertion, code transposition, and instruction substitution. = Newly=20 added instructions are highlighted.

4.1.2 Code Transposition

Code transposition shuffles the instructions so that the order in the = binary=20 image is different from the execution order, or from the order of = instructions=20 assumed in the signature used by the antivirus software. To achieve the = first=20 variation, we randomly reorder the instructions and insert unconditional = branches or jumps to restore the original control-flow. The = second=20 variation swaps instructions if they are not interdependent, similar to = compiler=20 code generation, but with the different goal of randomizing the = instruction=20 stream.

The two versions of this obfuscation technique differ in their = complexity.=20 The code transposition technique based upon unconditional branches is = relatively=20 easy to implement. The second technique that interchanges independent=20 instructions is more complicated because the independence of = instructions must=20 be ascertained. On the analysis side, code transposition can complicate = matters=20 only for a human. Most automatic analysis tools (including ours) use an=20 intermediate representation, such as the control flow graph (CFG) or the = program=20 dependence graph (PDG) [36],=20 that is not sensitive to superfluous changes in control flow. Note that = an=20 optimizer acts as a deobfuscator in this case by finding the unnecessary = unconditional branches and removing them from the program code. = Currently, our=20 obfuscator supports only code transposition based upon inserting = unconditional=20 branches.

4.1.3 Register Reassignment

The register reassignment transformation replaces usage of one = register with=20 another in a specific live range. This technique exchanges register = names and=20 has no other effect on program behavior. For example, if register = ebx=20 is dead throughout a given live range of the register eax, it = can=20 replace eax in that live range. In certain cases, register = reassignment=20 requires insertion of prologue and epilogue code around the live range = to=20 restore the state of various registers. Our binary obfuscator supports = this code=20 transformation.

The purpose of this transformation is to subvert the antivirus = software=20 analyses that rely upon signature-matching. There is no real obfuscatory = value=20 gained in this process. Conceptually, the deobfuscation challenge is = equally=20 complex before or after the register reassignment.

4.1.4 Instruction Substitution

This obfuscation technique uses a dictionary of equivalent = instruction=20 sequences to replace one instruction sequence with another. Since this=20 transformation relies upon human knowledge of equivalent instructions, = it poses=20 the toughest challenge for automatic detection of malicious code. The = IA-32=20 instruction set is especially rich, and provides several ways of = performing the=20 same operation. Coupled with several architecturally ambivalent features = (e.g.,=20 a memory-based stack that can be accessed both as a stack using = dedicated=20 instructions and as a memory area using standard memory operations), the = IA-32=20 assembly language provides ample opportunity for instruction = substitution.

To handle obfuscation based upon instruction substitution, an = analysis tool=20 must maintain a dictionary of equivalent instruction sequences, similar = to the=20 dictionary used to generate them. This is not a comprehensive solution, = but it=20 can cope with the common cases. In the case of IA-32, the problem can be = slightly simplified by using a simple intermediate language that = "unwinds" the=20 complex operations corresponding to each IA-32 instruction. In some = cases, a=20 theorem prover such as Simplify [49]=20 or PVS [46] can=20 also be used to prove that two sequences of instructions are equivalent. =

4.2 Testing Commercial Antivirus Tools

We tested three commercial virus scanners using obfuscated versions = of the=20 four viruses described earlier. The results were quite surprising: a=20 combination of nop-insertion and code transposition was enough = to=20 create obfuscated versions of the viruses that the commercial virus = scanners=20 could not detect. Moreover, the Norton antivirus software could not = detect=20 an obfuscated version of the Chernobyl virus using just = nop-insertions.=20 SAFE was resistant to the two obfuscation transformations. The results = are=20 summarized in Table=20 1. A =A1=CC indicates that the antivirus software detected the = virus. A =A1=C1 means=20 that the software did not detect the virus. Note that unobfuscated = versions of=20 all four viruses were detected by all the tools.

Norton®
Antivirus
7.0		 McAfee®
VirusScan
6.01		 Command®
Antivirus
4.61.2		 SAFE
Chernobyl original =A1=CC =A1=CC =A1=CC =A1=CC
obfuscated =A1=C1[1] =A1=C1[1,2] =A1=C1[1,2] =A1=CC
z0mbie-6.b original =A1=CC =A1=CC =A1=CC =A1=CC
obfuscated =A1=C1[1,2] =A1=C1[1,2] =A1=C1[1,2] =A1=CC
f0sf0r0 original =A1=CC =A1=CC =A1=CC =A1=CC
obfuscated =A1=C1[1,2] =A1=C1[1,2] =A1=C1[1,2] =A1=CC
Hare original =A1=CC =A1=CC =A1=CC =A1=CC
obfuscated =A1=C1[1,2] =A1=C1[1,2] =A1=C1[1,2] =A1=CC

Obfuscations considered: [1] =3D nop-insertion (a form = of=20 dead-code insertion)
[2] =3D code=20 transposition

Table 1: Results of testing = various=20 virus scanners on obfuscated viruses.

5 Architecture

This section gives an overview of the architecture of SAFE (Figure=20 4). Subsequent sections provide detailed descriptions of the major=20 components of SAFE.

=20

Figure 4: Architecture of the static analyzer for executables = (SAFE).

To detect malicious patterns in executables, we build an abstract=20 representation of the malicious code (here a virus). The abstract = representation=20 is the "generalization" of the malicious code, e.g., it incorporates = obfuscation=20 transformations, such as superfluous changes in control flow and = register=20 reassignments. Similarly, one must construct an abstract representation = of the=20 executable in which we are trying to find a malicious pattern. Once the=20 generalization of the malicious code and the abstract representation of = the=20 executable are created, we can then detect the malicious code in the = executable.=20 We now describe each component of SAFE.

Generalizing the malicious code: Building the malicious code=20 automaton
The malicious code is generalized into an automaton = with=20 uninterpreted symbols. Uninterpreted symbols (Section=20 6.2) provide a generic way of representing data dependencies between = variables without specifically referring to the storage location of each = variable.

Pattern-definition loader
This component takes a library of = abstraction patterns and creates an internal representation. = These=20 abstraction patterns are used as alphabet symbols by the malicious code=20 automaton.

The executable loader
This component transforms the = executable into=20 an internal representation, here the collection of control flow graphs = (CFGs),=20 one for each program procedure. The executable loader (Figure=20 5) uses two off-the-shelf components, IDA Pro and = CodeSurfer.=20 IDA Pro (by DataRescue [20])=20 is a commercial interactive disassembler. CodeSurfer (by = GrammaTech, Inc.=20 [19])=20 is a program-understanding tool that performs a variety of static = analyses.=20 CodeSurfer provides an API for access to various structures, such as the = CFGs=20 and the call graph, and to results of a variety of static analyses, such = as=20 points-to analysis. In collaboration with GrammaTech, we have developed = a=20 connector that transforms IDA Pro internal structures into an = intermediate form=20 that CodeSurfer can parse.

=20

Figure 5: Implementation of executable loader module.

The annotator
This component inputs a CFG from the = executable and=20 the set of abstraction patterns and produces an annotated CFG, the = abstract=20 representation of a program procedure. The annotated CFG includes = information=20 that indicates where a specific abstraction pattern was found in the = executable.=20 The annotator runs for each procedure in the program, transforming each = CFG. Section=20 6 describes the annotator in detail.

The detector
This component computes whether the malicious = code=20 (represented by the malicious code automaton) appears in the abstract=20 representation of the executable (created by the annotator). This = component uses=20 an algorithm based upon language containment and unification. Details = can be=20 found in Section=20 7.

Throughout the rest of the paper, the malicious code fragment shown = in Figure=20 6 is used as a running example. This code fragment was extracted = from the=20 Chernobyl virus version 1.4.

Original=20 Code
WVCTF:
mov eax, dr1
mov ebx, [eax+10h]
mov edi, [eax]
LOWVCTF:
pop ecx
jecxz SFMM
mov esi, ecx
mov eax, 0d601h
pop edx
pop ecx
call edi
jmp LOWVCTF
SFMM:
pop ebx
pop eax
stc
pushf
Figure 6: Original code fragment from = Chernobyl virus=20 version 1.4.

To obtain the obfuscated code fragment depicted (Figure=20 7), we applied the following obfuscation transformations: dead-code=20 insertion, code transposition, and register reassignment. Incidentally, = the=20 three commercial antivirus software (Norton, McAfee, and Command) = detected the=20 original code fragment shown. However, the obfuscated version was not = detected=20 by any of the three commercial antivirus software.

Obfuscated=20 Code
WVCTF:
mov eax, dr1
jmp Loc1
Loc2:
mov edi, [eax]
LOWVCTF:
pop ecx
jecxz SFMM
nop
mov esi, ecx
nop
nop
mov eax, 0d601h
jmp Loc3
Loc1:
mov ebx, [eax+10h]
jmp Loc2
Loc3:
pop edx
pop ecx
nop
call edi
jmp LOWVCTF
SFMM:
pop ebx
pop eax
push eax
pop eax
stc
pushf
Figure 7: Obfuscated version based upon = code in Figure=20 6.

6 Program Annotator

This section describes the program annotator in detail and the data=20 structures and static analysis concepts used in the detection algorithm. = The=20 program annotator inputs the CFG of the executable and a set of = abstraction=20 patterns and outputs an annotated CFG. The annotated CFG associates with = each=20 node n in the CFG a set of patterns that match = the=20 program at the point corresponding to the node n. The=20 precise syntax for an abstraction pattern and the semantics of matching = are=20 provided later in the section.

Figure=20 8 shows the CFG and a simple annotated CFG corresponding to the = obfuscated=20 code from Figure=20 7. Note that one node in the annotated CFG can correspond to several = nodes=20 in the original CFG. For example, the nodes annotated with = "IrrelevantInstr"=20 corresponds to one or more nop instructions.

The annotations that appear in Figure=20 8 seem intuitive, but formulating them within a static-analysis = framework=20 requires formal definitions. We enhance the SAFE framework with a type = system=20 for x86 based on the typestate system described in [6].=20 However, other type systems designed for assembly languages, such as = Typed=20 Assembly Language [27,=20 26],=20 could be used in the SAFE framework. Definitions, patterns, and the = matching=20 procedure are described in Sections 6.1,=20 6.2=20 and 6.3=20 respectively.


Figure=20 8: Control flow graph of obfuscated code fragment, and annotations. =

6.1 Basic Definitions

This section provides the formal definitions used in the rest of the = paper.=20

Program Points
An instruction I is a=20 function application, I : =A6=D31 =A1=C1 = =A1=AD =A1=C1 =A6=D3k =A1=FA=20 =A6=D3. While the type system does not preclude higher-order = functions or=20 function composition, it is important to note that most assembly = languages=20 (including x86) do not support these concepts. A program P is a sequence of instructions <I1, =A1=AD, IN>. During = program=20 execution, the instructions are processed in the sequential order they = appear in=20 the program, with the exception of control-flow instructions that can = change the=20 sequential execution order. The index of the instruction in the program = sequence=20 is called a program point (or program counter), denoted by = the=20 function pc : {I1, =A1=AD, IN} = =A1=FA [1, =A1=AD,=20 N], and defined as pc(Ij) =3D j, = ∀ 1 =A1=DC j =A1=DC=20 N. The set of all program points for program P is=20 ProgramPoints(P) =3D {1, =A1=AD, N}. The pc function provides a total ordering over the set = of program=20 instructions.

Control Flow Graph
A basic block B=20 is a maximal sequence of instructions <Il, =A1=AD,=20 Im> that contains at most one control-flow = instruction,=20 which must appear at the end. Thus, the execution within a basic block = is by=20 definition sequential. Let V be the set of = basic blocks=20 for a program P, and E = ⊆ V =A1=C1 V =A1=C1 { T, F=20 } be the set of control flow transitions between basic blocks. = Each edge=20 is marked with either T or F=20 corresponding to the condition (true or false) on which = that edge=20 is followed. Unconditional jumps have outgoing edges always marked with = T. The directed graph CFG( P ) = =3D <V, E=20 > is called the control flow graph.

Predicates
Predicates are the mechanism by which we = incorporate=20 results of various static analyses such as live range and points-to = analysis.=20 These predicates can be used in the definition of abstraction patterns. = Table=20 2 lists predicates that are currently available in our system. For = example,=20 code between two program points p1 = and p2 can be verified as dead-code (Section=20 4.1.1) by checking that for every variable m that is=20 live in the program range [p1,=20 p2], its value at point p2=20 is the same as its value at point p1. The=20 change in m's value between two program points = p1 and p2 is=20 denoted by Delta(m, p1, = p2) and=20 can be implemented using polyhedral analysis [24].=20

Dominators( B ) the set of basic blocks that dominate the basic block=20 B
PostDominators( B ) the set of basic blocks that are dominated by the basic = block=20 B
Pred( B ) the set of basic blocks that immediately precede = B
Succ( B ) the set of basic blocks that immediately follow = B
First( B ) the first instruction of the basic block = B
Last( B ) the last instruction of the basic block B
Previous( I )
=A1=C8B' =A1=CA Pred( BI ) = Last( B'=20 ) if I =3D First( BI )
-or-
I' if BI =3D < =A1=AD, I', I, =A1=AD=20 >
Next( I )
=A1=C8B' =A1=CA Succ( BI ) = First( B'=20 ) if I =3D Last( BI )
-or-
I' if BI =3D < =A1=AD, I, I', =A1=AD=20 >
Kills( p, a ) true if the instruction at program point p = kills=20 variable a
Uses( p, a ) true if the instruction at program point p = uses=20 variable a
Alias( p, x, y ) true if variable x is an alias for y = at=20 program point p
LiveRangeStart( p, a ) the set of program points that start the a's live = range=20 that includes p
LiveRangeEnd( p, a ) the set of program points that end the a's live = range=20 that includes p
Delta( p, m, n ) the difference between integer variables m and = n=20 at program point p
Delta( m, p_1, p_2 ) the change in m's value between program points=20 p1 and p2
PointsTo( p, x, a ) true if variable x points to location of = a=20 at program point=20 p

Table 2: = Examples of static analysis predicates.

Explanations of the static analysis predicates shown in Table=20 2 are standard and can be found in a compiler textbook (such as [3]).=20

Instructions and Data Types
The type constructors build = upon simple=20 integer types (listed below as the ground class of types), and = allow for=20 array types (with two variations: the pointer-to-start-of-array type and = the=20 pointer-to-middle-of-array type), structures and unions, pointers, and=20 functions. Two special types =A1=CD(n) and = T(n) = complete the=20 type system lattice. =A1=CD(n) and T(n) represent types = that are=20 stored on n bits, with =A1=CD(n)=20 being the least specific ("any") type and T(n) being the most = specific type.=20 Table=20 3 describes the constructors allowed in our type system.

int(g:s:v= )=20 |=20 = uint(g:s:v)=20 | =A1=AD =
=A6=D3 :: ground Ground types
| =A6=D3[n] Pointer to the base of an array of type =A6=D3 = and of=20 size n
| =A6=D3(n] Pointer into the middle of an array of type =A6=D3 = and of=20 size n
| =A6=D3 ptr Pointer to =A6=D3
| s { =A6=CC1, =A1=AD, = =A6=CCk } Structure (product of types of =A6=CC=20 i)
| u{=A6=CC1, =A1=AD, = =A6=CCk} Union
| =A6=D31 =A1=C1 =A1=AD =A1=C1 =A6=D3k = =A1=FA =A6=D3 Function
| T(n) Top type of n bits
| =A1=CD(n) Bottom type of n bits (type "any" of n=20 bits)
=A6=CC :: (l, =A6=D3 i) Member labeled l of type =A6=D3 at = offset=20 i
ground ::

Table 3: A = simple=20 type system.

The type =A6=CC(l, =A6=D3, i) represents = the type of a field=20 member of a structure. The field has a type =A6=D3=20 (independent of the types of all other fields in the same structure), an = offset=20 i that uniquely determines the location of the = field=20 within the structure, and a label l that = identifies the=20 field within the structure (in some cases this label might be = undefined).

Physical subtyping takes into account the layout of values in memory = [4,=20 6].=20 If a type =A6=D3 is a physical subtype = of =A6=D3' (denoted it by =A6=D3 = =A1=DC =A6=D3'), then the=20 memory layout of a value of type =A6=D3' is a = prefix of the=20 memory layout of a value of type =A6=D3. We = will not describe=20 the rules of physical subtyping here as we refer the reader to Xu's = thesis [6]=20 for a detailed account of the typestate system (including subtyping = rules).

The type int(g:s:v) represents a signed = integer, and=20 it covers a wide variety of values within storage locations. It is = parametrized=20 using three parameters as follows: g = represents the=20 number of highest bits that are ignored, s is = the number=20 of middle bits that represent the sign, and v = is the=20 number of lowest bits that represent the value. Thus the type int(g:s:v) uses a total of g + s = + v=20 bits.

=20

The type uint(g:s:v) represents an unsigned = integer,=20 and it is just a variation of int(g:s:v), with = the=20 middle s sign bits always set to zero.

The notation int(g:s:v) allows for the = separation of=20 the data and storage location type. In most assembly languages, it is = possible=20 to use a storage location larger than that required by the data type = stored in=20 it. For example, if a byte is stored right-aligned in a (32-bit) word, = its=20 associated type is int(24:1:7). This means = that an=20 instruction such as xor on least significant byte within 32-bit = word will=20 preserve the leftmost 24 bits of the 32-bit word, even though the = instruction=20 addresses the memory on 32-bit word boundary.

This separation between data and storage location raises the issue of = alignment information, i.e., most computer systems require or prefer = data to be=20 at a memory address aligned to the data size. For example, 32-bit = integers=20 should be aligned on 4-byte boundaries, with the drawback that accessing = an=20 unaligned 32-bit integer leads to either a slowdown (due to several = aligned=20 memory accesses) or an exception that requires handling in software. = Presently,=20 we do not use alignment information as it does not seem to provide a = significant=20 covert way of changing the program flow.

Code
      Type
call 0h
pop ebx ebx : =A1=CD(32) =
lea ecx, [ebx + 42h] ecx : =A1=CD(32), =
ebx : ptr =A1=CD(32) =
push ecx ecx : =A1=CD(32) =
push eax eax : =A1=CD(32) =
push eax eax : =A1=CD(32) =
sidt [esp - 02h]
pop ebx eax : =A1=CD(32) =
add ebx, 1Ch ebx : int(0:1:31
cli
mov ebp, [ebx] ebp : =A1=CD(32), =

ebx : ptr =A1=CD(32)

Figure 9: Inferred types from = Chernobyl/CIH virus=20 code.

Figure=20 9 shows the types for operands in a section of code from the = Chernobyl/CIH=20 virus. Table=20 4 illustrates the type system for Intel IA-32 architecture. There = are other=20 IA-32 data types that are not covered in Table=20 4, including bit strings, byte strings, 64- and 128-bit packed SIMD = types,=20 and BCD and packed BCD formats. The IA-32 logical address is a = combination of a=20 16-bit segment selector and a 32-bit segment offset, thus its type is = the cross=20 product of a 16-bit unsigned integer and a 32-bit pointer.

IA-32 Datatype Type Expression
byte unsigned int uint(0:0:8)
word unsigned int uint(0:0:16)
doubleword unsigned int uint(0:0:32)
quadword unsigned int uint(0:0:64)
double quadword unsigned int uint(0:0:128)
byte signed int int(0:1:7)
word signed int int(0:1:15)
doubleword signed int int(0:1:31)
quadword signed int int(0:1:63)
double quadword signed int int(0:1:127)
single precision float float(0:1:31)
double precision float float(0:1:63)
double extended precision float float(0:1:79)
near pointer =A1=CD(32)
far pointer (logical address) uint(0:0:16) =A1=C1 uint(0:0:32) =A1=FA =A1=CD(48)
eax, ebx, ecx, edx =A1=CD(32)
esi, edi, ebp, esp =A1=CD(32)
eip int(0:1:31)
cs, ds, ss, es, fs, gs =A1=CD(16)
ax, bx, cx, dx =A1=CD(16)
al, bl, cl, dl =A1=CD(8)
ah, bh, ch, dh =A1=CD(8)

Table 4: IA-32 datatypes and their = corresponding=20 expression in the type system from Table=20 3.

6.2 Abstraction Patterns

An abstraction pattern =A6=A3 is a 3-tuple = (V, O, C), where V is a = list of typed=20 variables, O is a sequence of instructions, = and C is a boolean expression combining one or more = static=20 analysis predicates over program points. Formally, a pattern =A6=A3=3D(V, O, C) is a 3-tuple defined as follows: =

V =3D {x1 : =A6=D31, =A1=AD, xk :=20 =A6=D3k}
O =3D <I(v1, =A1=AD, vm) | I : = =A6=D31 =A1=C1 =A1=AD =A1=C1=20 =A6=D3m =A1=FA =A6=D3>
C =3D boolean expression = involving=20 static
analysis predicates and logical operators =

An instruction from the sequence O has a = number of=20 arguments (vi)i=A1=DD0, = where each=20 argument is either a literal value or a free variable xj. We write =A6=A3(x1 :=20 =A6=D31, =A1=AD, xk : =A6=D3k) = to denote the pattern=20 =A6=A3 =3D (V, O, C) with free variables x1, =A1=AD, xk. An example of = a pattern is=20 shown below.

=20

This pattern represents two instructions that pop a register X off the stack and then add a constant value to it=20 (0x03AF). Note the use of uninterpreted symbol X in the pattern. Use of the uninterpreted symbols = in a=20 pattern allows it to match multiple sequences of instructions, e.g., the = patterns shown above matches any instantiation of the pattern where = X is assigned a specific register. The type int(0:1:31) of X = represents an integer=20 with 31 bits of storage and one sign bit.

We define a binding B as a set of = pairs [variable v, value x]. Formally, a binding B is defined as { [x,v] | x = =A1=CA V, x : =A6=D3, v :=20 =A6=D3', =A6=D3 =A1=DC =A6=D3' }. If a pair [x, v] occurs in a=20 binding B, then we write B(x) =3D=20 v. Two bindings B1 and B2 are said to be compatible if = they do not=20 bind the same variable to different values:

Compatible(B1, = B2) =3D ∀ x =A1=CA V=20 . ( [x, y1] =A1=CA B1 =A1=C4 [x, y2] = =A1=CA B2 ) ⇒=20 ( y1 =3D y2 )

The union of two compatible bindings B1 and B2=20 includes all the pairs from both bindings. For incompatible bindings, = the union=20 operation returns an empty binding.

=20

When matching an abstraction pattern against a sequence of = instructions, we=20 use unification to bind the free variables of =A6=A3 to=20 actual values. The function:

Unify( <=A1=AD, opi( xi,1, =A1=AD,=20 xi,ni ), =A1=AD>1 =A1=DC i =A1=DC m, = =A6=A3 )=20

returns a ``most general'' binding B=20 if the instruction sequence <=A1=AD, = opi(=20 xi,1, =A1=AD, xi,ni ), =A1=AD>1 = =A1=DC i =A1=DC=20 m can be unified with the sequence of instructions O specified in the pattern =A6=A3. If the=20 two instruction sequences cannot be unified, Unify=20 returns false. Definitions and algorithms related to unification = are=20 standard and can be found in [= 50].3=20

6.3 Annotator Operation

The annotator associates a set of matching patterns with each node in = the=20 CFG. The annotated CFG of a program procedure P with=20 respect to a set of patterns =A6=B2 is denoted = by P=A6=B2. Assume that a node n in=20 the CFG corresponds to the program point p and = the=20 instruction at p is Ip. The annotator attempts to match the = (possibly=20 interprocedural) instruction sequence S(n) =3D = <=A1=AD,=20 Previous2(Ip), Previous(Ip),=20 Ip> with the patterns in the set =A6=B2 =3D=20 {=A6=A31, =A1=AD, =A6=A3m}. The CFG node = n=20 is then labeled with the list of pairs of patterns and bindings that = satisfy the=20 following condition:

Annotation(n) =3D { [=A6=A3, B] : = =A6=A3 =A1=CA {=A6=A31, =A1=AD,=20 =A6=A3m} =A1=C4 B =3D Unify( S(n), =A6=A3 ) }

If Unify( S(n), =A6=A3 ) returns = false (because=20 unification is not possible), then the node n = is not=20 annotated with [=A6=A3, B]. Note that a = pattern =A6=A3 might appear several times (albeit with = different bindings)=20 in Annotation(n). However, the pair [=A6=A3,=20 B] is unique in the annotation set of a given node.

7 Detector

The detector takes as its inputs an annotated CFG for an executable = program=20 procedure and a malicious code automaton. If the malicious pattern = described by=20 the malicious code automaton is also found in the annotated CFG, the = detector=20 returns the sequence of instructions exhibiting the pattern. The = detector=20 returns no if the malicious pattern cannot be found in the = annotated CFG.=20

7.1 The Malicious-Code Automaton

Intuitively, the malicious code automaton is a generalization of the = vanilla=20 virus, i.e., the malicious code automaton also represents obfuscated = strains of=20 the virus. Formally, a malicious code automaton (or MCA) = A is a 6-tuple (V, =A6=B2, S, = =A6=C4, S0,=20 F), where

  • V =3D { v1 : =A6=D31 , = =A1=AD, vk :=20 =A6=D3k } is a set of typed variables,=20
  • =A6=B2 =3D {=A6=A31, =A1=AD, = =A6=A3n} is a=20 finite alphabet of patterns parametrized by variables from = V, for 1 =A1=DC i =A1=DC = n, Pi =3D (Vi, Oi,=20 Ci) where Vi ⊆ = V,=20
  • S is a finite set of states,=20
  • =A6=C4 : S =A1=C1 =A6=B2 =A1=FA = 2S is a transition=20 function,=20
  • S0 ⊆ S is a non-empty = set of initial=20 states,=20
  • F ⊆ S is a non-empty set of = final states.=20

An MCA is a generalization of an ordinary finite-state automaton in = which the=20 alphabets are a finite set of patterns defined over a set of typed = variables.=20 Given a binding B for the variables V=3D{v1, =A1=AD, vk }, the = finite-state=20 automaton obtained by substituting B(vi) for=20 vi for all 1 = =A1=DC i =A1=DC k=20 in A is denoted by B(A). Note=20 that B(A) is a "vanilla" finite-state = automaton. We=20 explain this using an example. Consider the MCA A shown=20 in Figure=20 10 with V=3D{A, B, C, D}. The automata = obtained from=20 A corresponding to the bindings B1 and B2 are=20 shown in Figure=20 10. The uninterpreted variables in the MCA were introduced to handle = obfuscation transformations based on register reassignment. The = malicious code=20 automaton corresponding to the code fragment shown in Figure=20 6 (from the Chernobyl virus) is depicted in Figure=20 11.

  
mov esi, ecx
mov eax, 0d601h
pop edx
pop ecx
B1 =3D { [A, esi],
[B, ecx],
[C, eax],
[D, edx] }

mov esi, eax
mov ebx, 0d601h
pop ecx
pop eax
B2 =3D { [A, esi],
[B, eax],
[C, ebx],
[D, ecx] }

Figure 10: = Malicious code=20 automaton for a Chernobyl virus code fragment, and instantiations with = different=20 register assignments, shown with their respective bindings.

=20

Figure 11: Malicious code automaton corresponding to code = fragment=20 from Figure=20 6.

7.2 Detector Operation

The detector takes as its inputs the annotated CFG P=A6=B2 of a program procedure P=20 and a malicious code automaton MCA A=3D(V, =A6=B2, = S,=A6=C4, S0,=20 F). Note that the set of patterns =A6=B2 is used both=20 to construct the annotated CFG and as the alphabet of the malicious code = automaton. Intuitively, the detector determines whether there exists a = malicious=20 pattern that occurs in A and P=A6=B2. We formalize this intuitive = notion. The=20 annotated CFG P=A6=B2 is a = finite-state automaton=20 where nodes are states, edges represent transitions, the node = corresponding to=20 the entry point is the initial state, and every node is a final state. = Our=20 detector determines whether the following language is empty:


In the expression given above, L(P=A6=B2) is the language corresponding = to the=20 annotated CFG and BAll is = the set of=20 all bindings to the variables in the set V. In = other=20 words, the detector determines whether there exists a binding B such that the intersection of the languages P=A6=B2 and B(A) is non-empty.=20

Our detection algorithm is very similar to the classic algorithm for=20 determining whether the intersection of two regular languages is = non-empty [51].=20 However, due to the presence of variables, we must perform unification = during=20 the algorithm. Our algorithm (Figure=20 12) combines the classic algorithm for computing the intersection of = two=20 regular languages with unification. We have implemented the algorithm as = a=20 data-flow analysis.

  • For each node n of the annotated CFG = PA we associate pre and post lists = Lnpre and Lnpost respectively. Each = element of a=20 list is a pair [s,B], where s=20 is the state of the MCA A and B is the binding of variables. Intuitively, if = [s,B] =A1=CA Lnpre, then it = is possible for=20 A with the binding B (i.e. for=20 B(A)) to be in state s just=20 before node n.=20
  • Initial condition: Initially, both lists associated with = all nodes=20 except the start node n0 are = empty. The pre=20 list associated with the start node is the list of all pairs [s, ∅], where s is = an initial state=20 of the MCA A, and the post list associated = with the=20 start node is empty.=20
  • The do-until loop: The do-until loop updates the pre and = post lists=20 of all the nodes. At the end of the loop, the worklist WS contains the set of nodes whose pre or = post=20 information has changed. The loop executes until the pre and post = information=20 associated with the nodes does not change, and a fixed point is = reached. The=20 join operation that computes Lipre takes the list of = state-binding=20 pairs from all of the Ljpost=20 sets for program points preceding i and = copies them to=20 Lipre only if there = are no=20 repeated states. In case of repeated states, the conflicting pairs are = merged=20 into a single pair only if the bindings are compatible. If the = bindings are=20 incompatible, both pairs are thrown out.=20
  • Diagnostic feedback: Suppose our algorithm returns a = non-empty set,=20 meaning a malicious pattern is common to the annotated CFG P=A6=B2 and MCA A. In this=20 case, we return the sequence of instructions in the executable = corresponding=20 to the malicious pattern. This is achieved by keeping an additional = structure=20 with the algorithm. Every time the post list for a node n is updated by taking a transition in A (see the statement 14 = in Figure=20 12), we store the predecessor of the added state, i.e., if [=A6=C4(s, =A6=A3), Bs =A1=C8 B] is = added to Lnpost, then we add an edge = from s to =A6=C4(s, =A6=A3) = (along with the binding=20 Bs =A1=C8 B) in the associated = structure.=20 Suppose we detect that Lnpost=20 contains a state [s, Bs], where = s is a final state of the MCA A.=20 Then we traceback the associated structure from s=20 until we reach an initial state of A = (storing the=20 instructions occurring along the way).
Input: A list of patterns =A6=B2 =3D = {P1,=20 =A1=AD, Pr}, a malicious code automaton A =3D=20 (V, =A6=B2, S, =A6=C4, S0, F), and an annotated = CFG P=A6=B2 =3D <N, E>. =
Output:=20 true if the program is likely infected, false = otherwise.=20

MaliciousCodeChecking(=A6=B2, A, P=A6=B2)=20
(1)     Lpren0 = =A1=FB { [s,=20 ∅] | s =A1=CA S0 }, where n0 =A1=CA N is the entry node = of P=A6=B2
(2)     foreach n =A1=CA N \ { = n0 }=20 do Lnpre = =A1=FB ∅=20
(3)     foreach n =A1=CA N = do Lnpost =A1=FB = ∅
(4)     WS =A1=FB ∅ =
(5)     do
(6)            WSold =A1=FB WS =
(7)            WS=20 =A1=FB ∅
(8)            foreach n =A1=CA N     // update pre = information
(9)           =20        if Lnpre =A1=D9 =A1=C8m = =A1=CA Previous(n)=20 Lmpost
(10)           =20       =20        Lnpre =A1=FB =A1=C8m = =A1=CA Previous(n)=20 Lmpost
(11)           =20       =20        WS =A1=FB=20 WS =A1=C8 { n }
(12)            foreach n =A1=CA N     // update post = information
(13)           =20        NewLnpost =A1=FB = ∅
(14)           =20        foreach [s, Bs] =A1=CA = Lnpre=20
(15)           =20       =20        foreach [=A6=A3, B] =A1=CA Annotation(n) =A1=C4 = Compatible(Bs,=20 B)     // follow a transition
(16)           =20       =20       =20        add [=A6=C4(=20 s, =A6=A3 ), Bs =A1=C8 B] to NewLnpost
(17)           =20        if Lnpost =A1=D9=20 NewLnpost
(18)           =20       =20        Lnpost =A1=FB=20 NewLnpost
(19)           =20       =20        WS =A1=FB=20 WS =A1=C8 { n }
(20)     until WS =3D = ∅
(21)     return ∃ n =A1=CA N . = ∃ [s, Bs] =A1=CA=20 Lnpost . s =A1=CA F=20

Figure 12: = Algorithm to=20 check a program model against a malicious code specification.

8 Experimental Data

The three major goals of our experiments were to measure the = execution time=20 of our tool and find the false positive and negative rates. We = constructed ten=20 obfuscated versions of the four viruses. Let Vi, = k (for 1 =A1=DC i =A1=DC 4 and = 1 =A1=DC k=20 =A1=DC 10) denote the k-th version of = the i-th virus. The obfuscated versions were created by = varying=20 the obfuscation parameters, e.g., number of nops and inserted jumps. For = the=20 i-th virus, Vi, = 1=20 denoted the "vanilla" or the unobfuscated version of the virus. Let = M1, M2, M3 and = M4 be the malicious code automata = corresponding to=20 the four viruses.

8.1 Testing Environment

The testing environment consisted of a Microsoft Windows 2000 = machine. The=20 hardware configuration included an AMD Athlon 1 GHz processor and 1 GB = of RAM.=20 We used CodeSurfer version 1.5 patchlevel 0 and IDA Pro version = 4.1.7.600.

8.2 Testing on Malicious Code

We will describe the testing with respect to the first virus. The = testing for=20 the other viruses is analogous. First, we ran SAFE on the 10 versions of = the=20 first virus V1,1, =A1=AD, = V1,10 with=20 malicious code automaton M1. This = experiment=20 gave us the false negative rate, i.e., the pattern corresponding to = M1 should be detected in all versions of = the virus.=20

Annotator Detector
avg. (std. = dev.) avg. (std. = dev.)
Chernobyl 1.444 s (0.497 s) 0.535 s (0.043 s)
z0mbie-6.b 4.600 s (2.059 s) 1.149 s (0.041 s)
f0sf0r0 4.900 s (2.844 s) 0.923 s (0.192 s)
Hare 9.142 s (1.551 s) 1.604 s (0.104 s)

Table 5: SAFE performance = when=20 checking obfuscated viruses for false negatives.

Next, we executed SAFE on the versions of the viruses Vi,k with the malicious code automaton = Mj (where i =A1=D9 = j). This=20 helped us find the false positive rate of SAFE.

In our experiments, we found that SAFE's false positive and negative = rate=20 were 0. We also measured the execution times = for each=20 run. Since IDA Pro and CodeSurfer were not implemented by us, we did not = measure=20 the execution times for these components. We report the average and = standard=20 deviation of the execution times in Tables 5=20 and 6.=20

Annotator Detector
avg. (std. = dev.) avg. (std. = dev.)
z0mbie-6.b 3.400 s (1.428 s) 1.400 s (0.420 s)
f0sf0r0 4.900 s (1.136 s) 0.840 s (0.082 s)
Hare 1.000 s (0.000 s) 0.220 s (0.019 s)

Table 6: SAFE performance = when=20 checking obfuscated viruses for false positives against the = Chernobyl/CIH virus.=20

8.3 Testing on Benign Code

We considered a suite of benign programs (see Section=20 8.3.1 for descriptions). For each benign program, we executed SAFE = on the=20 malicious code automaton corresponding to the four viruses. Our detector = reported "negative" in each case, i.e., the false positive rate is 0. The average and variance of the execution times = are=20 reported in Table=20 7. As can be seen from the results, for certain cases the execution = times=20 are unacceptably large. We will address performance enhancements to SAFE = in the=20 future.

8.3.1 Descriptions of the Benign Executables

  • tiffdither.exe is a command line utility in the \textit{cygwin} = toolkit=20 version 1.3.70, a UNIX environment for Windows developed by Red Hat.=20
  • winmine.exe is the Microsoft Windows 2000 Minesweeper game, = version=20 5.0.2135.1.=20
  • spyxx.exe is a Microsoft Visual Studio 6.0 Spy++ utility, that = allows the=20 querying of properties and monitoring of messages of Windows = applications. The=20 executable we tested was marked as version 6.0.8168.0.=20
  • QuickTimePlayer.exe is part of the Apple QuickTime media player, = version=20 5.0.2.15.
Executable
size		 .text
size		 Procedure
count		 Annotator Detector
avg. (std. dev.) avg. (std. dev.)
tiffdither.exe 9,216 B 6,656 B 29 6.333 s (0.471 s) 1.030 s (0.043 s)
winmine.exe 96,528 B 12,120 B 85 15.667 s (1.700 s) 2.283 s (0.131 s)
spyxx.exe 499,768 B 307,200 B 1,765 193.667 s (11.557 s) 30.917 s (6.625 s)
QuickTimePlayer.exe 1,043,968 B 499,712 B 4,767 799.333 s (5.437 s) 160.580 s (4.455 = s)

Table 7:=20 SAFE performance in seconds when checking clean programs against the=20 Chernobyl/CIH virus.

9 Conclusion and Future Work

We presented a unique view of malicious code detection as a=20 obfuscation-deobfuscation game. We used this viewpoint to explore = obfuscation=20 attacks on commercial virus scanners, and found that three popular virus = scanners were susceptible to these attacks. We presented a static = analysis=20 framework for detecting malicious code patterns in executables. Based = upon our=20 framework, we have implemented SAFE, a static analyzer for executables = that=20 detects malicious patterns in executables and is resilient to common = obfuscation=20 transformations.

For future work, we will investigate the use of theorem provers = during the=20 construction of the annotated CFG. For instance, SLAM [41]=20 uses the theorem prover Simplify [49]=20 for predicate abstraction of C programs. Our detection algorithm is = context=20 insensitive and does not track the calling context of the executable. We = will=20 investigate the use of push-down systems, which would make our algorithm = context=20 sensitive. However, the existing PDS formalism does not allow = uninterpreted=20 variables, so it will have to be extended to be used in our context. =

Availability
The SAFE prototype remains in development and = we are=20 not distributing it at this time. Please contact Mihai Christodorescu, = <mihai@cs.wisc.edu>,= for=20 further updates.

Acknowledgments
We would like to thank Thomas Reps and = Jonathon=20 Giffin for providing us with invaluable comments on earlier drafts of = the paper.=20 We would also like to thank the members and collaborators of the Wisconsin Safety Analyzer = (WiSA) research=20 group for their insightful feedback and support throughout the = development of=20 this work.

Bibliography

[1]=20
S. Staniford, V. Paxson, and N. Weaver. How to 0wn the internet in = your=20 spare time. In Proceedings of the 11th USENIX Security Symposium = (Security=20 '02), pages 149 - 167. USENIX, August 2002.

[2]=20
CERT Coordination Center. Denial of service attacks, 2001.
[ .html = ]=20

[3]=20
S.S. Muchnick. Advanced Compiler Design and Implementation. = Morgan=20 Kaufmann, 1997.

[4]=20
S. Chandra and T.W. Reps. Physical type checking for C. In = Workshop on=20 Program Analysis For Software Tools and Engineering, pages 66 - = 75. ACM=20 SIGPLAN - SIGSOFT, September 1999.

[5]=20
R. Wang. Flash in the pan? Virus Bulletin, July 1998. Virus = Analysis Library.

[6]=20
Z. Xu. Safety-Checking of Machine Code. PhD thesis, = University of=20 Wisconsin, Madison, 2000.

[7]=20
z0mbie. z0mbie's homepage.
[ http=20 ]=20

[8]=20
z0mbie. RPME mutation engine.
[ http ]=20

[9]=20
z0mbie. Automated reverse engineering: Mistfall engine.
[ .txt ]=20

[10]=20
TESO. burneye elf encryption program.
[ http ]=20

[11]=20
P. Sz=F6r and P. Ferrie. Hunting for metamorphic. In = Proceedings of Virus=20 Bulletin Conference, pages 123 - 144, September 2001.

[12]=20
C. Collberg, C. Thomborson, and D. Low. A taxonomy of obfuscating=20 transformations. Technical Report 148, Department of Computer = Sciences, The=20 University of Auckland, July 1997.

[13]=20
F. Cohen. Computer viruses: Theory and experiments. Computers = and=20 Security, 6:22 - 35, 1987.

[14]=20
D.M. Chess and S.R. White. An undetectable computer virus. In=20 Proceedings of Virus Bulletin Conference, 2000.

[15]=20
E. Kaspersky. Viruses Without an Entry Point, volume Virus = List=20 Encyclopaedia, chapter Ways of Infection. Kaspersky Labs, 2002.
[ = http=20 ]=20

[16]=20
G. McGraw and G. Morrisett. Attacking malicious code: Report to = the=20 infosec research council. IEEE Software, 17(5):33 - 41,=20 September/October 2000.

[17]=20
C. Nachenberg. Polymorphic virus detection module. United = States Patent=20 # 5,696,822, December 9, 1997.

[18]=20
C. Nachenberg. Polymorphic virus detection module. United = States Patent=20 # 5,826,013, October 20, 1998.

[19]=20
GrammaTech Inc. Codesurfer - code analysis and understanding = tool.
[ .html ]=20

[20]=20
DataRescue sa/nv. Ida pro - interactive disassembler.
[ http ]=20

[21]=20
Kaspersky Labs.
[ http ]=20

[22]=20
M. Samamura. W95.CIH, volume Expanded Threat List and Virus = Encyclopaedia. Symantec Antivirus Research Center, 1998.
[ .html=20 ]=20

[23]=20
C. Collberg, C. Thomborson, and D. Low. Manufacturing cheap, = resilient,=20 and stealthy opaque constructs. In Principles of Programming = Languages=20 1998, POPL'98, San Diego, CA, January 1998.

[24]=20
P. Cousot and N. Halbwachs. Automatic discovery of linear = restraints among=20 variables of a program. In Proceedings of 5th ACM Symposium on = Principles=20 of Programming Languages, POPL '78, Tucson, AZ, January 1978.

= [25]=20
B. Barak, O. Goldreich, R. Impagliazzo, S. Rudich, A. Sahai, S. = Vadhan,=20 and K. Yang. On the (im)possibility of obfuscating programs. In = Advances in=20 Cryptology - CRYPTO `01, volume 2139 of Lecture Notes in = Computer=20 Science, pages 1 - 18, Santa Barbara, CA, August 2001.=20 Springer-Verlag.

[26]=20
G. Morrisett, D. Walker, K. Crary, and N. Glew. From System F to = Typed=20 Assembly Language. In Twenty-Fifth ACM SIGPLAN-SIGACT Symposium on=20 Principles of Programming Languages, pages 85 - 97, San Diego, CA, = January=20 1998.

[27]=20
G. Morrisett, K. Crary, N. Glew, and D. Walker. Stack-Based Typed = Assembly=20 Language. In Xavier Leroy and Atsushi Ohori, editors, 1998 Workshop = on=20 Types in Compilation, volume 1473 of Lecture Notes in Computer=20 Science, pages 28 - 52, Kyoto, Japan, March 1998. = Springer-Verlag.

[28]=20
D. Wagner and D. Dean. Intrusion detection via static analysis. In = 2001=20 IEEE Symposium on Security and Privacy, Oakland, CA, May 2001.

[29]=20
J.T. Giffin, S. Jha, and B.P. Miller. Detecting manipulated remote = call=20 streams. In 11th USENIX Security Symposium, San Francisco, CA, = August=20 2002.

[30]=20
E.M. Myers. A precise interprocedural data flow algorithm. In=20 Conference Record of the 8th Annual ACM Symposium on Principles of=20 Programming Languages (POPL), pages 219 - 230, Williamsburg, VA, = January=20 1981.

[31]=20
W. Landi. Undecidability of static analysis. ACM Letters on = Programming=20 Languages and Systems, 1(4):323 - 337, December 1992.

[32]=20
D. Moore, V. Paxson, S. Savage, C. Shannon, S. Staniford, and N. = Weaver.=20 The spread of the Sapphire/Slammer worm.
[ .html=20 ]=20

[33]=20
David W. Currie, Alan J. Hu, and Sreeranga Rajan. Automatic formal = verification of dsp software. In Proceedings of the 37th conference = on=20 Design automation, pages 130-135. ACM Press, 2000.

[34]=20
Xiushan Feng and Alan J. Hu. Automatic formal verification for = scheduled=20 vliw code. In Proceedings of the joint conference on Languages, = compilers=20 and tools for embedded systems, pages 85-92. ACM Press, 2002.

[35]=20
George C. Necula. Translation validation for an optimizing = compiler. In=20 Proceedings of the ACM SIGPLAN '00 conference on Programming = language=20 design and implementation, pages 83-94. ACM Press, 2000.

[36]=20
S. Horwitz, T. Reps, and D. Binkley. Interprocedural slicing using = dependence graphs. ACM Transactions on Programming Languages and = Systems=20 (TOPLAS), 12(1):26-60, January 1990.

[37]=20
T. Reps, S. Horwitz, and M. Sagiv. Precise interprocedural = dataflow=20 analysis via graph reachability. In Symposium on Principles of = Programming=20 Languages, pages 49-61, New York, NY, 1995. ACM Press.

[38]=20
J. Esparza, D. Hansel, P. Rossmanith, and S. Schwoon. Efficient = algorithms=20 for model checking pushdown systems. In Proceedings of = Computer-Aided=20 Verification, volume 1855 of Lecture Notes in Computer = Science,=20 pages 232-247, July 2000.

[39]=20
H. Chen and D. Wagner. MOPS: an infrastructure for examining = security=20 properties of software. In 9th ACM Conference on Computer and=20 Communications Security (CCS), November 2002.

[40]=20
T. Jensen, D.L. Metayer, and T. Thorn. Verification of control = flow based=20 security properties. In 1999 IEEE Symposium on Security and = Privacy,=20 May 1999.

[41]=20
T. Ball and S.K. Rajamani. Automatically validating temporal = safety=20 properties of interfaces. In SPIN 2001, Workshop on Model Checking = of=20 Software, 2001.

[42]=20
K. Ashcraft and D. Engler. Using programmer-written compiler = extensions to=20 catch security holes. In 2002 IEEE Symposium on Security and = Privacy,=20 pages 143-159, May 2002.

[43]=20
U. Erlingsson and F. B. Schneider. IRM enforcement of Java stack=20 inspection. In 2000 IEEE Symposium on Security and Privacy, = pages=20 246-255, May 2000.

[44]=20
B.V. Chess. Improving computer security using extending static = checking.=20 In 2002 IEEE Symposium on Security and Privacy, pages 160-173, = May=20 2002.

[45]=20
J. Corbett, M. Dwyer, J. Hatcliff, C. Pasareanu, Robby, S. = Laubach, and H.=20 Zheng. Bandera: Extracting finite-state models from Java source code. = In=20 ICSE 2000: International Conference on Software Engineering, = pages=20 439-448, 2000.

[46]=20
S. Owre, S. Rajan, J. Rushby, N. Shankar, and M. Srivas. PVS: = Combining=20 specification, proof checking, and model checking. In CAV 96:=20 Computer-Aided Verification, pages 411-414, 1996.

[47]=20
M. Bishop and M. Dilger. Checking for race conditions in file = accesses.=20 Computing Systems, 9(2), 1996.

[48]=20
R.W. Lo, K.N. Levitt, and R.A. Olsson. MCF: A malicious code = filter.=20 Computers & Society, 14(6):541-566, 1995.

[49]=20
D. Detlefs, G. Nelson, and J. Saxe. Simplify theorem prover.
[ = .html ]=20

[50]=20
M. Fitting. First-Order Logic and Automated Theorem = Proving.=20 Springer-Verlag, 1996.

[51]=20
J.E. Hopcroft, R. Motwani, and J.D. Ullman. Introduction to = Automata=20 Theory, Languages, and Computation. Addison Wesley, 2001.

Notes:=20
  1. Note that the subroutine address = computation had to=20 be updated to take into account the new nops. This is a = trivial=20 computation and can be implemented by adding the number of inserted=20 nops to the initial offset hard-coded in the virus-morphing = code.=20
  2. Most executable formats require that the = various=20 sections of the executable file start at certain aligned addresses, to = respect=20 the target platform's idiosyncrasies. The extra space between the end = of one=20 section and the beginning of the next is usually padded with nulls.=20
  3. We use one-way matching which is simpler = than full=20 unification. Note that the instruction sequence does not contain any=20 variables. We instantiate variables in the pattern so that they match = the=20 corresponding terms in the instruction sequence.
Mihai Christodorescu <mihai@cs.wisc.edu><= /ADDRESS>This=20 paper can be found online at h= ttp://www.cs.wisc.edu/~mihai/my_work/papers/index.html#11.
Th= is=20 paper was converted from LaTeX using bibtex2html 1.57 = and LaTeX2HTML=20 1.67.
Last=20 modified: Mon May 12 18:55:12 CDT 2003 ------=_NextPart_000_0000_01C80F9B.F234E740 Content-Type: image/png Content-Transfer-Encoding: base64 Content-Location: http://www.cs.wisc.edu/wisa/papers/security03/images/overall_architecture.png iVBORw0KGgoAAAANSUhEUgAAAuAAAAFDCAIAAAD03CyDAAAACXBIWXMAAAxNAAAMTQHSzq1OAAAA HXRFWHRTb2Z0d2FyZQBHTlUgR2hvc3RzY3JpcHQgNy4wNULvQ9UAACAASURBVHic7L1/cNzmde/9 rGQpkiXRxmbsxqT72oOlWlq5jp3Bkpf+MaZHxEqKM7ZYO9h/MmRlKQNMqVs1k7cpUKqZ93Y6UYE6 k95I4baLuZR9qd5p3sXUkTrz2lYWVKw/bLPLxTSaXCtKyEWlxFy3ZkNIWktURUv7/nEiFN5f3B/Y xf44n9F4sMCD5zm7pInvnnOec3zZbJYgCIIgCII0E+u8NgBBEARBECQXFCgIgiAIgjQdKFAQBEEQ BGk6UKAgCIIgCNJ0oEBBEARBEKTpQIHSYkiSFAwGfT6fz+cLh8OmaVY6A9xb/vhQKOTz+SRJqnSh KtaqCEmSfD6fIAiV3ujLQ9f1elhYpiXlny+TQCCgqipZ67cl/yeb/8kUO6/rOkxuWVbVdiIIgpQi i7QOoigSQnieX15eTqVSNE1TFJVMJrN39ooXu9F5taKfeyqVgvE0TVdhcF1/x2DyVCpV3Y1O4vF4 PSzMVvJzKf+u0oiiSFFUtuRvS7bIT7bY34eCn9jy8jJFUaIoVmcngiBIaVCgtBI5zy14AnEcl3+p xI2iKJb/UJFlmRDCMAwhJBaL1WhwjeRokfInr/rG2mmwQHGKhhK/LdkiP9lK7eF5vjqNiCAIsiYo UFoJmqbhGROLxZxPBZZl4REiiuLy8jLHcfY3XVEUnVezjofN8vIyfNsmhBT7KswwDE3T8G2b53k4 aX/zdt6Yv65zLbgENsOjcWRkxPmNHMY77ZFl2bkcy7JOC+03xbIs3Aj2wOcDCxW8MVvETxCNRskd fwNFURRFLS8vFzSp4OfmfITbx6V/LvmfpO3IsWfIXz0ej9vTUhQVjUadbw1UCMxT7LdlzZ9sjpHF PrFsNhuLxZy3IwiCuAgKFM9YWFh4ywE8Dm3s885HSzKZhK+89vOJ53nnwzibzUajUZZlc04WfHzC wwyecHBs+/8BeHrBc5HjOAgc2DOkUqlkMlnmuvAkg6lYlrXDCsvLyxB6WF5eBuECNji/mtvvNycQ k/NG4GkKk4BqKX1j/uMWFoVPGMwoaFLBz63gJ1zmzyUejzvNts8XXB0ugc8jGo3mCBQwHn6XSvy2 lP7J2qwpUJaXl0m14T8EQZDS+LJY6t4jJiYmenp6Nm/eDC/37NnT3d1tXz1+/DgcBIPBL3zhC84b LcsyDMMwDFVVTdNkGCaZTEI+YzabtSxLURQYAAmM2WzWvkoIsY+dJwuiKEpObmwsFuM4ruBsa67r 9/tpmo7H436/XxRFWZYtywqFQoZhgFcgFArlJKvC+WJ2Fnsja77BEm88EAiYpinLMiiPgiaFQqH8 2wt+JhV9PvnHLMvmr25ZliAIlmVRFAXOIacKyX9rBX9byvnJlvmJ2ZdOnz599uxZ+/yBAwcCgYD9 cmJiwnnXkSNH7OOcG4eGhnbv3m2/PH78+MLCQsFpU6nU1NSUfam3t3f//v3Fpi1mT29vLyHEeePc 3NxPf/pT++Wjjz7a39/vnHZxcdF+6fzfNp1Ov/XWW/alnp4e5xvJmbbY/+9A69ozNzcHshUYGBgA FyAhxLKsRCJhX/L7/c4bU6mU8wftvBHeZs4ZpBO4y2sDOpo9e/Y4/2g6cf5FsNF1Xdd1nudZlmVZ lud5v99vGIZzTDgc1nVdluVoNFpscifwqDNNE55ezj8BmqblG+CMU1S0Ls/ziqIoigLHcIthGM4g FCkpmIpBUZRlWfBG4Nlf3R8ywzBgn4uu6yBQSpiU87mtOXnpz6eE2fmrg3DRdV3TNF3Xnc8DJyV+ Wyr6yZbJwMAAPOyBnp4e59UDBw6UeaPf73deffTRR51TOa/6/f6hoaFiNzrnzL/qvDEHv99fbMX8 l/YXDDgufaPzqvNGkvdx1cOexrC8vOwUTCsrK/Zv9crKivNS+TcSQs6ePdvb24sCpeNohJsGKUR+ WGdNnHkS2TsBiJwkWThIJpPxeDznJExiH4NKADc++Pmd9kDYwpleALGYYrOtuS6EFSiKYhgmm5ez ab8dCF6IoshxHNhT7BfVPm9vV8l+OtxT+kYnMB70GZgB4Y+CJhX83Oy14JaKPp9UKgV35YRUSqye TCadi9qA1IM4TrHfljJ/suV8YjCV84eIIPXgT//0TxcWFry2Amk0KFBaDGc2KMMwkH2ZvfNkEkUR Hkvw2OA4jmXZ5eVl+2r200myPM/DlxKapnM26cCT3nnSPlPwAVxw3ZwHHngaIG3C6TUheUmyEInI WSIH5xuxPxZntm9Fj1unsoEHNjzm800q+LnZHhdwCNnZJGv+XFiWhY/FVhLZT2uvnNVTqVR+sq0N qBA7MaXgb0uZP9k1PzF7ufpt0kYQAAVKZ4I5KEjjkCRJURTYJuO1Le2JZVnBYBByfRqwHKxlKxsE qRMTExM5WURIJ4CVZJEGYRiGpmm27wGpB7DxWNf1nMykeqDrOiQU13shBBkaGsrJtkE6AfSgIA0i GAxSFBWLxVCg1BtN02CPT71XIYTUmFqLIAhSDBQonoFOSwRBEAQpBoZ4EARBEARpOlCgIAiCIAjS dKBAQRAEQZqa48ePp9Npr61AGg0KFARBEKSpWVhYWFlZ8doKpNFgqXvPcLYjQRAEQRDECXpQEARB EARpOlCgIAiCIAjSdGAdFARBEARBmg70oHhGKpW6ceOG11YgCIIgSDOCAsUzpqamFhcXvbYCQRAE QZoRFCgIgiAIgjQdKFAQBEGQpgYD4p1Jveqg6PpSnWZuLRjmXora4LUVCIIgLczU1BS2Vu1A6uJB Mc3rqnqpHjO3FoJwroQ6GRoa8vv9jbQHQRAEQVqFunhQVPUix3Wz7H31mLxVMIwrpT+B3bt3N8wY BEEQBGkt6uJBMYwrHNddj5lbCFW9yPMPN3LFUCjkQ9aCZdlG/lAQBEGQ6nDfg6JpaYa5x/VpWw7D uNz4z2FkZGR4eLjBi7YQMzMzV65c8doKBEEqo7e3d/PmzV5bgTQa9wWKri+J4nbXp20tNC3tSYSr q6urv7+/8eu2CnNzcyhQEKTl2L9/v9cmIB7gcojHslYN4zJN3+3utC2HpqXXjO8cP348nU43xByk hTEMw2sTGoRlWV6bgCBIE+GyB6WcB3PbY1mrlrW6pkpbWFhYWVlpjElIC6HrumEYpmkahtE56gSg KIphGIZhWJZlGIaiKK8tQhDEM1wWKKp6MR5/0t05Ww5VvYQ5wkilSJIE0oTceU6zLCuKYkc9pHVd N01T0zRFUQghNE2zLCvLckd9CAiCAG4KFNO8jnXJCCG6vhSLBb22AmkZTNMMh8OGYYiiKIoiwzA0 TXttlDfYe6wsyzIMQ9d1VVU1TYtGoxzHeWsb4iHHjx/fs2dPdzd+8ess3MxBUdWLHV77hBBiGFdo +m5UaUiZKIoSDAYJIclkUpZljuM6Vp04oSgKfCepVIpl2XA4HA6HMUmlY8GAeGfipkDB8ieEEE1b LPNDOHDgQE9PT73tQZoW0zRDoZAkSTzPJ5NJhmG8tqgZoSgqFovFYjFd1wOBgKZpXluEIEiDcE2g YPkTQNeXynQjBQKBTZs21dsepDkxDCMYDJqmGY/HZVn22pxmh+M425UiSZLX5iAI0gjcFCi4f8er 8idIa2FZVjgcZhgmmUxiZdsyAVeKKIqKonTa5iYE6UzcEShlbqxte1ClIeWgKIplWdFoFDenVIos yyzLCoLgtSFIQ8GAeGfijkDBjbWkcpWWSqVu3LhRV5OQJkTXdUVRRFHEZNjqiEajpmlioKejwIB4 Z+KOQNH1JRQolaq0qampxcXF+tnjIoMOdu3aJctyJpPx2qiWxLIsQRCgwInXtrQqNE1DoEfXda9t QRCkjrggUFzZWOvznbL/+f1vCsI5y1qt3bZG0iEq7erVqydPnpycnCw2AHRM/jFCHMEdrw1pbURR hEAPbjxGkDbGBYFS/sbaMrGsm6p6UZLOFxsAOib/2EM6ofzJ7Ozs7OzssWPHCCEnT550ceYO6Ulk miYGd9wiGo1alqWqqteGIAhSL1wQKOVvrF2TbHZvNrsXiuWr6kVX5gRM87qLs+XjukprWiBVraur K5PJHD582A79RCIRQsihQ4dgWCQScR7Df3ft2gVBohMnTpA7/pVDhw6B3IGXL730EoyBu9oJ2HvC 87zXhrQDNE0zDGOapteGII1gYmIilUp5bQXSaGoVKPXYWEvTWwghFLXRslbD4aQd+gGfSij0HgyT pPPOY/iv3/8mBIkUZYHc8a+EQu+B3IGXgYAOY0o4aSqlCpXW29u7efNmtwxoACAgXnzxRULIyMjI zMxMJpN5/fXXZ2dnCSHT09OEkKNHj8Lg8fFx5/GJEyemp6ePHj06Ozu7c+fOyclJ22uSSCT6+/vt VY4dO/baa69dvXoVJmwnDMOgaRp37rgFwzC43xhB2phae/FoWlqWd7hiCiHEGazh+Yc0LW1Zq6lU iKbv9vlOKcq8LO+Ix5+AYbCufawoC4oyn0w+yzD3CMI5SXrfdmno+kei2GvPHI8/aVmrweDbMGHt Zlen0vbv31/70o2nq6trZGRkdHSUEJJOp2VZvnDhwpp3zc3NEUL27dtnn7EThEHf2HR3d7drxw3D MLBcrIswDAM9BREEaUtqEih1Kn9CURt5/iFR3E4IMc1rgnDOMC6veZeuLxFCgsG37TOmeQ0Ostm9 zpGuG+yuSmtacpTEoUOHEonEwYMHJUkCt0qlM3Qa0A7Qfunz+Qgh2WzWO4v+E9sYt6yCeZy4/k5B 7em6jsXuEKQtqUmguF7+JEdJhELv6fpHsvz5aPSxQCBexQwNoGOL1CUSCUJIf39/OZul+/v7E4nE mTNndu7cGYlEFhcXO62OhWmalmWV70FpKvli01RWQbzMMAwUKAjSltSUg1LvjbW6/hEhhGXvs30h JYAgi6alCSGSdD4cTjZmo3LHFqkDhbFv376TJ08ODw8PDAxAcZSBgQFyJzHWPh4dHR0bG5NleXBw MJFIjI6Obtu2zVPzGw1kS9TjUepuoqgoii7WaMk6KGd8pe8F82Q7hCNHjgQCAa+tQBpN9QKlARtr o9HHCSHB4Nuqeonjelj2ftAcLHs/uZMYax+LYq8obheEcz7fKV1fEsXtjdn0W7VKO336dKtUcYAN xjknR0ZG4Py3v/3tb3/720ePHgXNAZmw4+PjOcfj4+M/+tGPZmdnX3vttb6+vvxpnS8LrtjSlEhA 8fl8Pp8vEAj4fD6/3y9JUigUgkugAiVJ8vv9cNXOuoC7QqEQbLXNmQQOfD5fOBwuNoPzpG2Moigw ABoG+e5QplUF3xrcbt/rtA3kRUXvxQkKFARpY6oXKO5urIUNxjknef4hOB+LBWOxYDz+BGiOePyJ bHYvpH04j2V5x/Lyl7LZvcnkELRWzpnW+bLgipVSi0o7e/bs8vJyjQYgLUTp/TvxeDyZTFqWpShK PP6bgKYsy6AY4vF4NpvlOE6SJOcjOScDI5lMEkIsy4rFYvDbpWlawRkkSYKiLNlstuDOZ03TLMtK pVLg/KjIqmLAohzHZbNZWZY1TXN21Snnvay5BIIgbUP1OSi6vtQJmaGl6ZzyJ0i9KVa9DQq6B4NB +4xpmvbgnNCJrYGc3pqCM4DPA6I5IDhy1uU4zjRNQRAKbuUtbZVNjnmwiizLsLQkSc5y9eW8FwRB OocqPSj1KH/SirhYpA5BSuBM5qgukSVnBnj8Q5CxYKgxHA4risKyLHgyXLEqf1GsCoOUQwsFxBEX qV6g8PzDrlrSeqBKQxoAPPghuiFJUjgcrvQvdcEZIKwDLo2CO6rAt8GybMHYTXVWwaKwHCyNdXWR csCAeGdSTYinYzfW5lBj+ZOhoSFncmI+8P0V/dtNiGEYiqLQNM2ybJ32uLIsq+u6JEmyLEMP5HA4 zDBMNBqt1OsgimL+DBBngVQSjuPy30U0GhUEIRgMchzHcZxlWZZl1WgVRJQ0TfP5fBRFiaIIZiAI guTjq6KkgaIsUNQGnn+oHga1ClCGPx5/on5L6LoeCoW2bt3Ksuzg4OCaYiUUCnV3d8OWmapxNh/u 6+sbGxvbuXNnscGRSGR6enpkZESSJLixxNabggPWvMtdIpHIr371qzNnztQ+laqqdoLnU0899fTT T5cWK5IkGYZh55kiriBJ0j/+4z9+/etfR09MezMxMXHgwAHcadxpVBPiwcxQQoimpev9IbAsK4ri xx9/fPLkyT/7sz8LBoPd3d2CIKiq2pitlRcuXJiYmIAq9QWBdjljY2MVTQs9fWo1zmt4nrcr97/z zjvf/e53Q6GQz+d7+umnc3I/kbqSyWTgfwqvDUEQxGUqFiiGcYVh7m1MiZFmRlUvNkClybL87LPP rlu37pNPPiGEfPjhh/DFPRAI7NixQ5Ik2Avq+rpQhgSUB7QaLgG0zhkbGyutVAoOWPOuZubVV199 9NFH169fTwhZXf1NVcB33nnnO9/5DoiVL3/5y4qiVNrQTpIkn8/n3H9bBVA4pIpbAL/fLwhCS6Ql 9vX18TwvCEKn1SbuKNYMiCNtScU5KKp6EdNj/+7v/r9bt15XlDcbsNbjjz9uGAYUaXXys5/97Oc/ //nt27cJIQ8++OBXv/rV5eVld9vsjY6OTk9P2x6USCRy8uTJq1evdnV1jY6O2ucPHTp09OhR8KZA gAkcJD09PYuLi9BccHx8HAbYnQUjkYh9Eu6KRCIzMzNQOH94ePjgwYPd3d0Fp6r6HX344YdQAgRe WpZVhS9qTe/IrVu34OCNN9544403CCEURT322GN2A+fSODcAV4SLdegtywKfRDQaXXMtz+vfR6NR hmFAURUzGGlpdu/e7bUJiAdULFAM4zLUQCuNsy8xw9writtL+Bsk6byizPP8w9HoY3BjiRJqBQes eZe7/O//nbx9e8EwflnLJKlUqru7e/PmzWuOhBZu+Q+A27dvr1u37vbt2x988MGrr756zz33fPzx x7WYlANUhr169Soh5MSJE9PT01AEVpblycnJ119/HXoEHj16tODtx44du3r16r59+6anp21VcfTo UdAcOToD0lnGxsbGx8dPnDgxOTmZyWTsmQtOVQUff/xxJpMp4dIoJyXZOebs2bMFE2jWr19/69at z372s4ODg7//+78PuaX5hVBLUKwsipOCpUfyqaiLMvyaQf6TqqruPu/LNLh8DMOACSEHBdxOsizj 7mUEaQMqEyhQcr7SNQzjcjg8F48/WWxTrqLME0KgfXH5NFiUOPnkk13nzv0/NU5SZtpXfkLDXXfd BRGfRx555IUXXrBzMyGjtkarigH+EjvrghCyZo/A7u7u8j06TlfK6Ojo5OQkNCOsYqoSbN++fdOm TW5lquq67qxvBqKEEPLcc88988wzOUnNZbbetcvJh0KheDwOUTxw83AcJ8syTdPgsYD57V0wzjr0 9slAIAD3wn4ZSZJUVbUsC3bQlHDSwFOfoijYqmOXcIV5nGvZag/WLbiE02D4xGiaNk2Toiie52VZ hk8Sfs9hh1GZSa/OToE8z9M0HQ6HIRkZNQpSKbq+pOtLXltRL2h6S8ttbalMoFRaPRbUAzhIVPVS 6aohsG95TZlScECl4qYWNC1djg/JpbW0nBKfDzzwwNjYGMMwdq2t+nHixAlCCPTNAeq33aarq+vq 1auZTGbbtm0Qz+rq6qrTWq5gWRY4kAAQiwzDcBxXcHyZrXfj8Tg8zkGdQDV6qPQqSZJlWba60nXd qTDsG50bd+PxuGVZwWBQURSKohRFSSaTEA2RJInjuHx/hjNzhed5u+A9CCNFUWRZzlnLPoZNy8WW AIPh99lpmK14YrEYx3Hlp7vmN4hmWTYej4dCIZB3qFGQilDVS9HoY22ZYWkYV1T1IiEtJlAqSJI1 zevkjoyoCFAPoEwl6bzf/6bPd8rvf1NRFgghodB7MAwOFGUeHCo+3ymf71QgoMNgaA1oD7DvgvP2 XZJ0Hm7x+U6Fw0mwueBsur4UCr0H5/3+N1X1UplvR9eXGpOFY5rm1772NULI1q1b9+3bF41GU6lU Op2WZZnjuLr+8YWNNpOTk4SQkZERQkh/fz8hBDboRiKRw4cP56fF1AKsAiuCMIIzTYsgCD6fb2Rk BH4u58+fh59LiVsq7WyXUxiefDr9pZzirTRN289vuzi9z+cDEVDCGNv/wXEcqI1yUhRLL+E02GkY ISQWi1EUFQ6H/X6/rutlBqQKNohmGAY0XCAQqDQ9GWlajh8/XmYKV9VAfa+2VCeEEE1bbMWyohV4 UKounAo/csu6qSgLijKfTD7LMPcIwjlJep/juuPxJyBYU7CmSDz+pGWtBoNvK8q803lj3+U8Ca4a UdwuyzsUZUGS3resVee0ztlA0MRi/RzXXb46saxVw7jcmCJ1qqpOTEx4WKutp6dneHgYhMLo6Ggm k5FleWJioq+vT5IkyFCpgoGBgUQiAUmy9snR0VFCyMzMzODgYFdXFySjuPIu6oFpmjzPx2Kxiu5i GKaidncQYYFwiVuF4ddMZc0ZEAqFdF2XZTkajZZZhaKKbFmoFKfruq7rmqbpul5O2dBiuTWgUWw/ CpY6bAMWFhZWVlbqukR7VwZv0d55FRRqCwbPJpND5c776QQReMmy9+v6R85hkJjiHGwfFzxZegCQ s2ixGWKxfkE4Z1k3KWojy94nitvLCdyAlGnOSJ4rhdraGxcLtVWBpmnhcHh5ebm0zrA3xUCIh+f5 aDTqDPcU2zVTbGeNHYKRJAnCKNB8OKf8a8Fp4SS0WYZATLH5s9ksxKHylyhtWDabhUImyWSSpmlw 1ZTzdykUCtE0XSyN17KscDis63o0GsUybq1OAwq1hcNJWd7RlhXSTfO6osxHo495bUjFlBvigfIn 1a0BoRz79mx2r/3PXcVKURsJIZa1av8XzhSE47pTKRY8KLq+FAq9W84SjSl/grQl8FW+/BwLCLLo ug4RkzULw0Owo1g5EJgNIlOQDlKOPwYe/8FgUFVV8HOAL8e5ln1c3RJgG8dxwWAQ1Ek5W6xN07S3 8BSEoqh4PA4lUrCMG1Iay1o1zWttqU5IKzuHyg3xVFf+xOnY4PmHLWtV1z+CGqySdN40r7ubkcTz DynKvCSdj0YfgwhOCVeHIJxT1YvJ5LOyfJ+qXixnftO8jkXqkKqhaRqyRAtmp9rYzgPYz5IvSop5 F5y7k5xj7OOCs5Weluf5gu4H51rO4zUNLnhM03Sl8TJBEGATUOlh4MIRBME0Tez7gxRD15eq2KDa KmjaYvnRj6aiXIFSZvmTgtD0Fo7rBq1gWTcF4Vw4PMcw90ajj9fysIeAkSSdt0NrkI2raWmf7xRF bYRklGK3i+J2yEdx3lsaVb3YojoUaRJEUdQ0TRAEbMpTC7Anucx9OrA328UybtjHYE0YhnE3i//I kSMuzpaPri81citoIzHN6zS9xWsrqqSsHBRVvWRZq6LY2wCDmplQ6D0XuwOmUqmenp5Nmza5NSHm oKyJtzkoANRAk2W5ilqxCCHENM1gMAgFVMq/C3pEcBxXRTvoHCrtIdCBxOPxOnX5rhPu/m1vKlq6 uW9ZHpQWTQB2F9fLn0xNTWF/zg4EekCuGehBigHBnUrlHc/zDMO4VSLl2LFjsPEeyaflWoE2srRV 49H1pVgs6LUVVbJ2kmzV5U/ajIaVP2lOXGxB3B7djGsBskdrbAfYmUBwpzovCGw/Nk0zFAo1ph84 0hK0cQIKVHZp3bzJtQVK6yYAu0gjy58gbQ9FUdFo1DCMYDCIxcTKBLYNw16hqsMHDMMkk0lCCH7y rUUqlbpx40adJq8lw7LJafXHdzkCZbFFw1cuomnpTnafeEK9C0d6C8uyyWSSoqhgMFhsYzBio2la IBDQdT0Wi9W4GYem6Xg8TtN0KBRCjdIqTE1Nrdn8qzpa/RFeGl1faul3t4ZAqaX8STtRj/InQ0ND 5ZQPb2YikchLL70EIZvDhw+n0+lMJnP48OHBO0QiEXvkrl27BgcHd+3a5bzdPgnl7cmdANChQ4dO njzpwVtqIPCkhHwU/EJfDHCchMNhlmVTqVTpZgJlAiVSWJaF+i61T4i0LoZxuY3jO61e3GUNgVJd +ZM2o07lT3bv3t3Szcwikcj09PTw8PDs7OzBgwdnZmZkWZ6ZmclkMq+//jq0FYQexTBydHR0dnbW 7rBz4sSJ6enpo0ePzs7O7ty5c3Jy0ukySSQSHZKEKMuyHXRw9i7ucCzL0nVdkiTbcQL9etyan6Ko WCyGZdwQw7jSxvGdVtdepXbxQOJFu/7wygfLnxQExAfsah4dHZ2cnEwkEt/+9rehneGFCxdyRkK3 nfHxcXg5NzdHCNm3b589bHFxsbv7N26q+rVNbkIgMQLUCTQIpCiKuUMD2lY3CbquG4Zhmqau66DS KIriOE6W5Tp9AnYZN8MwnCVSQB654q1Bmpk2VieEEMO43OrFXUoJlDbQX65gGFdwl3U+XV1dV69e zWQy27Ztg87GXV1dhw8fTiQSBw8elCTpxRdfLDbSpqOESGmgBqtlWYZhwBPa1iudAygzaKHMMEwD tmHbZdzInbr+hBDDMHie7xxp2Pz09vZu3rzZ9WlbtMdvmZjm9ZaO75A1BUrr7p92i/beIl8pzu3B Y2Nj09PTk5OTkiRBBsnIyAh4R/r7+50ZbXD+xIkT4+Pjk5OTcLK/vz+RSJw5c2bnzp2RSGRxcbGW DsltA0VRLMvau1RAr3hrUsOgadqTwjBQLx+6G9qBpMuXL+/fv/+HP/xh4+1B8tm/f389pm3jEl/t kfxbVKCY5vWW3j/tFvUrgXz8+PE9e/bYQY2WA0I2MzMzg4ODXV1dY2Nj4+Pj3d3dsizv27dveHh4 eHg4k8lkMhkIA01PT0POysDAANyeyWRkWZ6YmOjr60N1UhDQK15b0f7klHGDANPJkyc1TcNAT7vS 3ltA2qNwV1GBgm17SZ3LnywsLKysrNRj5npQMBYzv5/sTQAAIABJREFUPj6eU1l/ZGTEToMtPbLY SQz6IJ4AZdxAozz66KMbNmxYXV09cOAABnralTLjO5qWhuZxtffbK1FQv5xa+z7fqWx2b5lrtUf+ aNFdPLq+hAIFy58gSOfAMEwqlSKE/P3f//3q6ioh5Pr169/4xje8tgupC4ZxpZxnnK4vMcy9hnEZ iqoDPt8pVb3k852SpPPFzmha2u9/E85Y1iohRNc/gkuWtRoKvefznQoEdFW95LxECJGk83DJMK4Q QmBaRVko/621R3yHFBMorV7dxS3QjYQgnYCu64qiCIIwODhoGMZ//Md/wPlbt2699tpr2L7Yc44f P+5u5cYyE0gta1VVL4ridpreoqoXcy6lUiGQFwXPKMq8KG5PpUK6vqQo86HQe4QQn+8UIURVLxnG 5eXl56LRxyTpfM4lRZlPJp9lmHvD4TlCiKpelOXPV/Tu2qa4S+EQj6alaXqLri812JqmwrJW61H+ BEGQZkPX9b/927+9cuVKwatjY2Pvv/8+Bno8xPWAeJk+Bk1LU9RGjus2zeuqetGZUSuKvYQQy7pZ 7IxhXI7Hn6SoDbFYfzB4dnn5S3aMhuO6GeYew7is60uWdTMef8K+pGlplr2fYe7h+YdCoUVdX7Ln kaT3y3x3bbPztLAHhecfboPwVY1Q1Ia6biI/cOBAT087iFwEaXVkWf6Xf/kXURS3bdu2YUPud5Kl pSXX93sfOnTIWWoZaTCatliOd1xVL1rWTZ/vlCS9b5rXKv3SDt9vafpup46BaQXhnNP74kTXP/L5 ToVC7+bMUybtVNylsAelbd5eMxMIBLw2AUGQ30BRlCzL0HlgcnJyZWXl1q1bcOmTTz5RFMW5/btG 0ul0IpEghMzMzOTnidcCFAIolmle+mrroqoqFM4pc7xpXqfpLeUMM4zLyeSz8EAMhd6rKLcDohAs e58knWfZ++EktBe+E8S5R9P+M24Fl2j6boa5Nx5/0jSva9oi7DPS9SXIYikH+642YO1mgQiCIB0C yJRf/vKXf/zHf7xly5b169fbl15++WW3VpmZmSGE9PX1LS4unjlzxq1p3aWFunXSNB0MBrdt2/by yy+rqrpms4gydYaqXqSojfbX9Rw9sSayvCMcTvp8p0zzOlQUY5h7/f43CCGiuD0YfNvnOwUBHUgn gEuyvIOiNvr9b4AHhaI2sOz9odC7ijJvq5zStNMGF182m/XaBsQdQqFQd3e3u1/I2oxIJPKrX/2q aR8JSFNhWZaiKN///vdXVlZu375NCPn617/+13/91z6f79ixY7X0itq3b18mkzl27NiLL744MjJi t7MGD0dPT8/i4mJXV9fIyMj4+HjBk5FIZGZmBsohDg8PHzx4sLu7+9ChQ+CYGRsbGx0dhd5YMPPY 2NiFCxfsq8VmsG0YGBjo6+ur7o/J4OAg9GKs+vPJJ5VK9fT0bNq0qdgASZKcYbgHHnjg+eefB6dX fvJQMHgWsjpctLBJMIwrqnoxGn3Ma0PcAT0onpFKpW7cuOG1FQiCFAa8Kb/61a+++c1vQp31//E/ /kfthX3T6fSFCxdGRka6u7uHh4fz5fKxY8dee+21q1evQl3m/JMF+3QSQo4ePQqDx8fH89t2Oq8W m8Gm2bp1BgKBEuqEECLL8rPPPrtu3W+eaB9++KGqquFw2O/379ixAxpdwSXLWqXpLW2pTkjb7cAt VeoeqStTU1MHDhzATBQEaWbs3JRvfOMb09PTTz31VI0TgldjcnLSbvsADR/sAd3d3fkFpp0nC/bp zBk/PDyc37bTZs0ZasxT+ed//mfnS9M0G9Cj+/HHH3/33Xdv3ryZc/5nP/vZz3/+c/CBPfjgg//l vzz/yCNPE9KeXVx0fWnNgm8tBAoUBEGQNaAo6tVXX/3Wt76lKIqqqr/+9a+rnsoOu9gkEgmnQFmT gn06c8YUbNtZ0Qy18Cd/8icuzlZ+k6aNGzeurq7m5y3cvn173bp1t2/f/uCDD/793//f//pff9s0 TU8aP9UVaFDjtRVuggKlrYB+N15b0dRU9CRAkBxqrIZy4cIFiO/YeScvvfTSmTNn7JflAN03c/p0 5owBj0hO286KZqgF13NQykGSpGIl9R555JEXXniBZVmGGQqHk//9v7ePj8FJ2xSQtUGB0j7IsmxZ ltdWNDtYbgupAl3XX3nllR/96Eewl/Wzn/1sdfNAxgn0ywSGh4enp6dzojylKdinEy4NDAwkEolI JCJJUn7bTvtqiRmak4mJidIBcU3TcmrVFMyTVdVLbfYId6Jpi/H4k15b4Sa4i8czWr2bMYK0PZZl aZr2/e9//6c//SkhZMOGDRcuXAgEAjXu4mlv6rGLp7RAMU3zi1/84tWrV7du3cqy7ODgIMdxBSM4 4XBSlnfUqf+rt1jWqiCcg/3MbQN6UDxj//79XpuAIEhhTNNUVfX73//+tWvX7JNvvPFG+yUutAGq qv7BH/wBx3Gla7VZ1qppXmtLdULaMb5DUKAgCII40TTte9/73jvvvLNhwwboaQyIotj4vAqkHHL2 SBdD15fao4VeQXR9qW3Kn9igQEEQBCGWZamq+r3vfe/DDz+EM7Y6Wbdu3TPPPFPmUxBpWnR9qa7t 1TwEnENttoWHYKE2BEE6HMMwBEHw+/2SJNnqxMnWrVtff/31xhuG2Bw5cqT2klGmeb2u8R1NSweD Z32+Uz7fKb//TegFCC/tf4QQXV+CEvg+36lAQC/WMrDSpdvSOYQCxTNOnz6Nm24QxHMoiqJpemRk ZMuWLYSQu+7K9SufOXMGN3+1OpqWrmsTXFW9FA7PMcy9y8vPLS8/xzD3Ksq8fTWb3Qv/NC0dCr1L URuWl5/LZvfy/MOC8BPTvF7j6u3Uf8cJChTPOHv27PLystdWIEinQ9O0KIo//OEPP/74429961vO BoGEEFmWy2+TizQt9U5AATkiyzsoagNFbYjHn0ilCmQsSdJ5exghRBR7s9m9tft12jX51+UcFF1f cnfCFoVh7m2/cCCCtDfQcG5oaOjs2bOEkPXr1z///POiKHptF+IChnG5rh4U07xGCCn2Zx+CO6K4 3TksFHpP1z8ihMTjT9ayAadd4zvEXYFimtdV9RLPP+TinK2IIJwrqJ0bQEX1KDsTnudxpyiSg2VZ 4XBY13Xwl5w9e/auu+7avn378ePHvTYNIYSQ06dPDwwMVB1oa8AWXJreYprX7DQXXV/S9SVZ3gFX s9m9cKDrS4Zx2TCuMMw98fgTpnk9EIjXuHQbJ/+6KVBU9SLHdbffVuyKMIwrHn4CiqL09fW521aj nUgkEizLtqhAsSzLMIy26SFCUVSThE4MwwiHw5ZlJZNJhmGgXPpnPvOZEydOYOpJk3D27Nne3t6q fxyGcbnePgaef1iS3leUeVneATXTGObe/GGiuD0cnhOEn8Ri/RS1wZmnUjWGcbkt4zvEXYFiGFds wdixqOpFnn+4nJFDQ0N+v991Aw4ePIg1LosxODjotQkVAIpE13XTNEGaeG2R+7AsyzAMTdMMw3ii VxRFkSSJZdlYLOZ8/n33u99tEv2E1E4Dnk2i2EsI0bRFv/8NQgjPP1xwRY7rjsX6FWUeHCeiuF0U t1PUxqrXbcv6bDauCZR650i3CuVHOnfv3l1vY5AWxTAMu/MZPLw5joNnedt8pzcMA1SXYRjQRYWi KJ7nG1ZuxLIsSZJUVRVFMWdRnud5nm+MGUi9gXhKAxYSxV6QKU7s4I4Nx3W7uONG15fK/Ercirgm UNo4DFY+7S1mkcagKIqiKDRNx2IxZ5+zNiPHZWIYBvR703U9FovVO4wFtU9M04zFYhzHlTAMaXU0 bbGN/yyb5vU2dg24s83YslbbOAxWPpqWbmMxi9Qb0zRDoZAkSTzPJ5NJjuPaVZ3kwzCMLMvJZJIQ EgwGczrTuoumaaFQiBASj8dz1AkhhKKozvnYW4VaAuK6vtSuAqVhziGvcEeg4IOZEGJZq5a1iioN qQ5FUYLBoGmayWSyY6uqMwyTTCZ5npckKRQK1SPtRpKkcDjMcVw8HkdPSauwe/fu6lSjYVwpmKza HrS3c4i4JVBg/44rU7Uuqnqpog9hYmIilUrVzx6khZAkyXac4FNTluV4PG6aJig2t6a1LCsUCimK IstyNBpFN0kn0N6P8DZ2DgEuCBTTvI51yUj7FhtG6o2u6/DUlGUZn5oAy7LJZJKiKEEQXJlQ1/VA IAAOKqy91jm08Z/l9nYOAS4kyarqxfYWceVgGFdo+m5UaUilWJYlCALLsvjUzIGiqGg0Cj6PGj+c YnuJq+bChQu1T4LUG/jy7LUV9aK9nUOACwIFy58QQjRtsXV1en51kGPHjnlSTAUsmZ2dXfNk26Ao imVZ0WjUa0OaEdBtiqJwHFfdvh7Qf5qm5e8lroXJyUm3pkLK4fjx43v27OnuruxvbHtvq+yEJ2+t AgXLnwDOqsZIMdpbalSBHdxpj+Kw9UAURU3TBEGIxyuuCF5iL3EtZLNZt6ZCymRhYWFlZaXSuzRt MZkcqoc9nmPX1G9vXBAo+GCuTqcfOXKkHsZUDeqGBoPBnXKoOtCjqqokSTRNJ5NJ1H8diGlep+kt XltRLzQt3cbRK5uakmRxYy3QHrusBx0QQk6ePDk4OCjLciaT2bVr165duzKZDCEkEons2rVrcHBw 165dJ06cyGQyzjORSMQ5m/P40KFD8DISiWQymcOHD9vL2XcBL730Esw2NzeXY2TO6nX9QOqNpmmm aVYR3IESqIFAwOfzBQIBZ4dIXx5wXpKkYDDo8/n8fj84FfKnlSTJ5/PVmJTqXNQtWJbleV5V1fJv kSRJEATYS4zqpDNp7/hOSycVlE9NAqXSjbVtSbuqtJGRkZGRkZMnT/7hH/7h1atXjx49um3bthMn TkxPTx89enR2dnbnzp2Tk5N/8zd/Mz09PT4+Pjs7OzIyMj09XSx/8OjRo3AwPj4+MzOTyWRef/11 cNtMT087R0qSdPDgwatXr+ZIkPzV0+l0fd59IzBNE9rQVHSXvVcWRIZpmpABWuIWqHtmGAbcrqpq KBSyLCtnGNRGq8KdUw9RkgPLsqZp5tucj2VZwWBQVdVoNIp7iTuZNn6Eg3OoE/Zk1CRQ2ngHV/m0 jUqbdQBnJEnq6em5cOHCwYMH+/r6CCHg0ti3b9/g4ODJkycJIa+//johZGRkhBACMgVGlmZ4eLiv r0+W5V27duVf7e/vHx0dJYQkEgnn+fzVFxcXa3rPnmIYRhUlT0BqMAyTSqWy2ezy8jLLspBpa4/J OpAkCcYnk0k4I4qiaZrFHBLlCKYya5OAJHIF+KDWnBD2EluWFY/HsZlOO3HkyJFAIFD++PZ+hLd9 +ROb6gWKixtrfb5T9j+//01FWci5VPsS9aNqlZZKpW7cuOG6PS5y4cIFUAA5oZZ8KQPRn3Q6PTc3 B8elOXz48PT0dH9//2uvvVZwAEzS1dWVf8m5ekv3bdZ1vYrog6ZphJBoNAr3QopGTgKpHd+RJMke b4shWZZBpjhvgdLv9oEdQvL5fOFwGBQJvAyFQk5xY9/o9OLAvcFgEE5KkuT3+yHAVF0Ne5qmKYqC 7onFgOKzUEAFi911OI1/hJumWdqR6SKd4xqoXqDUyYFmWTcl6X1JOu/6zPWgFpU2NTXVVA4AZw4K JIXIstzV1XXw4MFEIgHRFhAEZ86cIYREIpHDhw9/+ctfJoTA1cnJycOHD+dMWzBTBFwj/f39BT+B dDoNDhJwzNjkr16OGGpOwBlQxXMUtILzRpqmS/QUzBnv1C7OYbbEicfjkiTBzt5sNivLMmyisUfq us6ybP6Nzk288Xgcuuood4jH49lsluM4SZKqKw4LUZ6ClyzLCofDkEXrVqUTpKVp/CMcgq0NWMiy Vgkh7eocyqH6XTyub6yFttSmeT0QiCvKPExeS4fkBmzEauMwZyQSuXDhwtjY2Ojo6MmTJ0+cODE8 PDw6OprJZGRZnpiY6OvrgxjQhg0bTp48OT093dPTI0nStm3bCCFjY2PT09ODg4NjY2MDAwMw58DA QCKRiEQikiTJsrxv377h4eHh4eFMJpPJZLZt2wYjJyYmLly4MDIyAoEem/zVYa1WBASK80lfEaZp lvC+OPfBqqpqWVaOqigN/J0FwSGKoiRJTtdFOZtsnbbBvcFgsEzjS8wJ3qAcDMMIh8MQ1qn680Ta DJq+W1HmG7nipUu/JIQ05qs1zz/UgFWaAV91e/o1LW0Yl90SKPlBHFHcDpPDpWx2LxzQ9BbTvEZR G3n+IVHcLgjnNG2x4C0sez/D3GOa1zVtMZUK0fTdirIgSe/HYv0uSopg8GzV++wnJiYOHDhQUWB1 TXw+n1c11lqCwcHBJnmMwVMf3AwVEQ6HNU1jGMaO2sAZiGtAsqrzf2pBEFRVpWlalmWoBWIYRjAY zC9cZt/r9/sty1peXqYoyrIsv99PUdTy8nL+5Dk3FjxmWVbX9dprh+i6HgqFwCr7pL2XOBaL4W6d NiaVSvX09GzatMlrQ4oCv59YI8ddqgzx1HVjLU1vKeY4icefTCaftaybijKvaWnLWk2lQuB6ydHL uv4Ry94HWkTT0oQQXV+i6S0uqpN6b2NTFKV00B1pXarLkCWEiKJIURSIDAjWaJrGsmyx2WC8aZrh cBiyQOyskWJAeinEgMCbUkvCKchBcH5AJ+FyNuPkk58nKwgC7CXGSidtT7MFxJHGUI1AqdPG2mx2 bza7N5l81jSvhcOFv1nS9N124VqO62aYewThnN//ZsHZQKBQ1EZNW7SsVV3/yN1wTI0qbWhoyO/3 lxjAsmwoFPL5fE8//bS9TRRpQlRVhYRQTdPKf/pWlyrBMEw8Huc4Dm6nKIrn+VgsVmw8TdOpVIrn efsRzvO8KIolnuiiKIqiqOu6z+dTVXXNIvEgQYplCMJsgiD4fD5d10EwlfNOc3DeBY2ONU2DvcRV zIYgSPNTTQ5KXTfWgu7R9Y/WHBkOJ3X9I1n+fDT6WCBQtAw2zz+kKPPgX3HR61O7Stu9e3fpAQzD yLIsSdI777zzT//0T5Ikbd269Stf+coTTzzBsix+ZWweeJ53psj9zu/8zu/93u8xDFMid7UWGIYp pkgKephhp8+a09r3UhQFrZXLmZw48mRzxtjHBWerGkjapWk6Ho/jbh0EaWOq8aDUKUH6zjbjNwgh LHt/GWZ8RAhh2ftM81qJYSBKVPUSw9zrotenMeVPRFEcGRlZv379J598Qgj5+OOPX3vtNUEQAoHA b//2b1f6lR2pH7IsP/744+vWrSOE/OIXv1AUJRwO+/3+3/3d38Ufk4tMTU2Fw2GWZVGdIEjbU7EH xcXyJ8XguJ5o9LE1h0WjjwvCT4LBtzmuh+N6wKWRbxhN380w9xrGZZ53c8/R//pf/3P37tuSVFOl u3J4+OGH77777vz9tB988MErr7xy+/ZtQsiDDz741a9+tRlyPzuZM2fOPPTQQzk/KRArcPzggw8+ /fTTL774Iv6kqgAU3g9+8ANZlrF7UafR29u7efNmr61AGk3FAqUeG2shy7X0JecY+7jgbqv82Vj2 PsO47KLZhnHlM59Z+j//55duTViabDbr8xXYb3X79u1169bdvn37gw8+ePXVV7H8QzlMTU3Zqcfl p/UYhlG7C+SDDz74wQ9+8IMf/IBl2eXl5Rpn6zTg1/uv/uqvvvnNb3ptC9Jo9u/f77UJiAdULFBc L39SbwzjCmSzuuj10bTFV16Ra9zCc/z48T179nR3ryGbJEl6++23nerkrrvugojPU0899fTTT7Ms a38jb1gpw9blF7/4xb//+7/nny8dL3BepSiq4OCPP/74j/7oj375y1zZCj+v3/qt39q7dy/8sCiK WnMrDVKQL37xi16bgCBIg6hMoLRif0hB+AlNb3FXVLmi0hYWFlZWVkqP0TQtpzrhI4888sILL9Qv AbPtURSlThGWl19+2W5e6BSRzz//fIltwAiCIEhBKhYoreU+IYRUXUitGA1TaYZhfO1rXyOEPPDA A88//zzDMPbm0sYzODhICLGb7zTJVM2Dpml2XyH4ednOkoLjoTZJ1cvldw/2qgBdOdXbXAG32SNI p1GBQKlT+ZOWo2EqTdO0V155BXcUNz+maQqC8Nxzzz3zzDMcx5Xz8ypWuL0JcV1qVEeNzQGQlqbM gDjSZlQgUBqzsbbJaaRKc7F0RCuSTqdb6O9RpUmvLMsqimJZVi0uMc9FQyMxTRPDZB1LOQFxpP2o YJdsGzfGKx9NS7v1IRw4cKCnp8eVqbwiEom89NJL0AD58OHD6XQ6k8kcPnw4pysyjNy1a9fg4OCu Xbuct9sn7abHcOOhQ4egoXFLUIWLK79wexX4HBBCVFX1+XyCIEADHWipQwiRJMnv90Ode0hpsizL eRJyq+15nMd2Mq8kSdA0uGA/5EAgAFPlN2fIX706DMNA9wmCdBTlChTDuMIw93ZIi+cSqOpFtwRK IBBo5t5XaxKJRKanp4eHh2dnZw8ePDgzMyPL8szMTCaTef311yHFZHp62h45Ojo6Ozs7MjICt584 cWJ6evro0aOzs7M7d+6cnJy0M0wJIYlEor1bHlIURdO0u72WeJ7neV5V1VAoBN19KYpSFEVRlHg8 ns1mOY6TJAmK3iqKIstyNpvleb5EIwW7Sqwsy1BuLpVKgefGqTai0agsy5Zl5UiQgqtX9+50XcdY J4J0FOUKFFW9WL/ugK2CaV5HlWYD4mN8fJwQMjo6SghJJBLDw8N9fX2yLDs9JTASxsB4Qsjc3Bwh ZN++fYODg+AscTYDm52dbW+BQghhGKaWPFlCSNYBnIlGozRNG4YhyzI4aUADQWdBVVUJIXZVfmgB CDKlnOgJx3EMwwiCkN9DimVZKJ6WI7kKrl7FOwX9hCEeBOkoys1BMYzLdpe+Yvh8p3LOxONPerIt GSzJqdhW8GRFqOrFlttlXT+6urquXr2ayWS2bdsG5VO7uroOHz6cSCQOHjwoSdKLL75YbKRNm+3l qYh65MkahgEKANry2ecLZqtABoxpmmWmd4TDYV3XZVmORqOBQCB/NlKkA2LtuTIoUDqcNgiII1VQ lgdFVS9xXBP9ckDXnsavaxhXXMzCSaVSN27ccGu2xjDoAII1k5OThBDIIBkZGUkkEoSQ/v5+pzsE RsIYGA9jCCFnzpwhhEQikcOHD+eX829vOI4zTbMWjeLMQYGMEEEQoNWfruvgJoG8DVhFkqRwOGxZ FvhOYACcdE5bLFME3CEsy+Z7QUzTBAcJzGxTcPUq3qmqqhzHVXEj0h60ekAcqY6yPCgV1SWrxUXR zGhaek0fUkVMTU0dOHAg/5toqwAhm5mZmcHBwa6urrGxsfHx8e7ublmW9+3bNzw8PDw8nMlkMpkM hHWmp6chZ2VgYABuz2QysixPTEz09fVJkrRt2zaP31JjgcI2giC4VXNPkiTDMERRFEVRVVVFUTiO E0XRsixBEMLhMMMw0WgUFAwhBMbQNA0nRVFUFMXn84mi6ExHZVlW13VJkqLRqCAIwWCQ4ziO4yzL siwLRobDYcMweJ7P6ZJTcPXq3lc5DZkRBGknCnR4ycE0r0vS+VgsuPZceV6NbHavql4ShJ/w/MOy vCMQ0AkhqRRLURsk6byqXrKsmxS1URS38/xDijJvn+H5h0ASOeMycMyy90MfY1HcLorbBeGcpv3m y7oobnfeRdNbTPMaRW2MxYIse59zqpzVRbF3zXcnCOdEcbuLG4wnJiZcFyg+n+/YsWNtn71RNYOD g14VNCuIZVmBQIBl2Vgs5rUtTYphGMFgUBTFDt91jzQ5uq6HQqGO2vnfANYO8dRYOJXnH+L5h1X1 Yij0rmXdjMefpKgNirKgKPPx+JPZ7F6O65ak9ycmfqYo87K8I5vdC2LFMK4UnDAefwIOZHmHpqUt azWVCoHsUJR558ho9DFZ/rxl3VSUBef5/NVN83rpd2FZq4ZxGYvUIe5CUVQ0GtU0rVWKtjUeQRAY hkF1giAdyNohHk1brKhafH6IJxp9TNeXDOOyLH8eoiS6vkQICQbftsf87d/+C7nTnViWd5QZUeK4 btO8JgjnDONy/lWWvY9l75Ok98HjYpO/umleKy0+oN1gOSYhSEVAuMTFQE87AcGdZDLptSGIx9TD 34w0P2t4UKD8SY1rGMYV07xG7igDm2x2r/0PzljWKiHENK/r+hIclyYcTirKPMveV0xCwSQUtTH/ knP1NV1ELpY/sent7d28ebO7cyKtCGRXQIE1r21pIiDPVxRF3LyDIJ3JGgKlivInsMUG/knSeUKI IPyEojbK8ud1/SOItoAg0LQ0IUSSzofDyX37/i9yJ0YDZ3KmzQnTAOAaYdn7QADlYJrXVfUSueOY sclfvbQYqlP5k/3797dQKXekflAUFYvFdF0PBoPulm5rUaDQbSgUYlkWgzsI0rGsEeIpp/xJaSTp vGFchkRUVb2oKPMc1y2KvZZ1UxDOhcNzDHNvNPo4Td+9ceM6Vb2kKPM0vSUafQwEgShuV5R5n++U KG5n2fthTsiTlaTz0ejjgvCTYPBtjuvhuB5olENRG2BkODxnGJd5/mFR3O40KX/10uKjtcqfzMzM QA00pIVgWTaZTAqCEAqFYBtOx4Z7dF0HZ5Isyzl7ghAE6ShK7eJR1UuWtVrODpf2JhR6z87MbXJY lr19+7bXVjQv69atUxSlmUMGUBsekmebZ7dRY4BK+YqisCwLJXG9tghpFpo/BwV38dSDUgIlHE7K 8o4O37qiaWnDuFx+GRgEqRHTNAVB0HWd4ziaplmWZRimjR0qhmFAAVzo9QMOJK+NQpDKQIFSD4qG eGDnbYerE0KIri/lRIjc4vTp0wMDA2384EGqg6bpeDyuqqqu65qmQV1XmqYZhgG94rWBLgD19Q3D sHNuWJblOI7neXScIAgCFPWgKMoCRW3ISS/l17ZLAAAgAElEQVTtNCxrNRR6t6Jd1uXT/E5LpBmw LEvXdeMObbPTh2EYkFwMw7SH6kI6GfSg1IOiHpRKy5+0JVj+BPEciqKgVorXhiAIgjSUwtuMXSl/ 0gbUo/wJgiAIUhGnT59uG98hUj6FBUoV5U/ajzqVP0EQBEEq4uzZs8vLy15bgTSaAgIF+s6427m3 Fal3+ZOhoSG/31+/+REEQRCkdSkgUDQtzXE9jTel2TCMK3WN7+zevRu38CAIgiBIQQokyWpaOhYL Nt6UpkLT0uhDQhAEQRCvyPWgmOZ1itqAiRe6voRZOAiCIM0ABsQ7k1wPCu5bIXeycLBIHYIgSDOw e/dur01APCDXg6LrSyhQGlD+JJVKTUxMpFKpuq6CIAiCIC3KpwSKri+1UNve+oFuJARBEATxlk+F eDQtTdNbdH3JK2uaActaxfInCIIgzUMqlVpYWLBf9vb2OpuEzM3NOaukOHucWZaVSCTsS36/v7+/ 335pWZbzxp6enk2bNjkXddrgXDHnxpz8mHQ6vbKyUmxapHw+JVB4/mHLuumVKU0CRW2oU3dABEEQ pAqWl5cXFxftl36/3ykXcq6urKzYAmV5efns2bP2pd7eXqdASSQSzqs5zdGmpqacNhw5cqTYjUND Q+vXr7dfvvXWW0455Zw2lUo5p+3t7d2/f7/98vTp0yXsmZiYKGZPzo1DQ0POrJ3Tp087P589e/Z0 d/8mRJBOp9966y37Uk9PT1Ol+xRtFojUFfg1xWaBCIIgbYCzWWAJD8qNGzecWmHz5s22ViA1eHTq 5GHyHBQoCIIgCFITbdbNGDSN596Uwr14EARBEATpTHKiZl6BAgVBEARBkKYDBQqCIAiCIE0HChRv uHHjRiqVunHjhteGIAiCIEgzggLFGxYXF6emppohyIcgCIIgTgYGBvbs2eO1FYW6GSMIgiAI0rFQ FGVvRfYQ9KAgCIIgSMVomqaqav55y7IEQWi8Pe0HChQEQRAEqRiWZQVBePnlly3Lsk/quv7QQw95 aFU7gQLFG/x+/9DQUE4HBwRBEKRVoCjqkUceee2115566qn5+XlCiCRJoVAok8mwLOu1dTVhWVY6 nfbaCqwkiyAIgiBVIUnSd77zHZ/Pd9dddzl3ZS4vLzdDDkfVQPseZ58gT0APCoIgCIJUA8uyt27d +uSTT5zq5JFHHmlpddI8oEBBEARBkGrID+WsX7/+hRde8MSY9gMFCoIgCIJUyXPPPed8eevWrVZP QGkeUKB4QzqdPn78eDNkISEIgiBV88wzz9x116cqirWBQPH7/T09PV5b4UahNkVZsKybtc/T0ljW KsPcy/Pl7i5bWVlZWFhYWVmpq1UIgiBIXWFZVpIk++VTTz3loTFu0d/f77UJhLgiUDRtMZkcqn2e lkYQzjHMvV5bgSAIgjQUhmG2bt368ccfE0LWr1///PPPe21R+1BriEfT0ix7nyumtC6WtWqa1xnm Hq8NQRAEQRrNV77yFTjABBR3cUGg8PzDbljSwqBKQxAE6VieeOIJONi6dSvDMN4a007UJFAsa9Wy Vmn6bresaVE0LV1+9gnQ09Nz4MCBZshCQhAEQWrB9pq0jfsklUrNzc15bUVtAkVVL3Fct1umtCim eZ2iNlDUhoru2rRpUyAQ2LRpU52sQhAEQRoDTdOf/exnCSFf+tKXvLbFHRYWFn760596bUVtAkXX l1CgqOpF/BAQBEE6mVAoRNrIg9IkVC9QDOMKTd9dqeeg/UCVhiAI0uG8+OKLhBCapr02pK2oXqCo 6kVMj9X1perSY2/cuJFKpZztGxAEQZAqkCTJ5zXhcJgQ4rUVRQEHT8tRfR0Uw7iMG2ur3sS0uLg4 NTV14MCBQCDgtlEIgiCdRV9f38GDB7214de//jVkojQbMzMzUKalfAYGBpqhjmiVAkVVL3FcZTtQ fL5ThBBR3C7LOwghknReUeYJIdns3tK3ZLN77YPSwyqyp3aw/AmCIEgz0NXV1STFT5uQubm5SgUK RVHN0JC5yhBP1YkXur4EB4Zxpbqlmwcsf4IgCIIgdaIaD4ppXieEVFf+xDAu5xyUgyhur/pq/dD1 JfAGIQiCIAjiLtV4UGrxHLDs/bq+ZJrXnZ1rLGs1HE76fKfgnySdz7lLUeYhHmRZq5J03u9/0+c7 5fe/CSPtq4QQSTofCOgwTzicBC0FL2EAHOv6Uij0Hhz7/W+q6qVK30gtKo0Qsnnz5t7e3s2bN1d3 O4IgCIK0N9UJlMVKC6fa0PTdur6k60vO/cmalras1VQqBHkkttrIB7SILO/IZvfy/EOKMu8MFUFe C8d1Z7N7ZfnzmrYoCOcKzhMKvavrH8Vi/dns3uq8IDXGd7q7u/fv39/djfuTEQRBkObi9OnTx48f 99qKygWKYVyppW0vTW8xzeumec3pe+C4boa5RxDO+f1vlr4dtAvII5ApzhxVuAqCQxR7CSG6/lHB eWKxforaGA7P+f1v6vpSFe+oFpWGIAiCIEhpKhYoNZY/YZh7dH0pR+WEw0lFmWfZ+5LJoXImsaxV QohpXtf1JTgGKGqjfRX+C2fy4bjuVIqNxfo5rlvXl0Khdyt6FzWqNARBEARBSlOFB6Wm8icMc69l 3dT1j5whHvBzsOx9pnmt9O2gjcBTIknnw+Hkp68+BOfJp30tNoqyAAeCcM7vf4Omt8jyDsu6aVk3 K3oXWKQOQRAEQepKZbt4qih/kgNFbaCojZZ105nAEY0+Lgg/CQbf5rgejuuBJskFi+hD+EZVLynK PE1viUYfcw6D7Tyalvb5TlHURrvmiihuV5R5n++UKG5n2fvhjGWtBoNvO28sn9qL1FmWlUgkBgYG mmGvOYIgCII0G75sNlv+6HA4Kcs7qt660h6o6iXLWoUcl6pJpVJYSRZBEKR2JEn68Y9/fPToUa8N aVIikUg6nY7H414bUjEVhHhq3FjbNmB3QARBEASpNxUIFCycSlClIQiCIEhDqCAHRdMWy9xl08ag SkMQBEGQBlCuQMGNtYBbKs3v9w8NDfn9/tqnQhAEQRAXsSxrZWXF81Ki5YZ4cGMtcVWlURS1e/du 3MKDIAiCNBuJROKtt97y2ooKPCi1bqwtiGWtKsq8pqVN8xpNb+G47vILz0N7HaiO3xhQpSEIgrQ3 g4OD9nFfX9/Y2NjOnTs9tKeTKcuDUnv5k4JY1moo9K6izEN9NtO8pijz+Z0Cm4c6qTQEQRCkCblw 4cLExMTc3FyxAYODg05BUw5V3NKxlCVQNC1dj74zijJvGJcZ5l5oE7i8/BzL3q8o887q9c1DnVQa giAIYiNJkqIohmF4a8bs7Ozs7OzY2Bgh5OTJk16ZkU6nvVq6GVhboJjmdYraULCua41oWpoQEo0+ Drt2KWpDNPpYPP4kIUSSzgcCus93yuc7FQ4nYXMvnPf73/T5TjnbCjpP2sXsXcfd8iepVGpiYiKV Srk1IYIgSBvAcdyRI0eCweDnPvc5QRBUVTVN0ytjRkdHCSHgQYlEIrt27RocHNy1a9eJEycIIYcO HYJhkUik4IBMJuM8GYlE8m956aWXwKdy+PBhW47AmUOHDnmljXp7ex999FFPlnaytkBR1Yt1qksG kR1n0ISm72bZ+xRlXlHmOa47m90ry5/XtEVBOEcIkaTzijIvituz2b22R0dRFhRlPh5/Mpvdy3Hd kvS+rWZcNRXLnyAIgtQdhmFeeeUVQsi//du/qaoqCEIgENixY4ckSZqmNdiYbdu2EUKuXr164sSJ 6enpo0ePzs7O7ty5c3JyMp1O27Vrx8fHCw6Ak+Pj47OzsyMjI3Bs3xKJRKanp4eHh2dnZw8ePDgz MyPLsnP1RCLR39/f4LcMBAIBr5Z2snaSrK4vlZ+4WgWmeT3nwQ99/u600emVpPehmyCchxrzsrwD Xur6EiHE7qpDCDHNa64rCSx/giAI0hh4nn/vvff+7u/+7pNPPoEzP/vZz37xi1/cunWLEPLUU089 /fTTHMcxDNMwk8CJsm/fPvvM4uKicxduwQHT09OEkJGREULI+Pi4rU4AuAonR0dHJycnE4mEc8Ds 7Kzrb6S1WEOg6PpS/R7MHNejaYvh8Fw0+jj4UcLhpKYtdnVtuHr1N/0CIR+FojbCfy3rpvO8Tb33 8hw//o+/+MX/LQguT/uXf/mXLs+IIJ+GYRjczY60HJ988slnPvOZW7du2d3iQJ0QQt555513331X UZS77rrL7/f/xV/8xZUrV+pkBkRq+vr64OWaiqHggEwms23btnQ6vbi4aE9FCOnq6rp69SpczWQy cMY109uCNQSKpqXrt7FWFLfr+pJhXHb6P1j2foa5B7bzRKOPgZsEAjo8/xBEf2R5h73Zh2Xv0/WP NC3Ncd2SdN40r+e0OK4dw7jy5JMDk5Nudlr68MMP33zzzS996UsPPPCAi9O6jmmaHgaAEQTpTK5c ufLuu+8W7GW7bt2627dvE0I+97nPffWrX6Vpuh5/o5wbbUZGRjKZTCKROHPmzM6dOyORyOLioiRJ EAAC+vv78weMjIycPHnyxIkT4+Pjk5OTc3Nz//AP/+Ccdnp6enJyUpIkUELga0FsSgkUy1o1zev1 21jLMPfE408qyryuL1nWTYra6KyDomlpn+8URW0Uxe1wEv57J0Olh2XvJ4SIYq9l3RSEc+HwHMPc G40+7no+r6pe/G//7VHXPwdIv0IQBEFy+MIXvgAqJIdHHnnkhRdeYBiG4zj7pK7rdTKjp6dneHgY dEMmk5FleWJioq+vz1YnAwMDiUQiEomMj4/nDzh48CAh5OTJk9PT0z09PXDSvgUeATMzM4ODg11d XWNjYzkxIA+Zm5tbXl7evXu3t2b4CkpUQFUvWdYq5Hx0MsHgWWxChCAI0hhgp7H98oEHHnj++edZ lmVZtmC8UpKkH//4x3bKKpJDJBJJp9PxeAVBgNOnTy8uLu7fv79+VpVDKQ+KpqVjsWDDTGlOsPwJ giBIw9A0TVGUrVu3siz7pS99iWVZmqa9NgrxhqICpX7lT1qLem9iQhAEQWwsy0omk43coYM0LUUF Sv3Kn7QQ9St/cuPGjcXFxZ6enk2bNrk+OYIgSIvC87zXJiDNQtFCbe4WTm1R6lf+ZHFxcWpqanFx sR6TIwiCIEjVDAwM7Nmzx2srinhQ6lr+pIXQ9aV4/AmvrUAQBEGQxkFRVDPUTyrsQdG0dKWZoaHQ ez7fqWbuRVwphnEFa9sjCIIgiCcUEChVlD8xzetQjR76/5UGWgCWP3mZuD6tpi1ikAtBEARBPKGA QKki8QJ0CcPca5rXytEoLUFd41x+v39oaMjv99dpfgRBEASpDsuy7NbKHlJYoNi9gstE0xZpekss 1k/udO8DnF4NOA6F3oOXEAySpPOBgA6XwuEk7JqBl3De73/TOYAQYlmr4XASzthBpSqmLR2Nqnd3 QIqidu/e3QxBPgRBEARxkkgk3nrrLa+tyBMoVZQ/Mc3rhnGZ5x+m6bs5rqe0B8XOOYV+Oooyz3Hd 2exeWf68pi0Kwjl7JBRvtaybsVj/8vJzhBBNWySEaFraslZTqRA0CIRmPeVPG48/mUw+a1k34cZi 1LUJEYIgCIIgpckVKFWUPwFFIknv+3ynNG3Rsm6WGeUBiQBl0KCgPiSyALZIYph7nIKJ47oZ5h5B OOf3v1nFtDR995rpNZa1almrmCGLIAiCIF6RK1AM40rlAiW3mIczylMCitpICLGsVfu/cKY04XBS UeZZ9r5i/XGqm9aJql7C9FgEQRAE8ZBPCRRNS1fas9cwrkB8J5vdC/9oeku+B0VRFvLvhUwXyAUB t0c5uS/gDmHZ+0zzWsEB1U376SXqXqQunU4fP368GbKQEARBEMSJ3+/v6fG+Cd2nBIquL1WaeAHu E2c+Kcd121EeUdxOCPH5TlnWTZa9n2XvJ4TAfyXpvChuF8Xtur7k851S1UuiuL2crjfR6OOEkGDw bWjjx7L3g5ukxmltoPxJvZsQraysLCwsrKys1HUVBEEQBKmU/v7+3bt3e20F8WWzWTiyrNVQ6N1i cZPOQZLOs+x99S6km0qlpqamDhw4EAgE6roQgiBIeyNJ0o9//OOjR496bUiTEolE0ul0PB732pCK +c9S97hvBcD2xQiCIK1FIpEYHBz02ormhWVZr02ohv8UKKp6MR5/0kNTmoF6lz9BEARB3IXneQ8f wP/6r//653/+57/85S8PHjz43HPPeWVGaVq05tZvQjymeV1R/v/27uZVbrL9A3jG1h8tnBYyuLDP wRcydWG1PJZMjwgWRBLahSsh2biyQvInJPgXJJtCV5JARexCmWx0UWxJsLbSTU1A6aZq51YXp3U1 t7SKlZ9lnsVl7ydPZibzcjKTefl+FmVOTia5JzM9ueZ+ua4fg+DfdbenZqaZet6xBSwwfvjw4e7u 7vb29oEDB+Z9LgAAmIckSUzTVBTl+PHjn3/+eZqmiqLU3agKdLvdXq938uTJepvxzyTZMPwZPQeL TH9y4MCBVquF6AQAYEW5rqvruqZpcRyfO3dOlmXbtutuVDXu3Llz69atulvxOECZIf3J+kH6EwAA GIsx1m63fd/3PK/T6ciyLMtyp9NJksR13bpbtz6ekGZKf7KWFpD+BAAAVprv++12W5KkNE0dxxHb VVX1PM/3/SiK6mvdWnlCmin9yfpZTPoT4eHDh91u9+HDh4s5HQAA7BHn3DRN13Uty4rjWFXVwg6O 41iWZdt2lmW1tHDNPMH5/2fZb6g7E0W7i+w+2d3dvXDhwu5usUoAAAAsoSiKWq1WkiRxHHueN2pd jOd5iqKYpsk5X3ALK7Szs3PmzJm6WyHtj6K7svx/lBh+kyH9CQAADGKM2badJIlhGEEQlC/ZlWU5 juNWq6XrehzHK7q+l2bV1N0Kab9lPTdtqRoAAIBN4Pu+7/sUdkyYbYV21nV9pWOUZVCsZgwAAABJ krTbbZpxkqbpVLngVFWN45gxpuv6So/11AsBSj0OHjx49OjRgwcP1t0QAAD4HzSmo+u6LMtpmpbM OCmRj1EwZ3Y2/y0WCAAAsMk45zSmoygKLcnZ4wEZY6ZpMsY6nc4KFcS5cuXK7u7u2bNn620GelAA AGDTcc5d1221WmEYOo7T7Xb3Hp1IkqQoCk1e0XUdOdymhQAFAAA2Vz40sSyr2+16nlfh8SnJrOM4 vu/rus4Yq/Dg6w0BCgAAbCLG2GBoMqdFN57n0ZQUypE/j1OsHwQo9eCcX7lyBbO7AQAWj6oQt1qt KIrmHZoImqalaWpZFlUZxMzZsRCg1KPX6127dq3X69XdEACATUFzYCmLGmMsCILFhCaCLMvUlcI5 b7fbtm0v59fU06dP1z5DVkKAAgAAay+KItM0m82m67rUk0GdGbU0hhrgeR6lz3dddznDlNohQAEA gPUURZFt281mk9b6BkHQ6/WCIBis87d4Yq1QGIbNZtO2bcyfLUAelHp0u90LFy689957rVar7rYA ACyXJEnEY1mWp4onGGPJY5xzVVUNwzAMQ1GUObS0ApzzMAzDMGSMFVrLOY+iiF6IoiiKoizzC6kc ApR6cM5v3ry5s7NDY5/ffPPNrVu3xG+PHz9+8uRJ8SPlzBE/njlz5l//+qfw8t27dy9fvix+tb29 ffr0afFj4bD5J0qS9OGHH+ablB9xRHvQHrQH7Vlke954440oirIsy7JssCNBURTP8wzDGNUexliW ZdeuXbt06dLPP/8sSdKxY8dee+21999/X9zOl//63LhxI0mSMAwLL//kyZNvvvlmlmUUt+3s7Kiq 2m6359qenZ2dl19+WarV/npPv7FkWc5/MprN5vb2dv7H/M6FH/MJ8g8ePFj+xPxvC5n1878qQHvQ HrQH7VlAe+7cuXP9+vWvv/763r17kiRpmmYYxksvvXT//v2jR4/Snj/++OP169dN09Q0TVXVZ599 9v79+5Ik/frrr/fu3bt48eJ3331HczheeeWVnZ2dd95559SpU9SAfGfD8l8f6jsJgiBJkk8//fTB gwcvvPDCiRMntra2xP0iDMOPP/74gw8+OHLkyNmzZ0+dOjWn9vz5558DV2LR0IMCAAALxRgLwzCK IsYYDVtomlaeCZ66FmiwQ2xUVZUGPih22Zy6wVmWua6bJImmaY7jrFAS/akgQAEAgAWhuCRJElmW LcsyDGMZ5quuqCRJXNfNssxxHMdx1i8+Q4ACAADzRZM9fd9njGmaRqFJ3Y1aE1TdUJblIAjWrCsF y4wBAGCOKJ28bduapnW73TiOEZ1UyHGcNE0VRVm/eoToQQEAgLkIw5B6TRzHsSxrc9bH1sL3fUpD 1+l01mO4BwEKAABUTMzitCzLcRyEJouRZZmu61Q/eQ0m92CIBwAAZjEq86nruu12m3Mex3EQBIhO FkZV1W63K8vyehQjRIACAABTo8J7hY1ZlrXbbd/3aWLEms3ZXAmyLMdxTFNSCjHK4Pu15BCgAADA 1MIwvHnzZmGLruuSJFGJ4JraBSNjFMqmX2PDpoUABQAApnb+/Plvv/1W/Gjbtm3blmXRipIaGwaS JMmynKapYRgiRqEaAvkiR8sPAQoAAEwniiJKTk+pXU+cOPHJJ5+8/fbb6DhZHlmWybL89NNP67rO GIuiSJKkL774ou52TQGreAAAYDqvv/76jRs3JEk6duzY999//+jRo8OHD3/55ZdrsHJknbiu6/v+ vn37ms3mk08+effuXUmSVuimjwAFAACmwBhrtVr5LVtbW1999RWikyX07rvvfvTRR/ktnU5nVRLl YYgHAACmQF/K6XGj0Th06BCik6V17ty548ePP/HEP/f6/fv3r9A0FPSgAADApDjnzzzzzB9//CFJ UqPR2Nraunr1KqKTZcY5f+655x48eEA/bm1ticdLDj0oAAAwqTAM//rrL0mSGo3G888//8svvyA6 WXKyLF+9elX8+Pvvv69KDjcEKAAAMKnz58///fffFJ3QOpG6WwTjqaoaBAE93rdvH63oWX4IUAAA YCJidfH29jaik9VCRZEkSXr06NFnn31Wd3MmgjkoAAAwkRdffPH27dtPPfXUDz/8gOhkFb311luX Ll2SJKnb7S5/Pj30oAAAwHiMsdu3bzebTUQnq+vixYtHjhyRJGklRnnQgwIAAOPZtp1lWRzHiE5W WpZlr7766qFDhzjndbdlDPSgAADAGJxzRCfrQVXVy5cv//bbb8sfoKAHBQBgvrIso1JtWZatUJqs tSTLspqz/PMw5iQMQ0mSLMuquyFlEKAAAFSPMRaGYT4i0TSN7ogbe1NcBiJSpFwgFK9ommZZ1hJ2 DommziNzSa/XO3z48P79+ys/cgF95uk6T/VEBCgAABXzfd/3fVmWDcOgP83IZraEkiSheCWKIlmW gyCY9g46D1EUiaCERmHo87OE8dOEBl+Loij0X6P8iQhQAAAqwxizbTtJEsdxHMdZ3ZvKRlmSd41z bppmkiTiLj5Dr8PSyvcGUYcQXe2y5/QBAKAKnufJsqwoShzHdbcFplbv29fpdDbnw9Pr9Sg00TSt 1+uN2g2reAAAKmCapuu6lmWlabo233o3iuM4aZoqiqLruuu6CzsvdZyYprk5Hx5Zlj3Pi+OYMdZq tUYmZVlk0AQAsJbo6+AmfPfdBIt8N6njRJblzfzwiK4UwzAGu1IwBwUAYE+yLGu3247jeJ5Xd1ug GrquM8bSNJ3rfBTGWLvd1jQtCIJNnq4URZFt25ZlFf4HIUABANiTdrstSVKapnU3BCpDocPgLbNa iwmDVkIYhrZtx3GcH+HCHBQAgNm5rptlmahlD+tBURTP83zfn19iPTr4hvedCJZlGYZh23Y+vy16 UAAAZoTBnfVmmmaWZfPo4VhMD81q4Zy3Wi3DMES4jx4UAIAZ+b6vqurQe4zv+41GQywGCcMw/+MM Wq0WpScv0Wg0Go3G5Nv3Qhyz/ODzOPVsZmhJEAScc9/3K2+MaZrUSVP5kVcX5coLw1As6kEPCgDA jBqNRhAEQwua0NdBSZK63a4sy+12O8uybrc7W55713XDMOz1emPbI0kS/VUf9bgq4pjlB5/HqWcz W0toCC+O4wpb4vu+67ppmiK/8CBKVUcf9bkn4QcAWEuUDXPUPUaWZcuyfN+PokhRlCzLLMuaLTrh nIdhOEldtzF5OSVJkqQsy6q9L5afdJImLTNFUSrvQUmSxDAMRCdDeZ7XarWSJNE0DXlQAABmQf3z JTvQt0BFUSi26Ha7tF0kU6d0Vf1+P794gTq688fJp+Wgmiz9fr/T6UiSZFlWv983DEOSJNG/0u/3 xdEcx+k/7jMQ4RFtFPK/pWSmtMUwDHoVdPzCc8W56AEltBCva3A3x3HyRxZXQ+wgHpdfjVEnGnr8 /J75lhSufwlanJWmafluU8k3GwaJ9wUBCgDALCzL0jRt7D50X6RIov84rKEbnghcaJ9Op9Pv94Mg KNyS6as2pbEST6f7saIo/X6fqhL2h93s84+73a5YC50/vogw6EGapvlYh0ro0f1+6PFF4CJJErWc HtNrzP+W7sr0EsSlGzxm+dUYeqKhx6eNdKsTHTlDr3/5myhOVwl6uzczLduEqLh0HwEKAMBsVFUd +z1YBB/iLjiYyDyOY0onSt/pDcMofF/P38IpwvA8j85OTxe34fIAZfBxYcvQ3ajHQtO0fCfE4LMG D5vfrfDb8rZNfjVGbSw56dDrP9jsPIp1yveZHHV9lRSgAcdxVFXtoxYPAMBssiwbO6dE7FDYM//n WNM0GpLodDqGYSRJouv6qANSkVuqB2sYhizLtLQnPwpTLdM0fd/XNG2STHSUxIIxliRJPqEFBRy0 hf4tWbg7ydUonGjo8Qc3CoXrX/6iVFWl+UaVoDlAyH1SQlxwBCgAAFOj/F0zzHOk2yEtpHRd1zRN zrlt281mk9adcs4Ld1N6CmOMfjQMI9n5ce8AAAZhSURBVIoiWZZVVdU0LYoiilr2/qKGoleqaZpo wFA0XELzSel1Df6WVlnTPoNzfsVc1PKrMfREQ49f2JOePvT6l18BRVEqTNdW+STl9UPXJ0kSDPEA AEyNBlYm2ZP+5ua3iEmaqqrS+EW32x06EZXQtAkxDYLGCGgOR37uReFcdCfOT5Id1R6xZehuImuW YRiGYWiaRsMThWf1ej3Lsuh1KYpCM0jyvxWTWAtTRMXsEBpIovkuJVdj6IlGHV8cnFouLlrh+peb /L2exCQzc4GuEvKgAABMjYYeFvP3k3PebrcVRak2GwdMqNr3utFoFCrOLEaj0VAURWTFXZ78NEPp uq6qKoZ4AACWGi21TZKkwpkQsIEYY/PIijs/CFAAAJadpmmdTmfsbAmAcmEYrtCnCAEKAMAKoFkU o37rum6j0bBtu/LzDi1hszwVdqbium673V6hO3TlVFUtdKK4rttqtegNNU2zfCr0giFAAQCYu3nf 0emus+p55efNcZyVG+aoluM4vu+LKMR1Xd/3Kcuf53lRFM0jxp0ZAhQAgKnRgpE9rj6t/Nvq/BYb FziOU0kwNPMVmO2JlPktf4eeRJZla5O2hJK0ihCNHtAyMXpDK1xQvXcIUAAApqYoiizLE85aFanG KBsH9abouk4TAkzTbDzmui7n3HXdZrPZaDSazaZI4JHfWOgDEMenB0M77fNdOIOPaX9xuvy5hr4i 3/epDUOfnlfYge5/+SswqsGiDeJXYy/d0DPmjyw9zoMyVSdK5ZlL6h1jchyHLrs0ZQK9RfrnEtW6 1BkAYFWJiiGToD+7+ceSJMVxPFjpZmitmbEVZMTxR1W9GWxA/nG+TM/QEjajXs7g04fuGcfxYGNo ++RldMZeuvwZR5UWEhUcJ3zj+v2+oigVprqv9miTy18ikSWPri19jAs5depF/wUQoAAAzEJUDJnE qPhgsNLN0Nv82AoyhVt4+fbyx6OOMPZ05XtO+KzyNpRfuhle8lgU0Ii8c3tHk52rOtrk8q9alIgq SaBXI8r3k6YphngAAGZRSYmWUZVuhha1yf8RH7WiZ++d9iUlbPZiVGNKGlzehqmKBM2M3uIKh3iq rewzOfrY0GPRgUQJW6kXqtfrUQ9W7cQ1R4ACADCL/1YM2YPBSjdDa81MXkFmbNWbsdMvhpaw2SPG GM17GCzBM3kZnYIJiwQV0M1v8tqKSZLIslzh7GNVVTnnS7Wad9kwxv6JvxfcdQMAsDYm7xUfVRln sNLNTz/9NLSoTXkFGXHMsVVpRMmbwhPzj4eWsBl6uqFPH9yTgjnLsgp1fEY1mAru0G7iwdhLN7RI UKFt1ElQGCArMY8RGanSMaP1o6oqfQxQiwcAYEau64ZhmKbpwtb3rpzZar5kWcY5V1VVlmXOOS0m quRuRVWNqODiWFSFp9PpTN7jMmEbNE1bkvGUJdRoNIIgsCwLQzwAADOiXo2lym21HqIo0nVdrHmW ho0NzSBJEsbYhJEB59y2beqe2fup81RVXap0I0uFxjGpyw0BCgDAjEQZv01OTjoPnufFcZwkSaPR SJLEcRwxoLMXnPMgCCbs7rJtm/bf+3kLDMPIsgyfmUGUBEjTNApQMMQDALAnGOhZP1EUmaZZ+eCO gM/MUIXLggAFAGBPOOc0s4HyN8Cq45y3Wi2qID2/s7TbbUmS5rpGerXQjB+afUJbMMQDALAnYqCn ZPUvrArGGFUMmMfgTl4QBFmWVbWQe9WJGT/5yUYIUAAA9oq+bSdJ0mq1aJYfrCLf96ljI47jeVel ocW0vu/Xkrdt2Qyd8YMABQCgAoZhdLtdTdNM00RXysqhjhPXdS3LStO02uqAo3iep6qqruubHNRS 0ccoioIgKASFCFAAAKohy3Kn06GulHa7jaWkq4I6Thhjoi7jwsRxvMlBbRRFrVYrSZKh85ExSRYA oGI0oE5fi2nNpKIoqqou5ns5jJUkSZZljLEsy2iExXEckat38aIoomw6QRDMad3QshH/RwzDGOw7 IQhQAADmIvtftHFUkT9YDM55/r2g2FHTtNqX++Zv2JZlrfHnhEpg0uzg8oAMAQoAwCLQt/YN7MZf Nsvcm0VdKfQhoUZSa1c6XhE9VYReXUnHiYAABQAAYInQjZzu62szk4kiLeqvoipLY5+CAAUAAGB5 rUHH24QRScF/AKeVC1U6fD8ZAAAAAElFTkSuQmCC ------=_NextPart_000_0000_01C80F9B.F234E740 Content-Type: image/png Content-Transfer-Encoding: base64 Content-Location: http://www.cs.wisc.edu/wisa/papers/security03/images/infrastructure_executable_loader.png iVBORw0KGgoAAAANSUhEUgAAAYMAAACNCAIAAADae4GBAAAACXBIWXMAAAxNAAAMTQHSzq1OAAAA HXRFWHRTb2Z0d2FyZQBHTlUgR2hvc3RzY3JpcHQgNy4wNULvQ9UAABZdSURBVHic7d1PbNvm2QDw h5LqIFkqmy4cpIicIvKK5pOAHkajs7ycCglysUsOo4Ci6EmCvKHD9kkLIAIraqUnEVjE9bCDDen0 HQaYORQd0FQTT1sDa5v5HQJIS5FFRW11Mex+Ycw4zuyZ4nd4E0WTFFm2Jb8S8/wOgUWT7/vKevno 4cM/YUzTBIQQospBewD7UxQlEAjQHgVCVkY9IxmASERks1naQ0DIgorFYjqdpj2KwYlEU1NTtIeA EOoVG+0BIIQQRiKEUB/ASIQQog8jEUKIvoGpWKNmXq+3YUk2m6VS2icjKRaL+y7sfPMujgT1P8yJ 0P68Xm9z1EOoizAnGnj4/X8IlUrF5XLRHgV6BnOigeetAwCyLHu93qtXr+q6Pj09PT09res6AEiS ND097fV6p6ens9msruv1SyRJqm+t/udIJEJeSpKk63o8Hq91V9uKmJmZIa0VCoWGQTb03sn7kiSJ NOj1euPxeKVSeV7v9Y2377H2jmRZBsz1+glGIqvheT4UCi0uLkYikc3NzUwm43Q6s9lsJpPJZDLF YjEYDKbT6d/+9reZTCYWixWLRZ7nM5lMqVRq2WAmkyE/xGKxXC6n63oulyOJWO1XRDKZjMfjpNP6 5c29VyqV9u9CkqRMJhMMBovFYjwez+VyyWSyZe9kzXA4TN5IJz0uLS35fL4D/VVRrzHU7zfZF7nv DI9BmrWpzs7MzKyursbj8XA4DACRSGRpaal5teZt69us/Vz7Qdf1bDZbLBZLpdLm5mbzmi23au69 obLe/EYalpCXS0tLzb23XNPn87XsEevZzQqFQjgcph4HsE5kQaVSaXV1FQCWlpZIJCKad3Vd151O Z6VSqVQqHo/H6XS2bzkejy8tLcXj8WQyGQwGm1cgR4LDw8PNvzrQ/j88PLy5uUmGV2uzZe/Nax6u R0QXHp0NvPo6ESmdJJPJ2n5LSiTkYCSXywGAJEnxePzy5cvw9L7idDodj8cbmm1ZzSGJhs/na3l4 ValUSP2ldpRENPfeEDKa3whpgbwdMhKe51v2TtYk69QqRwfqsXk5On6YE1mNJEnkgCgcDsuynM1m g8FgOBzWdf3q1auk4ptMJl0u19DQkCzLmUxmfHx8bm6OJESRSCSTyZCybq2YQg52JElKJpPJZJLn +WAwGAwGdV0nyQhZMx6PF4vFUChUn4gBQHPv+yZfpIVcLuf1eoeHhyORSCwWc7lczb3HYjEAIFWh YDBIRnKIHhFdWCdC6IXWJ3UiPDpDCNGHkQghRB9GIoQQfRiJEEL0YSRCCNGHkQghRB9GIoQQfRiJ EEL0YSRCCNGHkQghRB9GIoQQfRiJEEL0YSRCCNGHkQghRB9GIoQQfRiJEEL0YSRCCNGHkQghRB9G IoQQfRiJEEL0YSRCCNGHkQghRB9GIoQQfRiJEEL0YSRCCNFnnf+NOpcz83koFs2VlSrtsRzGpUuO ixerwaDN46E9lBfAoM+WTgzWjLJCJNJ1+Phj88YNg+NOvfvuaY47ybJ22oM6mHJ5t1zeleUHmcxu PG4PhxnaI7IsC8yWTgzcjLJCJLpyBUolWFx8jeeHaY/lSFKps4KwJorrAAMwdQaUZWZLJwZoRg18 JJKk6s2b1eXl1znuJO2xdEEqdRYARHHd53MMRFI9WCw2WzoxKDNq4CvW169DInHGShMrlTrrdg/l cpatX1BkvdnSiYGYUYMdiSoVePCg6vefpj2QLuP5kdu3B/uj6UNWnS2d6P8Z1deD21elYgKA9b7i 3O6hL7/coz0Kq7HqbOlE/8+owY5EhPXOfbjdQ7SHYFnWmy2d6P8ZZYVIhBAadBiJEEL0YSRCCNGH kQghRB9GIoQQfRiJEHpG0wxBWJuY+Iphbk1MfCUIa51vyzC3GObW834rCGuTk/8g64RCK+Xybuct C8Iaw9yanf22800GDs27PRYWFmRZ5jjO7XZzHMdxHMXBIKRpRiDwtapuk5fl8o4orsPTGyaOgtz8 FY2+ks9fIL1MTv4jn7/Q4cVNZBiJxNgRh9F1XdyFaUaiaDQKALOzs7UlP/rRjy5duoSBCVEhihuq us1xpxYXz7vdQ5pmhEIrorieSIwd8SokEkrm588BAMvaeX5YFNdFcWNx8Xz7Dcvl3dqlQPteE1S/ 8vHo4i5M+Q7Yhndy8+bNv/zlL3t7Ty4GJe9qd/cAeWyHSBZtmm/WfiY47lQiMVZ/l3Yg8LWiPEwk zrT8YqzflmUdPD+cSp19Ma+dswBZ3gSA+flzZH9mWfv8/DlyGCUIa7K8WS7vAADPj5A7ucjyhYX7 mrbHss92pfqFicRYIjHmdp8ol3dCoRWeH+a4k6nU2dp0ap6Kpvkm+cHvf5njTqrqY7JmIPB1Pn+h ufH6lY+evh1UJ7twJ4GJfp0oGo3Oz8/XXtbeAwDcvHnz2rVrkiQBgNfrff/99yVJkmW5VCr1aDCq uh0KfaMoW+RlubyrKA/h6RxtT9P2Fhb+70CVBdRXSKCpP2Jyu4f8/tOiuCGK6zw/bJpvplKvyvID UrIhx1yJxJhpvhmNjpJNyMr5/AXTfJPnhwXhXrm8u7h4nuNOyfKDUOibiYnbo6Ol2dlv9y0VKcpD v/90Pn+BvMznL7RsvH7lrv9NOrHvLjw7Ozs5OckwzKVLlwRBWFhYUFW1oRHGNM1AIKAoyjGNunt8 Pt+5c77r19dN838Oum3LLyJ4Ord4foSkzaK4IQj3OO6Uqm63fKJN/baKshUIlGsvicMlzKSpYnHg n9nSVwoFMxw26j+dBuTTvHv3YsNHVv8pQ1Pm0rDc73+ZfHvV5PNuEiM0zVDVx6r6eGHhfrm8w3Gn lpe/3yYnau6RZOgNjTfPumaKshUIBAAKbf9Cx83v9/v9/rW1NUmSHACQSqU0TaM4oDt37ly5cmVn Z8cwjPZrjoyMXLhwgeM4n8/n8XhKpZevX1/p4kgSiTFRXK/lRLK86XafWFw8PzFxW1G22j9b62lK 74C6SUkS5ufl9qiv8PyILD8IhVbm58+RzCgUWpHlB06nXdcNTTNY1q5pBjz9lFnWoWl79ctrGuKC omwpylY0Our3n/b7T0ejo6OjxVpp/KDaB53n++9s9lHz0r///e+/+c1v8vn84QZD7LsLOxwOkiu9 8sorFy9evHTpkt/v5ziOZdlyuQykTkS3NqyqqiAIu7u79e/hpZde+ve//w0Ar776qtvtvnnzZjab 9Xg8TqfzP7c2Ac51cTCkxKNpewBQLu+q6nYq9arbPcTzI7K8SSqOzeqrRbUsHQAU5WEiMfY0hz+T Sp0lSZamGbWUG/WPRGJMUbZUdXty8k5tIfk6EcV1QVibnz8nihvw9FOORkdJ4Zl82Txd/7SiPJTl TZ4fFoS1cnmXFJtEcV3TDFJGXFi4DwA8P1LfO2m5vZaNd/z+/mtq6rlZtt/v77idRvvuwj/84Q/f eOONWuhp2NztdgP1irWqqm+//bau63b7kypv87gVRQkEAlNTU8c8NlIbEoR7gnCvtqRNWsSyjmh0 tP5UK/nuIskzKSUmEmOCcK8hwUZ9guNOklqMomyRkjA5BUF+K8ubDHOLZR210xe1xyGSI3q//2UA SCTGNM2Ynf02FPqG407Nz59jWXs0Olou78ry5uhoEZ6cGDlDpkoicUYU1xnmViJxhrTQRsvGe/o3 2Vcnu3An7TCmafZynO2oqhoIBL73ve9NTk62GTeJRMVisbmFfY/8n+d5daJaYWh5+fuTk/9oyJ+j 0VcavoIajudbLh8dLWna3v37XpLDj44WWdZx/367B3linagXDj1bLKDNjCoUCuFw+HBxoMNduBOU 5/rdu3cPN+6uazjCUtXHqrpdH3omJr5qc4DWBsnhm3N7hCygW7swzUjUh9cuut0neH44Gh0lh/31 p0XJ1WjtD9BaIkl4c26P0KDr4i78gub/9Sl6y3S9/tqz5y153rYNy1nW3nJbhFAN/SsbEUIIIxFC iD6MRAgh+jASIYTow0jUjxpuHUDoiPp/Rg12JHI6GQCoPTbBMlT18fnzg/3R9CGrzpZO9P+M6uvB 7cvjAbDi3FKUrenpwf5o+pBVZ0sn+n9G9fXgOhEK2ck9pbQH0jWyvKmq22+9Re0uHAuz3mzpxEDM qIGPRLEYY5pmKLRijemlqo9nZyvvvGMPBhnaY7Egi82WTgzKjBr4SOR0wrVrzPLyo4mJ2wsL9wc0 99Y0Q1G2BGFtcvKOx8N89FFfT5rBZY3Z0omBm1FWuNtjaoq5cYORJHN2tkJ7LEcyMmKLRGyxGO1x WJplZksnBmhG0XwqSIfaPBWkWakEut7Nd7S9DdeuVTc24Fe/sr32WhcbbuRyMS5XD9tHzbo+W/a1 vQ3Xr5s/+Qlz6lTP++pwRh3lqSBdZIWcqJ7HAwBdTkQnJ+2SVP3kk2osZuP5fs9yUed6MVvaKBTM Dz+sxmK2t9/GWdTIapGoF5xOmJuzAVSTSQPAjsEIHUI2a6bThtPJ4PxpCSNRp+bmbACAwQgdlK5D IlH905+qACBJA3+OqEcwEh0ABiN0UKUSfPCBsb5uAsCPf2ybmsJp0xpGooPBYIQ6J8umKFb/9S8A gBMnmF/8AhOi58JIdGBzczaPh0kmjVLJRgITQg10HUSx+umnVfLSbmfefx/PjbaDkegweJ4BsCeT BjzNkhCqqVTg5z83yuVnS0ZHIRzGedIORqJDwmCEWsrlzA8/rO7sgGE8u0Lngw9sjf9nKPpPGIkO D4MRaiBJ1Uym2rDwBz/Ay9D2h5HoSEgwkqTq6mo1ncbvvReXrsMvf1n9618bwxAAfPABhqH94Tf5 UfE8k8nYKxVzZsbI5fr91hnUC4WC6fcbf/tb46dvszGXL+OZ+45gJOoCjwcWF+1TU0w8bkhSVddp Dwgdo2zWDIeNR4/M5lu3TpyAn/0Md7GO4J+pO5xOSKdtyaRdls1QyCiVaA8I9Z6uQzxeTaeNsTHG 42FcLsbhYF566UkG5HAw772HZ+47hXWibuJ5xuezx+MGz+9FIrZY7LgDfS5n5vNQLJorKy0KFtZw 6ZLj4sVqMGgjT4OliHz9ANh0HZLJai5XjURsf/6z+dVXAAAsi2fuDwAjUZe5XLC4aCfnUIpFSCZt x/OtqOvw8cfmjRsGx516993THHeSZe3H0fHxKpd3y+VdWX6QyezG4/ZwmH4JplAwk8mqrkM6bQ8G mVu3qgAm4Jn7A8JI1BOxmM3nY5LJaihkzM3ZjuHBnVeuQKkEi4uv8fxwr/uiLpU6KwhrorgOQDkY ka8cn89WO3PKsgB45v7gMHvslakpplbGvnq1t2VsSarevLmXz7tfhDBEpFJnE4kz6TS1klypBKGQ kclU43F7JvMs/RkfB8Az9weHkaiHamXsXK63Zezr1yGROMNxJ3vVQV9Kpc663UO5HIWKWDZrRiIG AMiyozkpC4XwzP2BYSTqOZ5nFhftLhfD83u9OMdfqcCDB1W//3SX2x0EPD9y+/axzuHa+TLysTZX zZ1OBgvVh4B/suPgckEmY4tEbLJszsy0vubo0BGqUjEB4EVLiAi3e+jLL/eOrbtCwZyZMQoFM5u1 P+/EaDiMZ+4PAyPR8YnFbF98Yed5hsSjbPY/LoSTpOpRMiZLninbl9s9dDwd6TpIUjUcNqammC++ sOPBV9dhJDpWTifEYrbFRXswyKTTxsyMIctP4lE4bMtkqn7/syWoTxQKZiRiyLIZj9vx7sIewUhE gcsFc3O2XM7h8zHJ5JN45HLB5cu2R4/MZNK4fNkoFDAeUabrkM2aMzNGOGw4nUwm0xeXL1kVRiJq muPRa689meh37pjhsPHrX+MtbHSUSnD1anVmxkinDZ+PkWVHJkP/km5rwysbKSPxiOdt6XT1k0+M M2eY776DatUEgM8+M//4R+OnP7XhV/Hx0HXI5UxZrhaL5vg4Ew7beJ7BY7HjgZGoL3g8kMnYCgXm o4+qJAwBQLVqbm9DOm384Q+MIOAlKj1UqYAsV2XZ3Nw0QyFbPI5/7eOGkahfyLL5u99VNzZalIfI wdrly7ZE4kjlUk0zRHFDljfL5R23+wTPD6dSZw/fnCXIspnLmUtL1fFxhucZnrfjOXgqMBJRpuuQ zVZ//3vz0aN9StSffWbmcsZ77zGHu8Vf04xA4GtV3SYvy+UdUVwHgGMORgxzCwBM883j7LRZfRIU DNrIzat0h/SCw0hEGXnM4xtvMKurUEuI7Ham/nnsRLVqPn4MmYz56aemKB748EEUN1R1m+NOLS6e d7uHNM0IhVZEcT2RGOvba5HK5d1uXTFUqUClYhaLoOtmsQhLS9XhYUyC+ghGIsp4ngF4FlNKJdB1 c2nJBGD+939hddWshSebjSElpO++M8Nh4+JF5pNPDrAXyfImAMzPnyP7Nsva5+fPlcu75LeCsEaO 2gCA50fILV0kf3G7T5TLOyzriEZHSQLVcrkgrC0s3Ne0PZZ1JBJjicQYORisLYxGR1X1ca07sklz p7X2/f6XOe7k4VK2UglWV81SydR1WF2FpaUn96YNDzMeDzM+DsEg/seZ/QUjUX/xeACAach3GsJT sWju7Ji3b5vB4F4kYvP5OtqjyA5ff1OI2z1E9nzyhI1E4kwqdVYUNwThnqYZ+fwFslo+f0HTjMnJ O6K4Xh8X6pezrF0U15eXX+e4k7Oz3wrCPZ4fXli4L4rr8/OuaHSUdLG8/Prk5EOoe6zH8zoFAEV5 mEiMdfLWCgVzdRUqlSf/FotPYvf4OONyMV4v+Hx2rxc8HjwR1r+Y5ofv9htFUQKBQLFYpD2Q/lIo mJubUCqZ//wnfP55dd/KC0k07t692Hy801C7qb2sX97+Z7//ZUV5WN9mPu8OBMrQVBJq2Xib9ttT lC3SCwB4vYzLxYyPA/kXT351qFAohMNh6nEAc6JBRfa0YJApFMzPP99/fZ4fkeUHodDK/Pw5khmF Qiuy/GB5+XWWdWjanqYZLGvXNAMAWPYwE6Nl7CDNkmct1mdk3eoUAGTZgZcdDjq8xvpFkUiMsaxD VbcnJ+8wzC2GuSXLD0gtJhodBQBBWAMAUdwAALKkc+SZJKQUJQhrodCKphnR6Cu1BsnC+k2O3mkN hiELwJzoRcFxJ/P5C6K4oShbpIRcu56IlGNkeZNhbrGsg9RuDtQ4qU/Pzn4bCn3Dcafm58+xrJ00 QqpFbvcJspAcxwnC2tE7RVaCdaKBVyiY4bBB/QodKkidqFjEL9TD65M6ER6dIYTow0iEEKIPIxFC iD6MRAgh+jASIYTow0iEEKIPIxFCiD6MRGiAkdtEkAVgJBp4TicDALUHbrxQVPXx+fM4h60AP8WB R+66ejEjkaJsTU/jHLYC/BStIBSyk+f70B7IsZLlTVXdfuutfr9dCXUCI5EVxGKMaZrkDnjaYzkm qvp4drbyzjv4/GmLwEhkBU4nXLvGLC8/mpi4vbBw38JHappmKMqWIKxNTt7xeJiPPsIwZBF4E7NF TE0xN24wkmTOzlZoj6XnRkZskYgtFqM9DtQ9GImsw+mEuTlmbs5BnntNezi94nIx+L9xWA9GIgsi j+WnPQqEDgDrRAgh+jASIYTow0iEEKIPIxFCiD6MRAgh+jASIYToG5iz+F6vl/YQEEK9MgCRiOO4 fD5PexQIoR76fxKR4nAnAEc0AAAAAElFTkSuQmCC ------=_NextPart_000_0000_01C80F9B.F234E740 Content-Type: image/png Content-Transfer-Encoding: base64 Content-Location: http://www.cs.wisc.edu/wisa/papers/security03/images/cih-cfg-2.png iVBORw0KGgoAAAANSUhEUgAAAS0AAASGCAIAAADzYm8zAAAACXBIWXMAAAxNAAAMTQHSzq1OAAAA HXRFWHRTb2Z0d2FyZQBHTlUgR2hvc3RzY3JpcHQgNy4wNULvQ9UAACAASURBVHic7d0xj+PImf/x 4v/GwR2wiTexAUnhWsAFPsBJd78COfAEBshsAafO1NuYcKO74IBFw3oPm5GAgzEWZrThTCcGvJng sCVgL9nIwMLAGuA/eLbLNSQlUVKRT7H4/QQNik2VShR/qmJRJJOqqgxilGWZdhWmJc/zi5/7ymM9 EJprtgyc5cpvvf/nqx4ALkYOAX3kENBHDgF95BDQRw4BfeQQ0EcOAX3kENBHDgF95BDQRw4RlaIo LviXOnKIqBz6vXVRFCGfgEIOMQlpmmpX4RhyiDNIq1IURZIkSZLIQ5mQBeRhkiTS+MhDd7pZoCxf e7r70L7ikY6l+0K1enpdAX0hhzhDmqYSBnv6eJ7nVVXZpKVpWlVVVVWSBDkBUtoi+Zdbmi0qz3MJ TFEU8nR5mOe5bcfc6Rq3nFo9R3MGZoVI2Uj4ZbcZd+MxL8lsLin5aa1PLVcyM89zCY9bzvH30qxJ szKtczy6cm3THsKnZtdRmqaiKFpbM2lOhXkZZaktaRu3iJFDeOMGxg1emqbSZW0u39yjqy0mc9wl W193LPuBB13dICNQffRL7e5W5ewfut1IN3tdKuPu/tnCZcIt2S5mX7e1HPvXrWez5n24cm0nFddr i5QdJolPgG/tyirRL8XIFEURWgivRw4xMoEfkb8MOQT0kUNAHzkE9JFDQB85BPSRQ0AfOQT0cd+1 mIV8BvohT09PNzc32rUYGr9rQ1gC/M3aAOiXAvrIIaCPHAL6yCGgjxwC+sghoI8cAvrIIaCPHAL6 yCGgjxwC+sghoI8cAvrIIaCPHAL6yCGgjxwC+sghoI8cAvrIIaCPHAL6yCGgjxwiFHd3d4vF4quv vlosFovFoigK7RoNh+uXIhSLxWK329mHk9oyaQ8RiglextsihwiFe8Pt29tbxZoMjxwiFGmazudz mb6/v9etzMDYP0RAkiSRialtlrSHCIi0h7ZVnA5yiIDIUM3j46N2RYZGDhEQGapxB2wmgvuQ4gO/ /e1v//nPfypW4Kc//enr168VK/Dq1as//vGPQ7/owK+HwP3zn//84osvtGuh6c2bN8O/KP1SQB85 BPSRQ0AfOQT0kUNAHzkE9JFDQB85BPSRQ0AfOQT0kUOMWFmWJ+eMAjmEN8vlcsiXK8tyvV4fnzMW 5BDebLfbIV9utVqdnDMW5BDe2D6htEtlWS6Xy+VyKQ9lovnfWgn2XyNt2S5DDuGNTc5qtZLI2RZy s9lst1tZwP3varVy8+b+a6R7epfh/EP0xfYSW7uLMnOz2TT3KsfbvbwY7SGgjxxCU1mWE2z9msgh /JNdO9vhlBGa2v6eHcLZbDa1J7rjPd1fpXXOWHAdYXzg9evXF1+fZrlcdjx00X3J4b158+bt27cD vyjjNPCgv+5ls3HbbDbxdWXJITyQ435uD/OIsw4MBtts+kUO4UHHBLoLM0LjYpwGOgihixwC+sgh oI8cAvrIIaCPHAL6OG6Bukmd+Nf0k5/8ZPgX5XdtCEuWZXmea9diaPRLAX3kENBHDgF95BDQRw4B feQQ0EcOAX3kENBHDgF95BDQRw4BfeQQ0EcOAX3kENBHDgF95BDQRw4BfeQQ0EcOAX3kENBHDgF9 5BDQRw4BfeQQobi7u1ssFl9//fVisVgsFkVRaNdoOFxHGKFYLBa73U6m5/P58/Ozbn2GRHuIUNzc 3GhXQQ05RCjcy+nPZjPFmgyPfikCkiSJTExts6Q9REDm87l2FXSQQwTn9vZWuwpDI4cIiAzV3N/f a1dkaOwfIixJMsVtcorveSyyLNOugoJvvvnml7/8pXYthsZ9uYM2wTvjTlCWZewfAvrIIaCPHAL6 yCGgjxwC+sghoI8cAvrIIaCPHAL6yCGgjxyiLwFe6OlklbTqTA7Ri6IoLvudepIkSZL08Rv3WpWS F3Zms85Zlsky3itTQw7RizRNL35uVVXX/8C92bK5VSqKIk3TqqrkfCOJX7POeZ4PdEJShVDZDaUP eZ6naWo3d3lotzx5dfmXVEMeutPNAmtblHk5X6RWgsy0L1RTK7lWrK2V+7C1zNZt286s/ded79b5 SFEepWlKezhRaZpKi1G9fN/LRiwtQ5ZldkOU3prdOs3LNuqWZovK87zWi6teGhwbdePE/rhmsUVR SK3k4QVldiHv2ktRZ7wkwtRre1gdaBzMS3KaS8qG3lqxWgAOFSsTx9+X+6zWYvM8ly8F9ym2zGYU W1u21jd4fJ30h/YQB7XuXxVFIXtWzeVrfcJDbDvcUa3Y1h05t0y7vHFye301etdr0HENxfYwTVP7 6rYZrA7sGVYv27qdrhVrS5CmzC38SK2axbqvbj5sHptlttbTfBhO+6aadW6uk/6kaUoOw9X3OI39 IpaJWn+vNk5zslbuflprCbZwd6yl2RjUHrrFuqNKtQrXxm+a5VSHh3zcbxzjjJE2w9mTNE25TlS4 7OhI3GpvM8DrtfVdJa5PA2VFUTS/a3o6jn8BOY4/wAtxvTZoag75BNUYDtYfoT0E9JFDQB85BPSR Q0AfOQT0kUNAHzkE9HH8MGiBHM4e0n6/n81m2rUYWnC/IcLETeTXfDX0SwF95BDQRw4BfeQQ0EcO AX3kENBHDgF95BDQRw4BfeQQ0EcOAX3kENBHDgF95BDQRw4BfeQQ0EcOAX3kENBHDgF95BDQRw4B feQQ0EcOAX3kEKFYLBZJkhRFkSSJTGjXaDjkEKGoXca7eavgiJFDhMLN4Xw+V6zJ8MghQuFeTv/m 5kaxJsPj/hYISJIkxpj5fP78/Kxdl0HRHiIgU+uOWuQQAZFdxAned40cIiCSwHfv3mlXZGjchxQf KIpC/cCd7t1X0zQd/pAJ4zT4wOvXr7/44gvtWmh68+bN27dvB35R+qWAPnII6COHgD5yCOgjh4A+ cgjoI4eAPnII6COHgD5yCOgjh4A+cgjoI4fwZrlcDvlaYr1eH5ozIuQQ3my322FeqCzL1Wq13W7l FdfrdVmW8nCz2ZRlOUw1PCKH8MYGoCxLyYY0UPJQJpr/rZVg/3WkWVuv15vNRqYleKvVSh7aiXEh h/DGJme1WknkbAu52Wy2260s4P53tVq5eXP/dVmz5mZyRDgfH3053kbJzM1m09yrvCZIZVnapnJE aA8xPrbZrBlpY2jIIXRdlhy3N2v3FSWZUtrohmrIIfyTGNgOp4zQ1LJhh3DcbqQdyHEfNq1Wq9Vq ZYd5pHlcr9cyGmQHhEaE60ThA9dcJ2q5XHY8dNF9yeGpXCeKcRp40N+OWXMUZ7PZjHQn8AhyCA+k W9hxoPKs37sE22z6RQ7hwVmHCuywSnzN2sUYp4EOQugih4A+cgjoI4eAPnII6COHgD6OW+ADaZq+ efNGsQLffvvtz3/+c8UKDH/zQ8Pv2hCaLMvyPNeuxdDolwL6yCGgjxwC+sghoI8cAvrIIaCPHAL6 yCGgjxwC+sghoI8cAvrIIaCPHAL6yCGgjxwC+sghoI8cAvrIIaCPHAL6yCGgjxwC+sghoI/rJiIU RVEURfH09HRzc2OMmdTVE8khQrFYLHa7nX2Y57nKJX1V0C9FKGazmftwOiE05BDhuL+/t9Pz+Vyx JsOjX4qAJEkiE5PqlBpyiKDYHE5ts6RfioBId3RqnVJDDhEUGaqR4xaTQr8UASmKIsuyCW6T3Ic0 KnIoXLsWV/n444+zLNOuxVXSND13kIkcRqUoikn9DCVMWZadm0P2DwF95BDQRw4BfeQQ0EcOAX3k ENBHDgF95BDQRw4BfeQQ0EcOAX3kEAEZ+4/UL0YOERB7poU9Mb9VkiRJklx8WkaWZVLCZU/vA+db IEQnT0G85hxFOSUlqBzSHuI0OT23KAppRuShTMgCtoWRNkoeutO10mrNkbu8ndO9bq2lyRxbsiwW bL+X9hCnpWkq59RVVSXbt21SqqqSf9k5WZbleZ4kiZyDZ/8lJAlVVUkwZMK8tG82Th0D0yxNZtrS pHqycNDXgKsQEYlKH+ym4m4z5iU/zSVlo29WqZaE1gKbZR6qTLM0kee5hN8u31wz/W38F3wK9Evh QbP5StNULtLRbILyPLfb3/Uv3SxNGkD3daUy179Wj/x8AyAMKu2hXI5F5thmsHpprGrluM2URMhd rPVVjlSmS2nSNrqV7PIS17jgUyCHUekph3YHr3ppc2odP9v4NHuhrZW0hbhzavuT9r/NBsN9WCvN VlUmPvvsM/tlUSsnqBxy3cSoyBiJdi38q70vOyRzDS+FtLrgU2D/EKFrvQjd9cfxr66XTxy3QOia Iz1XtmMBdhloDwF95BDQRw4BfeQQ0EcOAX3kENBHDgF9/J4mKhHc//Dp6Wns9wO+4PgkOURYYv1p 3nH0SwF95BDQRw4BfeQQ0EcOAX3kENBHDgF95BDQRw4BfeQQ0EcOAX3kENBHDgF95BDQRw4BfeQQ 0EcOAX3kENBHDgF95BDQRw4BfeQQ0EcOEYq7u7vFYvHVV18tFovFYjH2C7GeheuXIhSLxWK329mH k9oyaQ8RCvcy3vP5XLEmwyOHCIV7/+3ZbKZYk+GRQ4QiTVPbDN7f3+tWZmDsHyIgSZLIxNQ2S9pD BETaw6ntHBpyiKDIUM3j46N2RYZGDhEQGapxB2wm4pV2BeBTlmXaVbjWxx9/PPZ3kabpuV8l5DA2 E7yJZ2iyLDs3h/RLAX3kENBHDgF95BDQRw4BfeQQ0EcOAX3kENBHDgF95BDQRw4BfeQQndgzdNEH cgjoI4e4SpZlSZIkSWJPVqrNKYoiy7KiKGSmfWKSJO4VSlsXaxZ+qDT739p8O0eeKEXJzLCuj1oh Imma9lRy66aSpmme53YBOe+uNkcmZKYscOQl3MVaizpSWp7nMl/O/JIJ91n2KXbJnlzwKZDDqAyc w9rM5je7ebni0/Fymv86VNSR0mqn/B2pc39rydbk3KfQL8VVmr07j/29c4tyW7lDy6RpGlaP1BjD /iEuUxRFURTuNi0Pm3Nqz+pyonqXopoz0zR1dyNljt1XlDm25OAuveG1QYaynnpcrdfasK8oD+1L N+fYp7jVMx82X62LNYs6VFpteVuyO8fuN9qZvtZPsxrnPoXrCEcly7IAr0+TJJ02M7+LKbrgU6Bf Cugjh+hXxz0xv4uNDjlEv2RX7eQQpd/FRoccYggdr+fpd7ERIYeAPnII6COHgD5yCOgjh4A+7vcU m1iPsLn2+/1sNtOuhU+h/0QIoyY/B/f+U7vFYjGbzd69e+e3WEX0S9GXnkJojHl8fHz//v3d3Z33 krXQHqIX/YVQLBaL3W53e3sbR6tIewj/+g6hFU2rSA7h2TAhvLm5kYk4oki/FD4N1hKaDy+pOvYO Ku0hfOp45Qsv5vP5MC80AHIIb7Isk6u/DPNycghxPp9/8skno24MDTmELwOH0Bhzf38/n88fHx9/ +ctfjv3XC+wfwoPhQxhaBa5Ee4hrhZCBPM9HfZI+OcRVQgihCPGqpJ2RQ1wunBCal4tljLRVJIe4 UFAhFGFeM78LcohLBBhC89IkjrF3Sg5xtjBDKAK8nHkX5BDnCTmEYowDNuQQZwg/hGacAzbkEF2N IoRidIcTySE6GVEIrRH1TvldG06TDXp0QyDh36HNoj3ECdLBG10IjTF5no+lSSSHOGbI83q9G1Ev mhzioFGHUIzlGAY5RLsIQmjG0ySSQ7SII4RiFE0iOURdTCE0I2kSySE+EFkIRfhN4mgOsGAYWZZF FkIR+LFE2kOYh4eHh4cH8/KjGe3q9CLwJjHoLwkMQy7I+8knn/zP//xPrDk0YTf1tIdTZy9K/7e/ /e3p6Um3MpNFDqduv9/b6ffv3yvWZADBdk3J4bhlWZZlWZIktYmOT7+7u9vtdjI99ltEnBRsp9Rw X+4xkpjZO0nUNi858GCjeORkpaIopAGUq2JHvGcYPsZpxkHSZR92PBWweCEL154lt/Kcz+fPz899 1DlAwZ7ARQ7DZbMnQbryNNxaKyrj+NH3RZvCHDUlh8E53u309RJff/31d999199LBCvMA/oh1mmC Bsheq8u6u6NGe4gPhJaB4zuT0QjzQjvkcFB+d/n6o9U+DyDMoRpyOIRRb9Zu5YP94uguzBNKyGFf at280D74C4TWkb5YgLuIHMf3qbmlhvZ5X8MNnv21QDTfMrpoDz2IrOd2rtH1ugM8dBFchcZidBvf MEbxlUQOxy2+Xb7+hLwzGeD+ITk8IeTtaSxC+/6iPRyNUfSvxiiE/nyA7SHnH/6LewqfMSbP86qq 8jyPO4QD358sz3NZsenLPbRlhY/rNmneTb09VOwyhdA7kmMPthr2xMWBj7gM3PkPsD3U3xSGxy6f y/06sNOK3xHDnG4SWg5NNRk2b/Jlr12dyq2D3Syac2zH2LZR0rU79CHK8rJAmqYnq+E+XVaRlODj LV7L/cg8VimQd+eayu9pbOcznN+42A6hNM5VVRVFIQ2R1LZymiZ7nQsbS/lv84qjcoKvbGpyQcTu 8jxPksS+tLrR3V77cnpfATp6+oq9gF35tSBVH7ZR7vK1Ch9quOzTu3y+5sNG2K4f3S5Drx8T7aE+ 2xjaX0jKQ929xC6jsmma1hoH7+cN2CZaho6rYVvFEA5pqNH+IgiCbQdsx3WAF7UrXzY4O129tJDu HNkndBuHI1/q5tL28NB0f9y938HWfIDtITmsG6zjaj4cIJGvRbshunNsUN2Z7pdpc8hH5phT3cva +JB92OsbV/nWcwWYwyket+iu155SCMcPhxROtzPA4xaT2z88S087k3ZbHEZz1HSwHwmFk73AkcNO 3ODVToG9IJPy9ME2yoFb3bhPhu7JtLpG3vF9L8Zy/StBvzQ27scpY/2TyiRfQ76QQ2/CPDLpXWgn E8aBHPp3ZGdypFstu3x9Y/9wOKPrxcV6MjT7h5N2aGcyqE18dF8WcSCHOmo7k7bXp5JJdvnU0S8N yJBt0ZRPhqZfimMGOApCtzNM5DBQtY6r3EP7sp1Jshc++qUjYG/Z170zKUvu93tjzGw2m1S38yT6 pTibBK/5W1bptbpHJpspNT2cK4w+kMPQHTo5o7kz2ZxvXjqlCBw5DJq9mujxxY60eNKbpVMaOK7n HS7pZF7Zq0wbV7VBgMhhuLy0Y7SEo0AOA2XHSH2V5qUc9IQchsgdI70e46XhI4ch4mDD1JDD4DQv le+rWO9lwhdyGBa/PVKL1jVw5DAsvfZIOYARLHIYkJ56pIIDiSEjh6HoqUdqkcOQkcNQDDBGShSD RQ6D0GuP1CKHwSKH+vrukVr8xi1Y5FDfwKelciAxQORQ2cAh5EBimMihpsF6pAgcOVQjl7FQaaDo moaG8/HVaJ0mn+d5886k0EV7qKPjBS96wgGM0JBDBV4ueHENchgacqjgrN1CG5jkQ+aK3213yaEs kGVZ7XXtVYkve2m0IodDO+unM7WFK4fM6SmK7mVR7cvJhHtrx8teGi0qDCjP8zRNuy/vfkB2Os9z G4+zSutek+aG0Zxz8UurC7DmXFd/UOcetU+Sf31A7iCnpKi2wLmOPLf5ry5zxoLr6k/a9T/m9rvd c4nhcLB/OBDdAxWtGDUNBzkcwsUHKty7rxljkiSpJeeaYB/JYa2q9tV9vTRqyOEQrvn9mv0NquzQ u1v/9T+LOxLF2q2j5NU9vjRc5LB31+wWHnmily6l3K3trNf19dJwMU7Tr+vPqDj03L67hUfKp0fq He1hjxTPqDgL7Zs6ctijURwVYNQ0BOSwL35v2NQfchgCctiLcZ1oTxTVkUP/xrJbaJFDdeTQv1Hs FrrIoTpy6NlYdgtriKIucujTuHYLXeRQFzn0YLFYyMS4dgtrnp6eFovF3d2ddkWmiN/TXCvLst1u lyTJ7e3t/f29dnUusVgsdrudTNsJDIkcXuvp6Ukm3r9/b8bZKYU6+qU+7fd77Spc4vn5eT6fa9di 0sjhVYqisB25+Xz+/PysW5+L1U6nUqzJNJHDqzw8PMjEqENojHl8fLRNIjkcHjm8ijSGYw+hoHeq iBxeTtqNOEIopHdqR54wGHJ4uT/84Q8xhdB82DvFkOI8bvH69esffvih71f59ttv//M///PXv/71 ySX//Oc/e3nFAe6XdnNz880334RzY7bx/i7iLHHm0Biz2Wy0q/Cj9XrtsbSJbJcinK+DvtEvBfSR Q0AfOQT0kUNAHzkE9JFDQB85BPSRQ0AfOQT0kUNAHznEVThZ0QtyiMsVRTGdn4D2ihziclwUyxdy aMqyXK/XZVkul0t7bsR6vV4ul3aOu8xyuVSt74Wk7SqKIkkS9073WZbJHGnZDi1mC2nOl5nydCnN lkCvtSNyaFarVVmWxpjtdisT6/V6tVptt1uZIw/tMqvVyu+pTMOwV+yuqipNUxubNE2rqqqqShLY upiw8/M8r0WxqiopLc9z20i60zgu2vMPz7Varex0WZb29MXtdmsbQFlms9mMtEk0Lz1JGyT3AuRV Vdl01RYTch+rk2W6E72+l5jQHraT1m8Kzuo65nlevTiyGHfLOBc5bGF7ocaYsizdprJ1zujYO8O5 gWneLq42x+2mNmNmF5aJWp8Wx5HDH5s+N3jSKZUhGbeP2pwzOjLKYruj8rc2s3WOMUbSVRunsXNk YbuTaV7GbwZ9e6PF/qGRIRmZthOtSbP/Ha9mf7J1L+5Qt7O5cG2OjPocKhaH0B4C+shhJ2M8UFHT cW+NnToV5LCTzWZjjy6OlAx1nhzG7LgY/CKHZxj7MKnp/Es0jr8PjBwC+sghoI8cAvrIIaCPHAL6 ov09zblH/P7v//7PGPOzn/2sn+p4o3V87+np6ebmRuWlpyA5/sP56ZBfRTJef0iSxLOpyHmS2rX4 AP3SHzXPNgAGQw6NIYQdsH56RQ6NIYfQRg5/RA5P4ken/SGHbF7QRw5N7axzHMIXVn/IIaBv6jnk tNeO2H/u1dRzaLiSCgIw6Ryyw9Md7WGvpp5DNi+EgBySQ+ibbg4JIcIx6RwyQoNATDeHQDgmmkMO G16A4eX+TDSHhsOG52N3uj/TzSHORXvYnynmkE4pQjPFHBo6pRehX9qfieYQCMrkckin9BqsvZ5M LoeGTinCE+11hJuSJLm9vX3//r12RUbJ3uk+pguZhmNa7eH79+8//vjju7s77YqMz3w+165CzKaS Q7tj89133+33ew6FncteVJ9A9mEqOdzv93aa6+dfwO5U73Y73ZpEaXI5vL29fXx81K3MSNES9mcq OZRv8fl8/u7dO+26jNVsNjOksR9TyaExZj6f0xJe4/7+XrsK0ZpEDmWQht3CK7H2+jOJHO73e3YL vZjP59yNtA+nj8nyU6bBnPVDn2l+Ll5+CxXgfUg7/Z4mtEpH6YJcTe1zifirZxL9UiBw5BDQRw4B feQQ0EcOAX3kENBHDgF95BDQRw4BfeQQ0EcOAX1TyaG93BgQoKnkkEv9IWRqOSyKIsuyoiiSJEmS RB7KhCwgD5MkkV/Zy0N3ulaaLGznNEs7Xhn36W6t3JeW+XFc6622/mVmbZ0fWswtpHW1y5wo11tf qlPSND25zGWMMXmeuxMyLS/qzpE62MXsv0Se53a+fUe1t3bkndaenue5fcv2WXIuv12yD+eu5+s/ F7s+0xfNdd5czD79+Gq3n6PH9eZrU+xvk76Ycg5rE3a6NUU2IbUq1a7XYJ/iLnYkh83LPRyqba+f n0oO3elD31zNT8dWoHWlSeoOfRDXCK0cj8LdP2x2Y9I0LYqiKIpmctyvW5kjEx3PHG0+vUle/Yw3 MELnvsHmerOXArLLTGG9eXAyqSrtodsFcjuK8gHXynG/fWv9n9aXOP509yVsafIdX+uY+aXYHtq3 1rrOa4vZpzdXu7vqzEsXxuN6i7g9VMuhvaZD9fJVWuvP2O/UZi+0tZK2EHeOLGx3/Kq2Dljz6bWH tlZ2vtc1cex9eVy+yX4X177mTKNL3/pBVAdWlHm5YMdnn33md72Rw6iE+Y509w+vX2wAEecw3P3D nhRFMbXLKyF8k8shF8MVHUewIr5EWlAml0MI2aM7OZLZcTFciRxOWsfeAZ2IvpFDQB85BPSRQ0Af OQT0kUNAX6f7PY3uINJ+v5ebSMdt+M9lv9+blxt0Dy/iYdvTORzjr08CvMGddypvUJIf/bodHv1S nIcQ9oEcAvrIIc7AD9x6Qg4BfeQQ0EcOcYaIjxzoIoeAPnII6COHgD5yCOgjh+iKg4f9IYfoisHS /pBDQB85BPSRQ0AfOQT0kUNAHzkE9JFDQB85BPSRQ0AfOQT0kUNAX6frCI/IYrHY7XbGmCRJjDG3 t7fv3r3TrlQMZH3KRFVVupWJT2zt4c3NjfuQEHo3n8+1qxCh2HLIVW77NoUbFgwvthy6bm9vtasQ D9sM3t/f69YkShHmkC2mV5yF2IcIc2ixxXgk3VF2DnsSYQ5lqIYtxi/JYW0YDL5EmEMZqmGL6QPD YD0J6/hhURRffvmll6L+8Y9/vH79+spC0jT99NNPvdTnGl9++WUI12j693//9+tX6TU+/fTTWPc1 gsvhF198cX05Xgoxxrx58yaEHPpaLWP3+eefx5rDCPulwOiQQ0AfOQT0kUNAHzkE9JFDQB85BPSR Q0AfOQT0kUNAHzkE9JFDHFOWpXYVJoEc4qCyLNfrtXYtJoEc4qDVaqVdhakYcQ7l27osy+Vyab+2 1+v1crl057iLLZdLvfpqaq4EOyFrzC4pD93uaG0N1+Y3n+sWa1+U/u1xI87harWST3e73crEer1e rVbb7VbmyKbjLrZarabZ0WquhO12K//abDbukmVZyjLuTLuG3cVk/mazsVGU52632+VyudlsbCHu NFqFdR7wBdwPuCxLu1XJ1lBbzN1oJqjjSliv1244WyNUlmWziavF1b5QLepoGnF72Ir+z5WknezS a9hsNtsXMsd2QOwyth3GcVHl0P3Uy7JsfpG3zpya0CP8fAAAGGVJREFUkyths9mcDI/bw7c7BbWS 5eFk9wXOMuIcysfvBk/6PzJU4PZRD82cmtpKWK1WdgCmNtAlcWquYVuUBMwdp5Gn2CGZ//3f/5Xd dfMyojPUuxylsO7dk2XZf//3f3svdrlc2r7TWd68efP27Vvv9TnX69evr79O1MUrIRyff/65lws3 ZlkW2gUgR9weAtGIP4fsnBhWQvDiz6EM60181I6VELj4cygYJjWshIBNJYdAyMghoI8cAvrIIaCP HAL6wjrfIk3TN2/eXF/On/70p9/85jde6nN9IdfztVqu8Ze//OVXv/qVbh1CuAdeT8L6XZsvAf5w aexiWqUBvhf6pYC+aHMYwo2sgY7IIaAvzhwGMr4CdBRnDmkMMS5x5jC00TDguDhzCIwLOQT0kUNA HzkE9JFDQB85BPSRQ0AfOQT0kUNAHzkE9JFDQB85BPSRQ0AfOQT0kUNAHzkE9EWYw4eHh8Vi8dVX Xy0Wi8Viwbn5V7q7u2N99i2s6wh7URTFbrczxnz//ffadYnBfr9nffYtwvbw5ubGfcg1o640m83c h6zPPkSYQ3dDmc/nijWJw/39vZ1mffYkzhzazeXx8VG3MhFwv9dYnz2JMIfGGNmfMXSivJrP56zP nsSZQ2kP6UT5wprsW5w5lKGa2oANLsb67FucOZTuE1cT9oX12bfgjh9++eWXXo4U//SnP339+vX1 5bx9+/b6Qjzy8qYu4Gt9nuvVq1d//OMfh3/dgQWXw6IovvjiC+1a/Gi9XmtXoe6HH37YbDbatRiO +o2QhxFnvxQYF3II6COHgD5yCOgjh4A+cgjoI4eAPnII6COHgD5yCOgjh4A+cgjoI4eAPnII6Bt3 DsuyXK/XZVkul0t7jtJ6vV4ul+4cd7HlcqlXX32tq6K2xk6urkNPdx/aEsqyHOSdjdu4c7hareRj 3m63MrFer1er1Xa7lTmyYbmLrVarAM8qHExzVTTX2PHVZf+12Wwke2VZytPl4WazWa1WsrA7jSOC Ow/4Au4nXZalPU3WbhnuYnbrmbLaqmhdY4dWV1mWtSbOfglazZJx3Ljbw1Z0hM517hrbbDbbF+bl qgW1ds82qugithy6H39Zls1OUevMaZJVcXyNNee4PVW7J9n6rInvApxl3DmUDcjdjKQjJMMGbh/1 0MxpclfFoTV2aHVJwOzAjCTQjsdIr1V2Ms3LiM7A726Mxr1/KAMMMm0nDsXMLoDaqmhdY0dWV235 5kcgE3zfdTfu9hCIwyRyyF6K1WVVsLqGN4kcyvgew3em26pgdQ1vEjkUDJNaXVYFq2tIE8ohECxy COgjh4A+cgjoI4eAvuB+T/Ppp59+/vnn15fz9PR0/f1rP/roo+tr4tdHH33kZf2cy8v6vMCnn346 /IsOL6mqSrsOvciyjPvXehTT+gzwvdAvBfSRQ0AfOQT0kUNAHzkE9JFDQB85BPSRQ0AfOQT0kUNA HzkE9JFDQB85BPSRQ0AfOQT0kUNAHzkE9JFDQB85BPSRQ0AfOQT0kUNAX4Q5fHh4WCwWX3/99WKx WCwWRVFo12jc7u7uWJ99i/D6pYvFYrfb2YfxvcGBxbc+uX7pENzLTs/nc8WaxEHlMt5TE2EO3a+6 2WymWJM4pGlqp29vbxVrErEIc+i6v7/XrsLopWlquxWsz57EmUO73bjf5bge67MnceZQsHPoF+uz P3HmUIYWHh8ftSsSCdZn3yI8biGSJNq3piKm9RngcYtXWZZp16EXH3/8caxv7axtyNdK8LU+QwtA IF4ZVs2oXBCGcD7fWL8Zrxfn/iEwLuQQ0EcOAX3kENBHDgF95BDQRw4BfeQQ0EcOAX3kENBHDn/U vPxRURQnr4lUW4BrKOEy08ph8qL5Q8fanCzLiqKwp722PrEoiiMPge4mlEPJVVVVcv7OkczIv+zP ow89sXZyesTnqvP90ju7hfmV53mapnZTlocyIQvYrVYqIA/d6VppsrCdY5/uPrSvaF/FVStTHton uv+1c6Q+rU88VKZUoKe1KhXudXn7LLsaa6va/UwvKPaCyngXSDVcr/qLd5Zl8oaTJLEbqJxOKv+y c+S8zCRJanESstNVVVVRFPZs1KIoZELm5Hluv7Nt5k+yJUs57r/cup373u1bHi+7eo0x8tHY1V5b 85q1jEl/3w3m5RvUNBoQ09bCSH6qxtdVLVR2fp7nkha3nCNvp/mizYq1TrfW9lCZrfM9GqY9rK1J X+8rkIYokGq41PYPm0OLaZrK122zNXO7QDJHmqnmHtqREcvj/z3i4ieOV3VqFxqeqbSHaZra17XN YNW2Z1i97JDY6dpi7r+kqEPvyC1Hur6t5VQvu3nukrUnNt/jkbfs1zDtoWi+HXeNsX/oUY/jNJLz ytnNqG3QrV8ErfWxC8tn7w4V1Ep2BxXsRLOcWvLdURm3/NpitdrWRo/cISJz0TBGF0OO07ijxM0P jhx61GN7GIL43t2Q7aF3gVQmkGq4Yj5+WBRFOJdIAo6IOYcRH1hHZGLOITAW5BDQRw4BfeQQ0EcO AX3kENBHDgF9r8wEfs673+9ns5l2LdR4+Xwnvg779moKvzipnV8/Kb7edYD37ozJJPql/LAGgSOH gL5J5NBwRUOEbSo5nOA59RiRqeTQ0CQiYFPJIWN9CNlUcmhoDxGwCeWQXUQEixwC+qaVQ+0qAO0m lEMgWJPLYfQ/ascYTSuHHL1AmKaVQ8PRCwRpcjlk1BQBIoeAPnII6JtcDg1RRHimmEPDaA0CM8Uc cvQCoZliDoHQTDeH/LAG4ZhoDumaIiivtCugIMuy/X7//v37JEmMMdXL/d/RarFY7HY7Y4ysrjRN +Rbzboo5dAdL5/O5Yk1GZz6fE8I+TLFfent7a6flmx5H3NzcaFchflPM4bt372gGu3MbQG5x0ZMp 5tAY8/j4aKPIwGl37969065CnCaawzRN+Wrvju5D3yaaQ+P0Tvf7vXZdQiffWe5+Nfyabg7NS++U HJ4kOby/v9euSLQmnUN6px3JUA0XvOtPEtRR7CzL/v73v2vXQtNPfvKTt2/fXlPC6Iadhj8gGeA9 VYM7jr/ZbLSroOnNmzfXFxLaRnbE6L41ejLpfikQCHII6COHgD5yCOgjh4A+cgjoI4eAPnII6COH gD5yCOgjh5cry1K7Cmq4ILpfwf2+9BrL5XK73Q7zWmVZrtdr+3LL5VImVqtV9D+RLYoiy7LaGQJF UdhwjugHroGIKoeDhdAYs1qt7HRZls1ARqz1BKhmMtFdVP1S6ShKS1WW5XK5XC6X8lAmmv9tlmD/ u16vO76uzWRZluNqDKVlK4oiSRK5PKkxxk5nWWZnyrTbHZVnyQkT8tctBGeJKoeSnNVqJZGzbdRm s9lut83/rlarWtjc/567+2cz7OOtDMTegq6qqjRNJU62WXO7l0VRyDLu06uqkqfLklVV0SReJqp+ qcu2UW4HsvbfzWbT2o1sfUqXV9xut2Psl0q68jw/3prVTp/l9HyPomoP0R9p6DhttyfRtoddlGV5 WdMXpaIojjdxJxtMXCzC9lD20Gz/UEZoarttdginNqxix3LchydfxY76LJfLcY3TCBlfKYrCdjvT NK2Nwcg4jQRV9gntmI09YmGfgnNF2B7KfppMu0cyDk13ee7xJd3pMWoOrjQPANb2DO1TmhO4QCQ5 7K+H2Rx32Ww29GbhVzw5XK/XXfqE3Y8KilE3dCfRjQxEJPuHcoSwSzMlS47rKF9/8jy3xwChKJIc nouOpYsjgeommkMgKOQQ0EcOAX3kENBHDgF9YR0/TNPUyw2PevXtt98aY37+85/3UfirVx4+EY4K jk5Y9z8ciyzL0jRluH+kArz/If3SS9jTZwEvyOElpCWk+wdfyOGFpGNDqwgvyOHl6J3CF3J4OXqn 8IUcXoXeKbwgh9fK85wc4krk0AN75U/gMuTQA/fqScAFyKEf9E5xDXLoDb1TXIwcesNhDFyMHPrE YQxcJqzzniIgF5/nLBachfbQvzzP6Z3iLOTQPw5j4FzksBdyGIMooiNy2BeOKKI7ctgjjiiiI3LY I7mGDVHESeSwX4zZoAty2DvGbHASORyCO2ZDINHE72kGIjuK8/m8KIqHh4fn52ftGiEg5HAgaZr+ /ve//+6774wx8/lcuzoIC/3SgSwWCwmhMWa3293d3enWB0EhhwN5fHx0H+73e62aIEDkcCBpmlZV ZXuku92OARtY5HBQz8/PNooPDw+6lUE4yOHQnp+fb29vtWuBsJBDBe/evbu9vWW0BhbHLTrx/oOY 2Wz2ySeffPPNN8H++pQbPA6JHHZSFEVod67sm9xrVbsWU0G/FNBHDgF95BDQRw4BfeQQ0EcOAX3k ENBHDgF95BDQRw4BfeQQ0EcOR4/ziSNADkcv2DM20B05BPSRwyEURZFlWVEUSZIkSSIzsyyTh9Kg tS5TK6T5dLcxtHNkSfqrI0IOh5CmqaSiqiq5oLCc3VdVVVVVksDmMm4J9l9y32/3oV0mz3N7xqA7 jRGo0IHNzMXcVd1c8/KwtkytAoc+uGZR19e28vGWgxXgW6M9VHNuvzHPc/uxHVnMtqsYEXI4tKIo 5NIv7p1nas1dc47bU5X/HtqH5KaLY0QOhyOjLHKpG9mvc+c0l3GfK+my4zSy+2fHadyRHnvHxdag IkxcJ2o4tf5k64WnjvQ5a8s3ny67PYdKRshoDwF95HAIXfbW2KObMnI4BBnqPD6M2WUZxIocDqfL gXUOvk8TOQT0kUNAHzkE9JFDQB85BPTxe5pOAvzF5tPT02w2m81mPZXPyO2QyGEnYd6UM8syfsIW B/ql48Zx/ziQwxHL85wcxoEcjl5oO664APuH4yaXq9GuBa5Fezh6AY7l4ly0h6NHkxgB2sMY0CSO He1hDGgSx472MBI0iaNGexgJmsRRoz2MB03ieNEexoMmcbxoD6NCkzhStIdRoUkcKdrD2NAkjhHt YWxoEseI9jBCeZ7TJI4LOYyQveWTdkXQFTmME3cjHRdyGCeaxHFhnCZaMmBz/CbeCATtYcwYsBkL chgz2zuVW3ZrVwcH0S+NXJqmv/vd777//vv5fK5dFxxEexizLMuyLPv++++NMbvdjmGbYJHDmOV5 7jaDDw8PipXBEeQwcs/PzzaKu91OtzI4hBzG7/n52U7TNQ0TOZwE20H9wx/+oF0XtGC8dBLkAMbD w8N+v9euC1rQHk6F3DqOXcQw0R6GqL9j7p988snd3V1/dy+9QJj3lhwYOQzUdG4wmmUZOaRfCugj h4A+cgjoI4eAPnII6COHgD5yCOgjh4A+cgjoI4eAPnII6COHgD5yCOgjh7GRS5UWRZEkib0BW5Zl 8tCeUdW6mFtIbb6dI0+UomQm19q4HjmMjb3DTFVVck9SObGoqqqqquwFhZuL2RLsfHsrRbkSsZSQ ZVme5/ZMJXcal6sQHhuby7gfa/NTtg9ri7mvXttCWrcTY8yV9bQvd30hgb/iSbSHk3Bu1zHPc7uJ HFqGW7t5RA5jVhSFXHXCBkbmtC5mH7rdVHlimqZ2X1Hm2JK5bYYX5DBOdkwlz3O5xIY7p3UxO1MC 5o7TyE6gnSM7mfYmNs1hHpyL69PEqdafPHS1m0PdzubytTnyxOlcRKdvtIeAPnIYm447bOzXBYUc xkaGOk+OZHZcDMMgh3HqeGydQ/CBIIeAPnII6COHgD5yCOgjh4A+fk8TKI7vTQo5DJHi78WSJDly jgV6Qr8UH+CIogpyCOgjh6hj13R45BDQRw7xAfYPVZBDfIAcqiCHqONkqOGRQ0AfOUQdXdPhkUO0 4NDFwMghoI8coo6rIQ6PHKIFQ6YDI4dowVDNwMghoI8coh1DpkMih2hBv3Rg5BAtyOHAyCHaMWQ6 JHII6COHaEfXdEjkENBHDnEQu4iDIYdosVgsvv7669///veLxUK7LpPAdYTRYrfbycR3332nW5OJ oD1Ei9vbW+0qTAs5RIv7+3vtKkwLOUQLe9BiPp/r1mQiyCHakcAhkUO0u7m50a7ChJBDtJOrY5DG YZBDHETXdDAcP4zZ69evf/jhh4uf/m//9m9///vff/3rX/uqz0cffcRFqFqRw8htNhvtKvzL559/ rl2FQNEvBfSRQ0AfOQT0kUNAHzkE9JFDQB85BPSRQ0AfOQT0kUNAHzkE9JFDQB85BPSRQ/yoLMv1 el2W5XK5XK/XMnO9Xi+XS3eOu9hyudSrb1TIIX60Wq3KsjTGbLdbmViv16vVarvdyhyJorvYarWy +cQ1OP8QH1itVna6LEt7+uJ2u3VbP1lss9nQJHpBe4hjpOlD38ghDrJdUGNMWZZuU3lkJi5ADvEj iZwbPOmUyniM20c9NBMXY/8QP5IhGZm2E4diZheAF7SHgD5yiPNwoKIP5BDn2Ww29gAjfCGHuATD pH6RQ0AfOQT0kUNAHzkE9JFDQB+/p4mcr8N9f/3rX//rv/7rykI++ugjL5WJDzmM2du3b30VlSTJ n//8Z1+loYZ+KaCPHKKTNE21qxAzcgjoI4foqigK7Sr4EeAbIYfoKsDNNxrkENBHDgF95BCdxDRe GuB7IYeAPnKITgJsQ2JCDgF95BDQRw4BfeQQ0EcOAX3kENBHDgF95BDQRw4BfeQQ0EcOAX3kENBH DgF9XL8UpyVJ4k5UVaVanQjRHuK0+XyuXYXIkUOcNpvN7PTt7a1iTbwI8IJX5BCn3d/ft07Dl4S+ Prqwu4gRbDBJEtxmT3uIM8SxoxjgNT7IITqRBN7c3GhXJE7kEJ3IUE2e59oViRPHD0cmyzKtl/6P //gPxVc3UX8LkMPxiXhzPC7LsljfO/1SQB85xORwHB9AC3II6COHmByO4wNoQQ4BfeQQ0EcOAX3k EJPD8UMALcghDrLtxrkNSIANjovjFhiNoijk7Ao70Z1d3p7Fj+PIIdrZRuOa1iO0y0+IAJtrcjhd RVEkSZIkiWyXWZbJw3OfbudICW7jqXu+4iEB9ks5/3CiiqIoikLaK7luUu3hyacbY+RZ9unmpQG0 4Qyw5QkTOZyoLMts2GTCZqkLifGhAnEu+qX4kfQhu/fZ8jyvXvRZr0kghxOVpqnbe8yyrCiK7iFM 09Tu+0nD6BaIc5HDicrzXJIj4ZEE2jEb2+10J9ynp2nqPt0tUPIpwW4+Ea3YP5yu2jWXaruLrRNH nt46hy5rR7SHgD5yCOgjh5iWMPdXySGgjxwC+sghpoV+KYB25BCTE+D5FuQQkxNgDvk9zfiEeVKf MWa/38vtSnsSYH58OX2mGdCR/Bg18FsUdjm7cnj0S+GNtFdhDkhaYTaq5BA+pWkacg6D7dKTQ/gU fpMYZrc5xL4yxi7MfTATcMVoD+FfnucB9gDPuuDAwMgh/AuzdxpyDgNtpjF2AR7DCLZTamgP0RNp ecLpnYbcGBpyiP5IYxhI7zTwHIbbUiMC4fROQ+6UGtpD9Eour6jeO82yLOTG0JBD9C2QsdMQ2uQj yCF6l+e5Yg7VvwK6IIcYgmLvNPARGkEOMQTd3mn4OQx6EAmRGX7QUkZoyCHwL8Mfxgj8cIVFvxTD Gbh3Gv7hCmsc3xaISZZlwzSJY2kMDe0hhjfM2OmIGkNDewgVAwyfjKgxNLSHUNH3ZWzG1RgacggV 7oBNH4EM5Mfl3Y2p7UZkkiT57LPPiqKYzWbv3r3zVazsfJJDoJNf/OIXf/vb32Ta43Y4rj1DwXX1 oWOxWOx2O+/Fjm7PULB/CB2Pj4/z+dw+vLu781XyuHqkghxCR5qmbhT3+/31ZaqfcHwxcgg1tSh6 McbG0JBD6LJR3O12V7ZmI90zFIzTQJmE5+Hh4enpyUtRYzS+EV4Mb4D9rv1+//79+4uD9PT0NJvN er0Lqumz00t7iE5Gut/lUa9fRuwfAvrIIaCPHAL6yCGgjxwC+sghoI8cAvrIIaCPHAL6yCGgjxwC +sghoI8condJkqiXEDhyiN5xbt1J5BC9s9cLzrKsKIokSZIkkYcy0fxvx5KlhCRJ7ElJzTmXlTy0 CjglTdNrnm43M2NMnufuROt/5dYXrSXUauUWIs+qzTlZcndXroTjOB8fp115pzR7Yd/mxMmZrQ8P LWM+7AN3Kbm7Xm8XR78UI9a8N8ZgNzn1ixwiOEVRHL9Qjdzf271plDxszjm3ZC1cnwYDkYS4fcVm N0/6lrKb13yiu6TssMmojPuU5pwjJYeDHGIgMs4h0+4e2qHp1ifWNEN1KGaBj4PQL0WPRrq3Njxy iB7JUbsuXcH+rko4iptekEP0yB6167hkH+1nfyV7RA4RkP4GM8McJrXIIaCPHAL6yCGgjxwC+sgh oI/f06CTURyFG6//D79y7o5wYrdYAAAAAElFTkSuQmCC ------=_NextPart_000_0000_01C80F9B.F234E740 Content-Type: image/png Content-Transfer-Encoding: base64 Content-Location: http://www.cs.wisc.edu/wisa/papers/security03/images/cih-annot-cfg.png iVBORw0KGgoAAAANSUhEUgAAAkMAAASPCAIAAACCoWw5AAAACXBIWXMAAAxNAAAMTQHSzq1OAAAA HXRFWHRTb2Z0d2FyZQBHTlUgR2hvc3RzY3JpcHQgNy4wNULvQ9UAACAASURBVHic7N0/j+TIff/x auMUeH3w2ZAAW8DMSJEwcNAMlOyun8AcYF0ggJ1xH8Ps7egJ2Mo0K01wgUP9moADNqBgJUOTyIAD 722ioBuwsalvFpCTg23YUGIB/AXfm+/Vkt09/FMkq4rvV7DgcLvZ1WSxPqwim1yUZWkAAAjWH01d AAAAeiHJAABhI8kAAGEjyQAAYSPJAABhI8kAAGH7YOoCAPDXbrfbbret3pIkyXK5HKg8/thsNpvN ZupSBCxN0zRNXS1twe/JgHnK87wyR0IrSRKHn6JBWFls6IG3Wq2Kopi6FAFzuwLpkwHxs7tW2+22 EipZllUmHKosU0uyvaeFCT3bMCH6ZECE7MAIIiq0g2gXeIhkdYU+WU9uVyBJBkSiMlroc249qJ7E vqUaSdYTo4sAvmYHWNDpZVsul/JFsiyTVMvzXFItmu8Ih0gyIEh2gPnWX3Grnmrr9TpJkri/NVoh yYCQ2NduzLApl1TLsizPc8lyumgwJBkQCs2wGQZYna4EHXUkz+aMJAN8pxlGe10nkUaezRxJBnhN Yow2+jg7z+izzhD3XQT8JTGWZRkx1kSWZUmS1O9dgujRJwM8JS0yPYxWJPLzPGe9zQp9MsBHu93O RB1jw91+d7lc0jPrLNDbIpNkgI/6n+9pfrP2xWLRarFdS/TeQlar1YMl6fxZy+Wy7S38YQ5vF9Oy koyPJAPi1DzJmt+ybrVaOXkSR/OFdA6zJEmkX4vmHD5mZWScJwO8039kbLPZFEWxWq02m400T3q4 Lc+FqkxLmMnrZQlyTzyJLllIURR2rujC9ZX6p0zr6+3b69kfoaWSV9bvwidl69a8Zln24x//eLie 2be+9S2HS1ssFrqejTH1FWuMKctSpnVjVdbYoS1i7utDfXNU3qt/Ht8uPioBeEZuyNRnCUVRyL9p msocY4zOtKflv8qylFTTF+i77An9006X+qL0o2Wm0sLUl1yfPjSnofV63XMdHmd/l/4OrXD9FN0o 9hqulOfIFtm7OfZ+l73bpf5nf25XIKOLgHd6nuaRA+rFYmEfaJdludls5GyHPa2kTyDTTQ7D7SbJ 1AYz7aXZBev8pTqI4HoZXWMaVJWutu3IFtm7OeqfEjCHqQjAlfV6vd1uu73XPtrVnpbd36r0vYzV 09LXC3O4T2Z3F/b2GOy+nS7E7iNWJkp3fbL+ndoHjdYn27vG6r2r41tk7+awP/34dqn/2Z/jFehw WQAc6hZmcjyuLZ0csGqA1f/VBk5fo8NQlf8q3x/RkiVXAlLfLvPlNZ9++qm9HP10bV41a7W09tdp uwYkxjofBzTksCGubAt7o5TWannwow9tkSdPntibo55qx7dLeWDT9P/WrhZVliVP2gT8lef5aPep qlwbsrdlkGGoDpdgLBZdmpq9Y2jH6V1R2n5WW6E/abPbFnHI7QrkPBngryzL5CGTI3yWfVLtUH50 vkq7W4y1fQv3XWwusj4MV+EDXpNncV1dXQ39bMmyLOUaEB3d2mu0nxy1+iAeeTNzJBnguyzL9NmS gw42hnicTobBkGRAKPTBJev1euj+WRB00JVVAZIMCIn0z3a73Wzbcfus4dy+Ow4hyYDwLJdLfXyJ tuxxP42TAMMRJBkQMG3T5bI9GXg0UaSangCTR2YbAgyHkWRADKSVl38lA+z7XQURbBpd5v30IsDw IJIMiI2OPQrprmm2SUKYqbs49mih5pa5D13SC61wjw9gdio/ta4knK1DZ87uWj34KeEmVuj3+Jic 2xVIkgF4j9tbioSbVcc1f5DpVN69e3dycjJ1KY4hyQAAx8yq18h9FwEAYSPJAABhI8kAAGEjyQAA YSPJAABhI8kAAGEjyQAAYSPJAABhI8kAAGEjyQAAYSPJAABhI8kAAGEjyQAAYSPJAABhI8kAAGEj yQAAYSPJAABhI8kAAGEjyQAAYSPJAABhI8kAAGEjyQAAYVuUZTl1GQAAzmw2m81m8+bNm8ePHxtj iqKYukSDI8kAICqLxUKnT09Pv/jiiwkLMw5GFwEgKqenp1MXYWwkGQBE5eTkZO90xEgyAIjK8+fP dfr169cTlmQ0nCcDgNjoqbKZtPD0yQAgTk+ePJm6CCMhyQAgNnLRhz3MGDeSDABiI78kS9N06oKM hPNkALrY7Xbb7dbtMrMsc7tAf+R5vtlsRvu43/3ud//5n//5V3/1V6N9ojHm1atXY36cjSQDsF8l q2Q6SZL6K5MkWS6X/T/i+GcFnXOffPLJT37yk6lLMaDLy8tf//rXU336B1N9MAB/2IlSSRHNjyGC ZLlc7o3AymfleS7/VsrWOUERGZIMmCmJh+12WwkGD7s+9TSV6FXyFTwsOcZBkgFzoR0vbfqTJAm0 9bc7c1mWyVeTTpt+Nbpr80GSAZHT7ot2XAJNryM02OSrSaSt1+skSYi0OSDJgGhpHyXcvlc38mW1 r0akRY8kA2Kjo4hzC7A66atppEmuk2fxIcmAeGiGzTzA6nT4Ufup5FlMSDIgBmRYQ/aJNPIsGiQZ EDyJMdrl5jTPjDGstAhw30UgbBJjWZbRIreVZZlcuz91QdAXSQYETEbJGFHsLMuyJEkIs9CRZECo drudmeLE2Jh3wh2BXgkydUHQHUkGhMq+0VRnq9VKny/c/C09P7SubRnc8rZTe35+PvLyh/7EgZBk QMD6nxsrisJJSdqaNrpC8fbt26mLEAauXQSCJEOLDm02m9VqVRTFZrNJ01Sni6JYrVYysxJ79nyZ lqdEyXvt/5WFp2kqc2R8crGoPlVKyyDdvrIsF4uFvkteUF9gURROHin54x//+Dvf+U7/5ez14Ycf tn3L7e3t5eWlhNn5+fnFxcXt7e3FxYX818XFxc3NjbxG/uvm5kb+V9+rb7m5uXH7dXzD88mAUF1d XV1fX/dfjiaKHS06LYEkMyUz5L8OzTf3eSP/1hcuE5UYe/BlhyYk0vq3Y3meDzrG2O35ZOfn55pk xyfs2Dv03iPLPzKnIZ5PBqCLJEl2u90IF98fyonKfOlLpWmq/aehC2aMcdIbc97BHZ/2xuaJ82RA qJIk2fvAZef0Eo9KOFXm67ChzNEhQf85uXYGEyLJgFAtl8v+v4XSU1ZPnz419+EkM2VaTkotFgs9 LyVvrMyXmdohE/Jf9jiknYWyfLl4Us/AyYCkXQB9u70csVgs6ifb2srz3PPbo1xeXsq/t7e39oQ6 Pz+vjArqCyqvrC/2/PxcX1OfEwrOkwFh03t8TF2QsfXPMDNijPU8T9bzNSPgPBmA7uwf9s4wzzqT 26M8e/bMz96YXqboSuWHYpFd0EiSAcGTR5bkeX51dTWTZ5L1+TmaPjfAyZWfA5FrER8Mm+Y/ZPah 3zYcRheBqHje1XCr7UUlEz77ptvoYkPOO3AdTDu6SJIBEdLnb3l+LcM47AdGT9VhHTTJfMB5MgCO ZVmWZZm04Ov1ep6RZgfYTAZdZ4skA6Il588qkWaMiTjVdPyQAJsVkgyIn0aauR94jCnVKull7ruk U5cL4yHJgHmRJl7+1fE3+yYXnmeb5paxokvKTHrNFkkGzJf01cy+YJO0sO/hNHLC2YlljKmUh+iC jSQD8BUNNmNdp26P3emE/JfbexUeWSyjhTiOq/ABuNH2DpCzCie5pfKYn/jmzZvHjx+P+YlTPbXV kGQAECV5ZunUpRgJ98IHAISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCE jSQDAISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCEjSQDAISNJAMAhI0k AwCEjSQDAIRtUZbl1GUAADjz9OnTd+/e/f73v3/06JEx5osvvpi6RIMjyQAgKovFwv5zDo08o4sA EJXT09O90xEjyQAgKo8fP947HTGSDACikqapTJyenhZFMW1hxkGSAUBU0jSdyaCiIskAIDZ3d3fG mJOTk6kLMhKuXQSA2Jydnd3d3c2neadPBgCxmcmFHoo+GeCFPM+P/O92u+3/EUmSHPnfLMv6f0Sg VqvV1EVwb7vdHt/igdp7DQtJBgyuklIaS4camtFC5VB8HiphrGm3Wq1mco1f6A5tKZIMcGO329k9 J5m2kyBJkuVyOUHJeov4qwmSLBQkGeCS3bjbwziht+ltHVoPYfXeSLJQkGRALzoQN+fcamJvtnm+ okiyUBzaUh+MXxQgFJVRNWmOw+ptjG+5XGpoybqS1Wife/M51RAikgx4z970mrA8EbCzjVTDEEgy wBirhU2ShOZ1OHaq5Xm+3W7X6zXrHD2RZJg7PQGWJAkjh2OStZ1lmd1RYxOgA5IMM2WPItJ6Tks7 anmeX11d0UVDWyQZ5kj6YTSXvsmyTLtoxhi2DhoiyTAv2krSD/OWdNHkLBqbCU1wB2HMiMRYoOfD NpvN1EXYY7hSZVmWJEme57vdbqCPCIufFUBNWzySDDMix/hOxqw2m03zXXexWLRa8t6ZDe9yu1gs FouFw2bFLvzi3t5SrVaryv/2/Gj59d56ve65nAg0rwDKn5owApIMc5HnucNbg7dKsuZ30lmtVvro etvemUc+rtXrKwU48l9FUZRlWRSFvKzyKfK/9hwnzaj0zPovJ2jdNqg/NWFwJTAP6/Xa1aJkR03T VCZkjuxQaZpW/pRpeZm9wxdFIf8lM+WN9i5pz5f/kjlFUdgT+gJ92aEl2H8eWkK9TdA59n/ZM+vl tAvQdJ0e9eLFCyfLOcRVOW32mqmv+UqV2FuMbhXALsChRY1fE1w59GW57yLmIs9zV6fHNptNmqbS LZM0WiwW0hbIf9X/LMtSjl6LopChGNn15L90Qv9crVbaoZFFyX/pe+XTtZVUR5ZgF8bcdysrS9D/ rc+x/6s+8/jLevrss88++uij/ss55De/+c3Pf/5zt8vcu2b0yGZvHZDNIW/vXAHqBfChJrhy8A6Z bgMT8NZ6vd5ut/2XU9mRdL4e29b/lJdp50x6Y/Iy8/5xrjlweF5/pUzocuovqy+h0kG0l1BvDY98 7qHCHHlZH3IfkP7LOWKgPll9Yu/KkXSpbMrOFaD+eh9qgisHO6DOPwnwk6sG0d6XdEDGHlSs/2m3 KYfGcPa+UqbtOLQXXmmMKgusLKFemL1LaLJAaXn3lr8cIMmGjrFyxCSzN8reILGL1K0C1AvgQ01w hSQDyvV63bNZrHS85KB17+ku/dPuimlbpifM6901/SxZuN1S6DGytiBagHrzVFmC/elSgMoS9C31 71s/YJf/rZe/soT+CdF/kzXhPMkqW9+uBpV1eKQA3SqA/fa9i5qkJrhyaEtxngzzoj8pG//+EXoW xD5HUn+N6XqhWuVM21T2nvXpYLfbrdfrZ8+ejbCl4ng+WWXT+1MTHDq0pbgKH/MiP1HabrfjX9ht n3g/1MT3afq11XD7K6Lm5FdEMt2zAHKDj+vra25Y1Vw9NnyoCeOgT4aZkkP+kW9WK7u3jjWhQu8l NnKnOY4+2RzwzGjgPcvl8vr6euTniXDgeIj9bB36YWiLJMOs2c8T4Qb547PHeEO8GSY8QZIBxtw3 o3YXzZBqw7Afz21YyXCBJAO+pl00Q6o5ZT/X1PB4brhGkgH77U017UkYRsOO0mFD+l4YAdcuAq1p D0MvtDPzbqnruWWCWiFcuxiKQ1uKJAMcsK9csONNBNSmH1EZIax/zXA7qSRZKEgyYALHE842bQxU Uspmd7NEuIl1CEkWCpIM8M6h+4zUQ8XJM0KbLza+rDqu7eOYg/Du3buTk5OpS+EeSQYAczGrjib3 XQQAhI0kAwCEjSQDAISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCEjSQD AISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCE bVGW5dRlAAA4s1gs7D/n0MjTJwOAqJyenk5dhLGRZAAQrSdPnkxdhDGQZAAQlcePH+v08+fPJyzJ aDhPBgCx0VNlM2nh6ZMBQJzmc8KMJAOA2EiG2cOMcftg6gIAABw7OTkxxhRFMXVBRkKfDABiI0k2 H1zxAeBhu91uu93ac+w/kyTpvOQjy0mSZLlcdl6yVz755JP/+7//m7oUw/r1r3891UeTZADek+e5 TEjGjJ8uh1JTS5Jl2aAFGMInn3zyk5/8ZOpSDOjy8nLCJOM8GTBrdm7ZoZVl2VSBsVwuK2GpJZGQ kzLbBQ4x2+AQSQbMTiUJpJsVRBjYIScF1mzTrxPEF4FbJBkwC5X0mrDL5ZZmm3ydPM9JtRkiyYCY 6eChiSi9jpAvaKeaPR+xIsmACNkXTcy2EdcvrpEW08WQsJFkQFQ0w2YbYHV2pMnAI3kWGZIMiAQZ 9iAddSTPIkOSATGQ0TMyrAk7z1hjcSDJgODleU4Po60sy3a7HasuDiQZEDAZUaQt7kau4JfuLCsw aNxBGAiYjI+N3ApvNpsxP25oWZZtt9vdbjd1QdAdSQaEyv6tWGer1UqfL9z8Lf0/t6JtGdxKkqRy p0dPnJ+fD7pwUZ9zeXk53OcOgSQDQuXkgoWpHmE1bXRVLJdLP5Ps7du3Ay358vLy5ubm7du3Nzc3 klu3t7dv376VObe3twN97kA4TwYEabfb9XmWSt1ms1mtVkVRbDabNE11uiiK1WolMyuxZ8+XaXm2 hrzX/l9ZeJqmMkfGJxeL6rM4tAzS7SvLcrFY6LvkBfUFFkUh/9tHkiSfffbZRx991HM5h3z44Ydt 33J7e3t5eSlhdn5+fnFxcXt7e3FxIf91cXEhkXN5eSn/dXNzI/+r79W33Nzc1BcuMy8uLiTJ9L0X Fxc6HQqe6gKEKs9zJxeRa6LY0aLTEkgyUzJD/uvQfHOfN/JvfeEyUYmxB192aEIirX87NvRFjN2e 6nJ+fq5JdnzCjr1D79275PoLNC9b4akuALx2KCcq86Uvlaap9p9GKFv/3pgK+vJFh72o4DpkhvNk QNCcXPTxIL3EoxJOlfk6bChzdEjQf+OsRq/ooGK3HphvSDIgVPqArj4L0VNWT58+NffhJDNlWk5K LRYLPS8lb6zMl5naIRPyX/Y4pJ2Fsny5eFLPwMmApF0Afbu9HLFYLOon29qSFejzzT4kci4vL+VC DJ1QcsFhZYSwMlEhJ8nkMkWZvr29PbcM81WGwnkyIGyzvUtF/wwz9z8tHyHGep4n6/maEUx7now+ GRA2+WHvDMfH+vP51ovOr4M/f19wvxg7jis+gODJLQSvrq6SJPGzXXau58/R9Hb43q4uuRaxfvV8 RfNhQB/6bcNhdBGIh/8NtFsdLiqZ6tk33UYXG/Lhqo1pRxdJMiA2OtI4kzxraNrVMmiS+YDfkwFw SZ+/RaSZ96+wn/N6iBtJBsRJW2070uZzlSMBNiskGRA5O9K22+16vZYbNkaWanoCTM4UGgJsTkgy YC6kZdexx+12a98APrhg0+iyZVlGgM0QSQbMkd3cSyQo+xb7/qSCPVpoFzK4AMYQSDJg7pbLpYaB PRSp/0rXp/4QGbcpUv9xd/1z7W4loLgKH0A7R+4n0vx5lUcerhZlUP3whz/8wx/+MOYn/u53v/v2 t7895ie+evVqzI+zkWQAECF5ZunUpRgJ910EAISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCEjSQD AISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCEjSQDAISNJAMAhI0kAwCE jSQDAISNJAMAhI0kAwCEjSQDAISNJAMAhG1RluXUZQAAOLNYLOw/59DI0ycDgKicnp7unY4YSQYA UTk5Odk7HTGSDACiYqfX8+fPJyzJaDhPBgCx0VNlM2nh6ZMBQJxmcpLMkGQAEB/JsJcvX05dkJGQ ZAAQGzlVlqbp1AUZyQdTFwDAxHa73Xa77bmQJEmWy6WT8nhutVpNXYRGHj16FERR0zTtn7hc8QFE Ls/zyhzJrSRJHnxvJZ/qi9rr0PKzLGvyds+tVquiKKYuRTycrE+SDIiHnTTb7dbOkgn7TJU+X6Vg wcUbSeYWSQbMnURXoNmwN3f9LzxJ5hZJBsyO9m8CavqbqwSzn+feSDK3SDJgLuzui5/tu3OVc3L+ BDZJ5paT9cm1i4C/7Nbcn6Z8HPb3zfNcV8Xc1gOaIMkA78w5wPbSlaCRNpOOKRoiyQCP5HkuZ4kI sL3sSJMVRZ7BkGSAJ7Rpvr6+nrosAZBII88gSDJgejJiRoa1ZecZvdg5I8mAKclV9fQq+siybLfb 5XlOmM0WdxAGJiMxlmUZMdbTcrlMkuTq6mq3201dFkyAJAMmE+6Y2GazmboIVcvl8vr6uv+tkOPg 4QYaFEkGTKPh3XiP22w2DdssfYhww8Ue/1+5yXp9mYvFYrFYtG1GV6uVvLHJpz/IyYoNmm4gc3S7 O9lYniDJgMn075A1T7Lmd/NZrVbHn7Kh/7t3mWVZtn1IR1EUlUV1DrNA+7hu2ev/+HZ3srF8wBUf QKg2m01RFKvVarPZpGmqR+JpmhZFYf+ZpulqtZIGSF4vS5C7BEl0yULkjfoRunB9pf6vLP9Io7b3 vfKnTOtnVW5WJKXt/MyqoQcYHz165GpRi8VC17wxZu+qLstSpnXzVVZXfcXqfz24jepL2LvAQ1vK IyWAKbx48aLnEuTouCiKNE3LsjT3x8t61Fz5syxLSTWZqbt/ZUL/tLNEP2XvKyvz6++tFEOLLTMr 79275ObW63Xn9zZhr4eeDm0C/QjdTPZ6qxRG1/OTJ0+OLPnQp3fYUscX25aT9UmfDJhGkiR9Lhy3 z4WIsixXq5Ues1f+FNLdkekmx9eldTjf9gHE5dEBw54dr0PiOEmm60rXj3SP9j5eWddz53NXk2wp x/qHIYBu+vQe7CPZ9F5pHcVX/jRWb0lfLMzhPlmlc6B/mgZ9Mvu9e/sWdr+wdNQnG7pDVo7VJ6us 6tLqZFcKY79y77uOf3qHLXV8sW05WZ8kGTCZ7XbbreWVA2Rt1/TAVObrYKP+qS2RBpKOF1X+q3x/ /EoXqy/QxX766aeVN5a1NKq8V0PUHhE11kUE9hfssFrW6/V2u+3wxlZcJVll69ibqbQG/R78XF3P lQ1kJ1A9isr3067Vlir9SzKeTwZMSX8cPc7H2RcXHLoWQAaXWo0mLRZftSQ60YG+VwvZ3G63W6/X z549G+E35oE+n6y+aZxsrP6crE+uwgemJDenyPN8nJtT2Ne2HUqLDmdE7Eat80+UZLrD9fdy38Xr 62tulXLE3uDpubH8QZ8M8MJoj92SZsgejArXJA/BCbRP5i2eGQ3EY7THlMRx8MpDcGAjyQCPaJ6t 1+skSbhHfoWcViTDUEGSAd7JskyeVCIXNxJpdoDxQG3UkWSAp5bL5XK5rESamc2tBeVbG2MIMDyI Kz6AkMiFIdK4m1GuEBlTJb2Ml7HNFR9uccUHMDvSssu/Ouam7b7xsuk/RHPL1LI5oG8BH9AnAyKx NxjE5F03u2zm/eJNXra26JO55WR9kmRAzCp31NU4sXOuomG0VMKp4tAHRdDZIsncIskA9OX25vER BNWD2j4TYCpv3rx5/Pjx1KV42N4b/LdFkgFAhGbVd+S+iwCAsJFkAICwkWQAgLCRZACAsJFkAICw kWQAgLCRZACAsJFkAICwkWQAgLCRZACAsJFkAICwkWQAgLCRZACAsJFkAICwkWQAgLCRZACAsJFk AICwkWQAgLCRZACAsJFkAICwkWQAgLCRZACAsC3Kspy6DAAAZ54+ffru3bsvv/zym9/8pjHmiy++ mLpEgyPJACAqi8VCp09PT+eQZIwuAkBUTk9Ppy7C2EgyAIjKycmJTj9+/HjCkoyG0UUAiMpms1mt VjI9kxaeJAOA2Jydnd3d3c3kJJlhdBEA4nN3d2feH2aMG0kGALGRiz6eP38+dUFGQpIBQGzkQo80 TacuyEg4TwbA5HnudoFZlrldoCf0SgrPyS+jkySZuiCNpGnaM3RJMmAWdrvddrvVP2W63tIlSbJc Lp18xKEP6vMRk1utVkVRTF2K2PRfqyQZECHtY1WCZPKukrcFa4gkGwJJBsAYqz+03W7Digc726Tk PnfaSLIh9F+rH7gqCoCR1dMry7Ig0sumBZYJ+VI6SulzqsEfJBkQGG3rkySRhj649DpiuVxqdNmp RqThCJIMCEMlwGJKr0M01eS7r9drDe+piwa/kGSA73QUcSYBVieRlmWZHWnzXBXYiyQD/KUZRqst NNLyPM/znP4ZBEkG+IgMO05WS57nMtxKns0cSQZ4R2KMBvpBmmfGGNbVnHHfRcAvEmNZltE0N5Rl 2Xa73e12UxcEkyHJAI9ojE1dkMBImDm/eyRCQZIBHiHGOtMfVk9dEEyAJAN80bNLsVgsXJWksljV /E7wm82mw2d1e5dKkqR+C2OfDbTJZogkAzziZ4esLMuyLIuiaJg0q9Wq80M6+oTZcrkMK8ngCtcu Ar4YYmhxtVpJNsi9wit/bjYbSZ3NZlMURZqmi8Wxu4pLPj24EDuN5MVpmsotYu33yrQGpLy3Twoa Y7773e8Oerbs0aNHwy1cVFZvZU59bRtj7K229wUPLqHy6Q2319CrooUSgB/W6/V2u+389vrunKap TEijU/nTfpc9p75YIW9vshCdb7eS0mjqe4uikAlpNI98i1bW63Wftz9Iv4ITD26yvXOabLJuS+iw vZzov1ZJMsAXch+mzm+vt2v2nMqRayWE9r597/wmC7GT7HgJZaarJFuv16EnWX31Hpnz4CZru4QO 28uJ/muV82SAL9z+gEzGiOxrNCp/dtN2IfpiGZKq/CmjVf1Lpfw80dhQfZMZF1ut1RJG3l7OOElU AE5It6zDGGP9pIUcOOt4kbys8mdpDR7qn3sXax+G/wKcRwAAIABJREFUP7iQSh/LWAOb9p86flUp cNvvLnqOzTbksE/WcJPV59Rbb1PrN7ddQuW9zbeXE/3XKs+MBrwz5r1xj1/i0W0hcvze4cIN6ZR0 KMBoa8yHZ0Y/uMn6v2Bk/dcqo4uAd0K/Y0W3NOp2/T03xYfhKnzAT/Isrqurq0EfxOXkl7l7F9Ih zNq+ZbfbzfBBZQ9usv4vCBF9MsBTy+Xy+vraGHN1dTXQTZjkHEPPO2s4WUgru91OnudyfX09qxgz DdZ2/xeEyK/RUgB76YO45jyMJrdXlvUwVYD5cJ4sPv3XKqOLQACyLJPxRrm4cVaRZgfY3MYS0RBJ BgRjuVwul8tKpJnAf0R1iFzwQoChCZIMCI9Gmrm/eE9afGNM0H01O72MMQQYGuI8GRAPTQJjjISB 8bjHpj8zsAvsbWkF58mG0H+tkmRAtOxfpGlHR42WGTIWeqgknkdXBUk2BJIMQDv1H1xX+nD9aW5V lhn0yKcgyYZAkgEYRKUjdUQE+dTctDfPffPmzcnJycnJiasFbjab733vew4PYrrpf3BAkgFAAOTx mH0eQ1r39OnTzz///NNPP3358qXDxY6Pe3wAgO+GiDFjzPPnz40xP/3pT1+8eOF2ySMjyQDAa52f LfAgXWboYUaSAYC/NpuNPN9yoOWfnp7KRNBhRpIBgKeGjjFjzOPHj3U63DAjyQDARyPEmDGmKArt lhljPv/880E/biDcrQoAvDNOjNkePXp0cnLy+vXr0T7RIfpkAOCXkWPs5OTk9PT05z//+eQ/LOuM PhkAeGT83pjdDwv0Jib0yQDAIyPHmE0uyg/xcdIkGQD4Qn4BPWEB0jQlyQAAHQ10I49WAu2Wcd9F AJieDzGmFovAooE+GQBMzKsYM8akaTrtXf/bIskAYEq+xZhx8ZiVkZFkADAZD2NMhNUtI8kAYBre xpgJ7dIPkgwAJuBzjImArsgnyQBgbP7HmAmqW0aSAcCogogxEUq3jCQDgPEM9wDoIYTSLSPJAGAk 498duL+iKEgyAIAxYcaY8jzMSDIAGFzQMeb/2TKSDAAGF26MmRDOlpFkADCsyZ/V0p/n3TKSDAAG FNA190d4Xn6SDACGIrcu9DwGGvL5ToyBPYQGAPz39OnTzz///G/+5m/++I//ONzTY3XePreMPhkA OPbu3TtjzK9+9avT09Opy+KSt2fLPA1YAAiUdMj0z8ja2NVq5WEv84OpCwAAUZEOmTHm9PQ0jjNk /iPJAGA/+TmzTkssHe+RnJ2d3d3dGWNOT09fvnwZZZJ52C0jyQDga3Z6GWPsC+jlv1arlaRa/dr6 zWYjMfbkyZPXr1+PWezR+HmqjPNkwCB2u912u9U/ZTpJkrbLqb8xSZLlcumijPiaXF/esONlDgSe dMg+/fTTly9fDlraaXl4BaN3BQICpdG13W7t4MmyzNVH5Hmu0/anOPyIWWmbXods7j169ChJkufP n0c5qKg8HF0kyYDuJFomDJV6fJJqx7lKr6mW7wMP71pCkgGt2X0jr5LD24JN68ipr0HFmmryvbz6 OiQZ0JR96svznCDSpkqv44WJJtV8G2AkyYCHaTAElwp5nsvAY3Al78b/bpBXEdsZSQaEJI4kiONb HOJ/eh0Sbqr5dvmiX6UB/KFjidG0/vKN4riIP9z0OiSsb0SSAQGIqdGvkJHSEOM53B5MW/6nGkkG +C7ctr6hgL7gfNLrED9TjfNkgNcCauX7kE6nt1/Tz+Z7Wl6FOkkG+Mvz9t2tPM+9Gj4lvZqbPNVI MsBfeZ73jLHFYmHCeSTV5GFGevU3yTr0Lcl4ZjTwFfvXxM2tVitJL5kuy1L38OFuGb64p3M6f5Z9 m+NxyO3khTEmTVNZacO1jPaKGof9iVJDFouFfF/nZNVpxZOP09U7E/TJgK907qDodVz2BV3DHbTq Xe9kiEk+RZKswxBT/25oE5OPho3Mrgl7q8cIBu2r+dYnI8mAr1xdXV1fX3d4o7RQ8tgqY0yapkVR VFJNGhQ9atZXShTJnw2bhr2tpOnauFxdXXV41kxzv/nNb/7f//t/Ew4eShdQ1lLzDaH/Zd4fK5al yet1OXWV7SITUzX99mGEq2MI35LMlADKsizL9Xrd7Y26H9k7lE7bDUdRFDKYVt6Pp8kcGR2qvFjs bXfqn1KZbu7Fixcd3tWcfEGl33RMsmZabQj7v/YurTy6wu3/0hbfyXdpS6NLJhwu1tWinCDJgK+s 1+tuYfZgku19sT2nVfvuMMm2223n/O5goFb1QZpk9ZmVOZUNsbeQbZPMztF25e5qhPXsW5JxxQfw leEG2XR8SQZ57D9lUEvn6NUBqj4nTdPKKZDORr7cY/JrE5pvCON6AG3osTgZ9tTrSka4jsYvU0cp 4JH1er3dblu9RVuKJ0+e2PuUfXQv8/UwVv/UMS6Z07xnpgFWmdmq5CN3yI4Yug9hrF5Uww1hB4C9 XWS+PbH369hvrPzp0IQjt771ybjiA3iPq2v5Ol9M2Pnj2n7W5D8mqxvoEkffbhLYkw8/wvPtio+o NjDQn8PbfPQf/Wv+QaZlanoYYxVOUk02wXBJVvml2pGrGXvyIb0q5Zm8DDaSDKja7Xbr9frZs2c+ N/SdhXib/87tuFxML6NwA5ZvGL6ll40kA8IQ5a2EI/hSPrfv/YXyE3KSDAhGTA/bjOm7qDhSLZT0 spFkQGDyPJfhuEAzQO8nGWj5mwg0DEywMUySAUHSPAvoDFPoGdzNZrP52c9+9vnnnz958uTk5MSr VAs6vWwkGRAwGaPzOdJ0FFHMKsNM7SaHPiSHD2VwjiQDYmBHmvEgMGQI0eeIHVrl7sB7X2DGSpQQ RztbIcmAqGgfSFPNjBJsevZLP3eeAWYaZFiFxoy8y1XMRJ9eNpIMiJkdbOb9ezl2S5rKaKG92NlG l9Dk6JMZ/eMnysHDB5FkwOxU0kjtnXnoRsaTD2D6w0mGHVqsZtKRhc8zvWwkGQB0pBk2dDNazyrS y0aSAUBrbU+Guf1omZDn6dBmGv+S7IOpCwAAx2iGTRUhXjXZ2IskA+Aj+1oMf7pBEV+OGDSSDIBf Brqgw5XRHtaD5kgyAL6wz0h5mxYkmYf+aOoCAIBZrVby1MqiKHx+nBjnzPxEnwzAZPw8GYbgkGQA JuD5ybAj7HuCwBMkGYBRVW5XH5ywcncmOE8GYCSVk2FTFwfxoE8GYFj2QGIcAebbHS5AkgEYSrgn wxAWkgyAexPeJnFoejdh+IMkA+AMV9VjEiQZAAdmNZDIhfi+IckA9BLxQOIh0Ud1cEgyAF3MfCCR uy96hSQD0M6sBhIPIcm8QpIBaGqGA4l7pWnKqTKvkGQAHjDzgcQ6ksw3JBmAgxhIPIQk8wpJBmAP BhKPI9e9wh2EAXxts9msViu5h0VZlq5iTHswi/eZqDs38tXkvsk2Wb0Rf/HxkWQAjLnPMO2HOeyK rVYruwdTWvSjXX3WmI7fs0q/dVEU+mVlQtdGoF/cQyQZMHfSadhsNhJgzsfN9rbX9hm4KBv0I19K 1nCsX3wSnCcDZmqqKxJlUNFYT9oMsUF3ch/hEL+4n0gyYHamvSKR6/jhHEkGzAhXJDpEj8ofJBkQ v2l/2lwZRVwsFpWzcYFe0X682JVjBf3u9voP9It7iCQDYqbncqb9abN0BNM0reeoXGkySamGZt+b sf7dI/7i4+PaRSBOckWiMWagKxKbO/LRoQ/QHbno4/gKD/2L+4Y+GRAV+2oOfw75DzXrcQ+vHfl2 cX/x8ZFkQCQ8GUicDycX4sMJkgwIHlckYuZIMiBUfg4kzgqnuzxBkgHh0UEtAmxajOJ6gmsXgWDI TX7tKxKnLhHgBfpkQAAYSPTWarVii0yOJAO8xtUcwINIMsBH095fCg1xIb4nSDLAL9PeqB4NvXjx QjbTl19+eXZ2lqbpy5cvpy7UfJFkgC8YSAzI48ePf/rTn8r073//+7u7u2nLM3MkGTAxBhJDlKbp 6ekpAeYJkgyYDAOJQTs5OdEkY/NNiyQDJsBAYgRev359dnYmYUaSTYskA8bDQCIwBJIMGAMDiVF6 +fIlV+H7gCQDhsVAYsS47sMTJBkwCAYSZ8K+7gNTIcmmlOd5Zc52u62/LEmSJi+rvzJJkuVy2aOA Hvn444+nLsLXvvGNb7x69cqe8/Tp088//1wSy/+BxN1up1WoUpcqVejI/2ZZNlgBXRp69O/k5OR7 3/veyGOM9O8rFhwtjkATS9qFejK5bRTsdmrv54aYcB9//PHNzc3UpfjKj370IzvJ9AK2b37zm19+ +aWHA4lSA11VA/sIbLvdygK9rVTx3eHXh2/kQxls9Mncs3Orcgw7zmHscrmstCn6uRJyKuhs84E8 ZkX//PLLL/05NNT00qRxVf3s5ch0nufb7Xa9XnueaogVSeZGvdVYLpceDr/YIafFoxnqRkYU6zNf v349SXmE3SMf7eBJPsVONSmDh7sAokSSdaf9G+fHvCOrN0OaaoF+oxHoiKI4PT09OTk5OTmZcMhF tl2SJNMei9hHSHKEx7ERhkaStWYHWLjpdUgl1XSkNLKv2dO//Mu//Mmf/MmTJ0+mjS6lgeHVZqp0 +skzDIcka0Gbdd+ajIFwcH3IX//1X1euXZxKENvF7u57XlQEiiRrZOZdE/nWdmeUxmhyej4slDqZ ZZmWmfoDt0iyB8iB5LNnz9j39GoRDq4nJ5EQ3CaQKiT1J5QARhD+aOoC+Gu320lX7Pr6Oqz2YmhZ liVJst1u67/sxggkxrIsC7RaSv2h8sAh+mT7aWMxdUE8JQfXEvaspTHFUTO1cx/6F4En6JPtEUdj MYLlcsnB9cjW63UcNVPCbLfbTV0QxIAkqxo/xvQ+syEizMaU5/mzZ8+m+nTnFTXLskN3EI1J0Dt4 KEiyqp4xtlqtFotF27d0/jgnBeiJg+vRbLfbVufGFpZW1aze+Fbuy9XwXU3EfRjUZL2hP5LMsQl/ JysB1rwADgNPLgBxtTTstdvt6veeflBZlmVZFkXRPGZWq1X9Fv4Nb+rfNsw6fKOw+PkwhPiQZO9x eHi42WwWi4UckdnT5r7bVD9Ss+fbXav6u2SB9hx5sR1O9dfIHJ0pr6+UrZvlculbkt3e3p6fn19e Xsq/xhiZPj8/r7/g9vZ20sI20nMNa5MqW18qgF1JNIR0ol5R7Vfq9Gq1ktekado2yTysOebAanlw vakj662+0uAET3V5jyRZz5Nki8VXa1Un7GkJNplZFEWapvJfh+YbY+SJw/JvfeHNJxaLhSz50KL6 +Oyzzz766KOeCzniV7/61d/+7d+2esv5+fnbt29l4uLiQh4Kc3t7e3l5qfPfvn1rz2mo8lSXcXSo nHpkow+a0YdxSGMqtaIsS/3T3NeHQxXSfqU8ic1+DFuHunR1dTVoz+yXv/xlh5GSypdtst7Eg+ut vtLa8uGJKj6UwcZV+O/JsmyEUftDu3plflEUMs4jNb7/eWPZIQeqgh999NGgl8n8wz/8Q5+3a6/r 4uKi8l/1OX7qNoRbqVRaiyrN6N5WdW9FtV+ZpuneocjmZMh00Jrzy1/+svN79as5XG/9VxrqGF2c gI4qVMKpMt8eq9HjuJ6fWxSFBGSf5dT5f9L+4uJCxhjD5WQgTprRhi8+VFHt9r1ndfJwaHEvh+ut /0pDHUlW1fOacqm7i8Xi6dOn5r5O63kFY4yce5dhdDurKvNlZmUIQv7LHr6oZ2GlAPZrdHTePmVi l60buWrRtx85aSdMJmRoUc6T2QOJ9Tk+S5Kk+TWiWhPsSiLdcTnfYw9923/Ka/ZWSHmljrPp0ZV+ SoeDLd9qjrJXS5P1Jo6vt70rDf1xnmwPfhnd3Gjr6uOPP5Y0cqhzhk1ynkxcXV1dX187XGC9Le6c SW2HDfI8H+HWkd2G0x884efk7HI3Ppyj8qEMNvpke/Br34aI/PE9e/Zs6JrZbRC7bfdCOpeB3joS vuGKj/1kB5Orqmip93JyneeE9HL8sLi9Y+Gh3xR2CLNWb/H8GOjBn1qOfPMBPIg+2UHL5VKGcfI8 5wYWNnnwZugZ//btW7kEf+qCtOZwzEB+Nz3y2Rr/n+ry4GqZZL3hCPpkD+Bxt7YgnlDcSiiX4Fdo z8zJthjzivBxzo058eBq4Up6f5BkjejjbtfrdZIkoeyKrtiH/z4fSs+K/dTKICpkfIdB8AdJ1pQ0 HJVIi7tZ12fVGwLMV0GMGZBhGBpJ1ppGmpwukhbExNLW2+llaH0CoQdYsu08qYpal6hFGBpJ1p20 F3pQbKdaWLtu5dqBsAoPIQdY5r4qysxJIk0zNfpBC/iDJHOjkmrbe3prVH/iQZs5PV42PhUPPWl4 jDlmUO/Kk2EYE0nmnu7DMqGHqDr4U7nz9xD7fKVlMe/nlnwobU3cjowZ2C/owO7EBzoOgchwt6rJ HAob2/GnXey9++oIMTmJjz/+uM/b/+M//sMY85d/+ZdOCvONb3xjqrtVOVGJIp3eW98OvSCU6Br5 Xr3v3r07OTkZ+lMmv1OUb3erIskwC3r75qkLgsj51sQPxLevyT0+AMAl7v0xPpIMc+HVISQiRpKN jyQDAGe4hdUkSDLMAofJGAdJNgmSDAAQNpIMs8CRMkbDAMD4SDIAcInDpvGRZIgfx8gYGVVuZCQZ 4kezgpFR5UZGkgGAS4wujo8kwyzQuGA0VLbxkWSYBRoXjInRxZGRZADgGEdOIyPJED8OkDE+at2Y SDIAcI8kGxNJhvgx1IORUeVGRpIBgGMk2chIMkROnhYNjIzRxTGRZADgHt2yMZFkiB9tChA3kgzx I8kwCQYYR0OSAcAgSLLRkGSIHK0JEL0Ppi4AMJSzs7O7uztjzGKxOD09/eKLL6YuEeZisVjoRJqm RVFMW57o0SfDLLx8+XLqImBGTk9PdZoYGwFJhmg9fvxYp7noA2Oy6x5GQJIhfvYBMjCCoiiodWMi yRAt7YdxgAzEjSRDtDTJOFGB8Z2cnExdhBkhyRA5BnkwidevXxuq31hIMkSOQ2MgevyeDJF7/vz5 1EXATD158oQDqXEsyrKcugxeyPO8Mme73RpjkiTpv/C9i0qSZLlc9l/4hHhgSh8enr2z94IO9b/y Fv9reJQVeJx6tVqtvKrAc0yy3W4nu5wxZrvdVvbVLMuGLkAlNe0y+L/z23yrzQHxZNXpvmBXwv67 wN5dbIQ9qy1PtoJDo30j31bdLEYXNTnqu+ske1flQ+VP2fmVlDOsYEMoKsdSWZa53RGWy6XWW1ly nud5nkvFplbDuWiTTPbVSiR4eFSo6ju/ZpvMZP9HT3aAjbwv6IGj1Or1ek2kwaGokkybfk0vn6Pr QXa2VVIt6O+F8WmGTV5zpFbbkfbs2TPyDD1FkmS6o4aeXofYqSYDNTId5ZeFQ1JVPOz9aKTpqKNv JURAwk4y+8TyfNp0/aZ5nl9dXTFKg70kIfzv8eiJNPIMnYWaZJph8wmwOjlRrwOPc14VqJCu2PX1 9dQFaUrzzBhDmKGtIJPM2wGTScgozW63k9VCniHP80D3Dh1spBqjlcDuViXttZwMC3FHHY5emXl1 dbXb7aYuDqahO0i4e0eWZUmS1O9UABwRUpLJMBoZdkSWZdfX19vtljCbpzhONUn5qcNoLpgk0xib uiAByLKMMJuh0HtjNuowWgkjyYixtmgI5ka2dRwxJpIk0SuTgePCSLKeMbbZbDabTcMXLxaLVkt2 u8BuH7EXDUFFzw3huQ7bemFpdS/dbnWy7btiSmUMLYwk66lVkjW/pfJqtdKnEjtZ4CHdGo7lckmS 2aK/WXaHo72yLMuyLIqieR1rWO336lCTGVdAEwFchd/zKqbNZlMUxWq12mw2sgduNhs5Ak3TNE1T nZb9ebVaSZMnb5GFfPrppz/96U/TNJWFyE2g7d1Sl39kgfUXa8GMMWVZyrQ2K/IpssBubcd3v/vd Qa8Be/TokcOlLRYLXcPGGHtV19eSbiP7htz2irW38qENUX/X3s+qbBE/9Ry30Apmf31ZabJRiqLQ 7WK/srLe7FrtpCa/evVquAOyb33rWx3etXe1PLje1KH1Zu4bK/8rm49K78nN2Tq/vSgK+TdNU5lj jNGZ9rT+b1mWkkn6Gp2/d6JSU/cuUNkvfvLkiRbSLqe+d+8Smuuz3prQVerEoTWsn1LfFvaKslfs kS1b/wr2u+zPOrJF+nO76spO21q/uBbm0KrWP0tro+yt8Pam6V+T1+v1oHW481aorJYm600/8ch6 61/ZnNeryT+ooQCSrCzLFy9edHtj5bhG50t9khpjT5e1Kqgt2vEksz907wLt/9XpvTu2/d4jL2si jiSzv/6RHb5emOMbYu+7Gm6R/oZIsu122+ot9S9bX9UPHl4o7WTYS+ifZK1e31bPJCsPV9H6ejv0 iZX11rOyzTbJwjhPliRJt+FyGYYSOu4nQ1IyBmJP22/UcULt+B+nb3/69OneBe59sT0tQwo6Ftr+ u1ZF89tS3XD2nL1nXOyVeWjLHn9X/U+HW2Ro/Ufh6qv6CHtFCXtPcbLeQjnR62q9BVTZvDNhirbS 4ZCzfkhurCH7+r/l+10xc59n0iCa9/tq5ftDW7LwBxdYeXFlWke3tKj6Rdqurp5Dsg05PC7TlaZr 3l5juqsf/+jKin1wQ9TfVTbbIk6+r6tFqVZbXIcr6l1be1XXG4rKDmWv3vSek5o89NBi2a9PVq+Q x9db5b319fbkyZP+lW22fbJgkqzsFGadaTWqN3z2a1rVtm7HDR0q9DgxVvpXmxvy4QBuiFU3xHav r6u21d5+Y/MXe34o9mAVmqqOzTbJwhhdFGP+2te+cPHQ6GLbq7DK9leBd7hqmRuwPqjDhgiC/ABr 6FHlttVetK3J1GG0EsBV+LbRnstXlqX8kFaHufbqtlc312r5u91OHipPEzBbsoPkee6kDhz6LXmH at/8LXJDH3nsu58e/Il93L/B91NIfTIht8rebrdDH3tKpzWIX3XIHdC32+319TUxNnMO7yUvu0Dn u8x0EMRdwh9cLeOvNwTWJxOVJ3JFc9fUDrTBIsCgZHeQ54n3rxhDDzyosB6w9+BqGW29wQSaZELz TE4OJ0kyn0jTR2abcPZ8jGm5XF5fX8tIo//7BYek6CngJBOSZ1mWSeMu7Xusu4R+x1nFNjqTo5xx Ti13Q4bBieCTTEmkmfv9VnppJvydxO5+GWO4oANtaZ75M3Rhn8ajPqO/eJJMyY5hH41qqpkQdptK dJnwwxg+yLJMhy6mijQCDAOJMMlsdqrJPiz7kn2Z77R7lJ1bdqlILwyhPhqvtW6IHUGr96CfAkSe ZDYdfjQHsk3+q/JDlv6JUu9j6cm8yqewk2M0ujtoqskAhnm/Zraqk3aXq3KwSN3GoBax3u+gp3r8 HFI5iXXklfF1s7y91em7d++MMScnJ1MX5CC51d7UpTjmyC/SDh32iYBCq9UzeHt68+bN48ePR/ig cX4CW3k04ORIMkRInnPoeVRgVnxr+nvy7euEd48PoAliDJgPkgwAxuDtaHwESDJEiFveAbNCkgEA wkaSIUKcJINvvLo+Ij4kGQAgbCQZACBsJBkAIGwkGQAgbCQZYsMl+PATNXM4JBliw4WL8BM1czgk GWLDkS8wNyQZACBsJBliQ58MfqJmDockAwCEjSRDhDi1Dg9RLYdDkgEAwkaSIUIc/AKzQpIBAMJG kiE2XCEGb/HY6IGQZACAsJFkiA0nyeAnauZwSDIAGANJNhySDPFYLBaLxWKz2ci/UxcH+NrZ2ZlW Ts6WOUeSIR6np6c6zfEvvPL48eOpixAzkgwRsiMN8EFRFHun4QRJhghx/AsPcYA1HJIM8ZAAOz09 5ZgXHjo5OZm6CNEiyQBgDM+fP5+6CNEiyRAPucqDI1/4iauQhvPB1AUAnJGW4vXr11MXBNiPU2UD IclGkud5zyVkWeakJL7ZbDb9V47tk08+6fP2NE0jW9W73W673fZZQpIky+XSVXm8kuf5mD89/Iu/ +It//dd/7VlFO/jggw9+8YtfjPyhY1qUZTl1GWJQbyzkzyRJjrxrbwPxYLO+d8nhNr6r1erv/u7v pi7F1370ox+9evVq6lK0U6l+Ou28+h1acrjV75NPPvnJT34ydSkG57xWr1Yrr66rok/Whd1wbLdb e6/WXbrzvv3gGysvkKZH/rULE/Fx9Mxp2FSOaXSL98mV4++t/K/uCHmeVwoTbrYhRCRZU4eiYvI9 1i6ATEv7oqTAk5cTnWlg2NUvy7LJt+lyudSjJS2MHlppaTmowtBIsmMq6eVD29FEvX3J89xuWYL4 FjNXGTD05LCpicqwhB5XyUxSDUMgyap0xwsrvY6Tr1BPNZoV39gBFsfWsY+rKqkWwZ4FT5BkX9PT D0mSRLyPaarZzUrE3zcI9vFTxNvCTjU5ojKxBDamRZK9dxQccSNSp80KbcqE7P7xrKqfPfa4Xq/n 9vXh1tyTjBbc1M5n0KCMQw+hrq+vpy7LZORwKsuyPM+vrq7IM3Qz3yQjwyqkTdntdrJmaFCGoxnG SlZyQlqGB9gr0dYck2y3263X62fPnrG31Nl5Rjs7BIkxGuu99KIkxgbQyuzuICztyPX1Ne3IEcvl MkmSPM93u93UZYmKVL8sy6h+R2RZJtVv6oIgGPPqk2k7MnVBAiCdM2lNaHadYNi2Ob0WidWFJubV J1uv12PuGGPemXQgWZat1+upSxED6d1S/ZrTMJu6IM7c3t5OXYRozSjJdrvd8RuqHrdarRaLRdu3 dP44JwVwIkkSxhj76zkYMM/qF1OH7Pb29vLycupSRGtGSVa51W9bE974WVqQVgVw2OgkSdLzmSDo L6DqN8nxlv8uLi6mLkLMZnSezOEZss1mIw812Gw2aZrqtPxXmqaV3X61Wul8mZaH6eiTEewXyMLT NJU5stjF4r3n71ReI2UwxsgSpClZLBbycbK7l++PAAAgAElEQVTMzk3hcrkcdIDxW9/6VqvXy7Ht xcXF7e3txcXFzc2NMeby8lKGbt6+fWu/4ObmxocWxO0QmVa/n/3sZ59//rluX7sW2a8/Xv2O1D27 LjWsflpdndQ98eMf//g73/lOnyUc8eGHH3Z+b6WmGWMuLy9vbm6k76VV0RhTqaj6p6jMtP+U6Zub G6ne9rtgm9Hzyfr/TsXen+vT0iLInKIo0jQ9Pt/cx55O6PzKRH1+5TW6ZF3g3rd0M/RlMh2eT3Z+ fv727VudkObD3Lcsb9++lfn6Z6uFD/R8sv4XLxza+hpL3arf8bq3d8IcrX4O654Z/qKPns8nq9S0 Ss08Pz+XYylJpsr/yhL0IMwYs/fFt7e3MqfPMRnPJ4uHjJINehneof22Ml8OV+VgVub0PDkvDdZA davnqOwItCGo7Oo+9MbG1K369b8wZNDqF4RDNU0Os/RI65DKwVblwpCLiwvp+fUvZ8RmdJ5suVwO fb5Hz7FXWofKfB2E0Znal+r8uUVRSAvVeSGHDB3//cmuPnUpHjb0ZXgdql//umeGrH6hX7goGabj jUdeJhO39+zcktHFIGr4hGaUZMaYZ8+edd439ASADtfIfqvTckpAXmA3FpX5MtPukAn5X3ssyG6P dL5dAH2NnLewl2+/pU/7kuf5s2fPOr99CHrEao/JGGPOz8/tQZvKn5PTm1t2e/ve6mdv3z7V73jd M+2rn5O6N/7vFrqxa5qMCtpVVPpkMmaoM+23yzmw8/Nzu+Ml0zLwqMFmLxkVMzpPJvhxdHPj3AGv w3myB/XJsIHOkwl+HN3QaPupk/NkDsszEM6TxUZ/bsmN747g3oADkcfCceuK48h7tDW7JDPWfZho rOsiuE37+fn51EU4xr51RbgreSBhPaHC85o2K3NMMmE/NJkGxUSRYUKvXfb2ci994ABP5FLyhIqw 1ob/NW0+5ptkpvZELhN+I96Nfv0/+7M/++d//meZDuW4+BD/G5flcnl9fa1PmAx9hXcmoyPGGPmx x263C2s9+F/T5mDWSSYkz4wxc2tTtBNmrAj/wQ9+YO4bFzlGNuGnms/kCZM6PGBmczilx092P4yK h25Isq/ZbYruS/E1K9KCyDnCQy2FfGu9cFwbWUPjMgz7cMpu4iNb1fbBk9m3c1Hx0A1JViVtikaa XhhiQk41O73M+0fBD9JG1tC4DC++3ommV9v9aG/Fi2BnxBBm93uybup7o/F7X9Ljeju93DaFnVuo iiF+T9bHoL8n66b+c36fg83ueA0RPK4qnuj5e7JQRP97MpKsi737qpmoffGhMJXGpdVHk2Rt2Ycp 5v5CCTNR9bNTdpLjvMp4Q9vPJcm6IcniVDkBYI+/HXrLkXansrSKvQv35yC9UvgHC0aS9VfptFUS bq8jLf7xO7rtXbgn4xP1UfQHdwqSrBuSbI6OJ1ND/mRVc01SbbPZ9L9R7G9/+9vvf//7PRci0jT1 pF12xcl9eINbJ5VvfWj3yfO8/9MAWvntb39rjHFVXRv64IMPfvGLXzhcIEmGmeo5CnSck0dhIVZt xwmG5lsMdODbV2D/xwSa/BKgFZIMDfmQar7FQAe+fQX2f0xJr67u2VHzbb9CEKZKtQgOvHzb44Jf oYiGfWKjbaT5tl8hOGOmGknmHL+Mhi/sHwXP6rZh8MGYdwDo+ZBu1JFk8E7bWxH2eTAxUGenWuVm K8FdwzkTwXdyET29YdihRmSz2Ww2G6/GOhAlV9ff+jY014FvX4EkQxgkz549e1Yf55E+mVf7FeL2 4K2Qj/MtBjrw7SswuogwyJCj5Fm94eDEA8ZUGX7UYQPO7E6FJENI5PxZnuc0GfCEPonGfiAU9XNk JBkCI4fDcsaCxgL+qDwQalbPTZ3cH01dAKCLLMu22+1ut5M/GV2EPyTPJMPsR6faRr7ZY/TokyFU ctps6lIAB+nDr6Wi2v0zjr3cok+GgEnPbOpSAMdIF80Yk+e5jiLALZIMYUuS5N///d+nLgXwgCzL kiSxh8ThEEmGsMlFHwwzwn/SOdtut//93/89dVliQ5IheN/5zneMMRzqIghZlv3Xf/0X1dUtkgwx 4IQZAvKd73xnvV5PXYqokGSIB2OMCMWzZ8+org6RZIgEP0FFQOT8LmOMrpBkADABhsQdIskAYDKM MTpBkiEqtAsICEPirpBkCNhmszk7O/vHf/zHs7Ozp0+fTl0c4GFaY6m0DpFkCFiapnd3d7///e/v 7u7evXsnT/IFfKY1Viqt4boPF0gyROLk5ISHvCAsjx8/NsZw3Ud/JBnCdnp6KhOvX7+etiRAE1pj jTFFUUxYkpiQZAjb3d3d1EUAWjg5OZEJiTQu+nCCJEMMnjx5MnURgEaeP38uExpp6I8kQ9jkwFZb B8Bz+oxNKq1DJBliwBN4ERwqrUMfTF0AoC/7FDrgP2qsc4uyLKcuA+bohz/84R/+8If+y/nd735n jPn2t78tf3744Yf/+7//22E5r1696l8YhM5VtTzu3/7t3/78z/+8f6VtaIi6vVqtvLrwkj4ZpvGH P/zhJz/5ydSl+Mrl5eXURYAXvKqWTsykbnOeDAAQNpIMABA2kgwAEDaSDAAQNpIMABA2kgwAEDaS DAAQNpIMABA2kgwAEDaSDAAQNpIMABA2kgwAEDaSDAAQNpIMABA2kgxhuL29PT8/v7y8lH+NMTJ9 fn5ef8Ht7e2khcXs1Ktf8/pp1+rKn/r6y8vLmTyfpRueT4YwXFxcGGNubm6MMbKfy7Q0EG/fvtUX yG7/9u3baQuMWalUv4uLi4b18/LyUoLt9vb29vb24uJC3isRKK/X+ZN9Pe+RZAiSHtXWd292eExI ql+r+lk56qr02C4uLiQaHRc0LowuIkiye09dCmC/VvVTX3l7z86t29vbm5sbavtxJBnCoAeqMqHD jDJ0oy+rzwFGo9Wvef2U8UMZMNcAk2kZeNRg4wTwEYwuIgwXFxey/+uENBYVZBgmVImo4y/YO1Or tz3n0Buh6JMBAMJGkiESerkzML4Hqx/1c1AkGSLx9u3bt2/fciIBk3iw+lE/B0WSISpcrIwJPVj9 qJ8DIckAAGEjyQAAYSPJAABhI8kAAGEjyQAAYeMeH5gMt5KDh6iWISLJMI1Xr171X8hqtTLGFEXR f1GAcVQt21osFmVZjv+5MWF0EQCmlKbp1EUIHkmGsNEKIAKbzWbqIoSNJEPYSDJEgCTriSQDgClx orc/kgwB40gWcaAm90SSAcDEGCTviSRDwNj/EQ35SQm6IckAYGIck/VEkgHAxEiynkgyAJgeF330 QZIBwPTolvVBkgGAF7joozOSDACmR5+sD5IMAKaXpimnyjojyQDAC3TLOiPJAMAXnCrrhiQDAC/Q J+uMJAMAL3CqrDOSDAB8QZh1Q5IBgC9Ism5IMgDwBafKuiHJAABhI8kAwC9ci98WSYYgPX369Ozs 7J/+6Z/Ozs7Ozs6mLg7gTJqm2+327Ozs6dOnU5clGB9MXQCgi3fv3t3d3Rljvvzyy6nLAjhzdnYm FRut0CdDkE5OTnT69PR0wpIADr18+XLqIgSJJEOQ7CR7/PjxhCUBHErTlCOzDkgyBKkoCpk4PT3V aSACdMs6IMkAwCP8pKwDkgyhkkEYe5gRiMOTJ0+mLkJgSDKESjLs9evXUxcEcOz169ecLWuFJEOo 6I0hYlTvVrz+PVme5/WZ2+22PjNJkobL3Pv2vUtIkmS5XDZcbKBCv5XA9773vXC/QkAXqux2u8qO 0203rL8r+v2uc/08OTn58ssvnVfvgGpdK4uyLCf8eHsP0YlpK/eR+LQLlmXZOOUZ1Gq1irVme87P NV/fHys743B74qGw1AIEmnBebWiHhfHqe5kx+2SVncTeQ6SOepINe4thz5Sok38rO5snXwFo4tAu Ocn+uFwuK0GlBZByqtCzDUMYMMm0c2NXPqmdQbf4duErO1ue5/UvC/ij3vXxf5e0Q07LKfvaer2W 3Y1UmzmXSaY7SaU193kncUV3NvmydrCZ+07bHNYDPKR9mpgafTuA9QvKf8XxBdGKgyTTcTZtr2my DwWb/C/rByOwAyxJkohrnd1pk28tfTUibT66J5l9ZUTc+0l/9p6W57kOQrKnwbn5BNhesq9lWWZ3 1Oa2EmaodZLZ4+zUjw50VMQ+eGRNwgk5vqRGGevwUY4dDe1V1FokmXbC6Ek4oQeP7GnoTzOMfbNC R/jZyyLWKMl2ux1dh+Gwp6EPMqwJOXDc7XZXV1fPnj1jXUXm4SST/eT6+nr4wsya7ml5ntMqoSFq SyvL5fL6+lraNFZaTI7dd1FbVXoJo5HxRvtCR+AQYqwbdrH4HEwyuR5hhvvJZrOxp+0/m7+xJzl0 YE9zyIfN6lZYMebbasyyLEkSdrFoHEyy7XabZVn//aR5q7FYLFottmuJvvosUV+s3rJTJvSpd/W3 2NP2G52UUE+e9VnI3IywWT0hFaPn7jn0vrlarXT1OlyTlQ3UeTmhHASgif1J5rABbb63NL+X8Wq1 6vNYVbn3ZVmWRVFU9i67gSuKQv888pbKG209wyxJkkN37kfdaJvVB5U7l3Yz6L4pSy7v2avdrZ57 WRDdslZHEvO0/174rgYuNptNmqZSubV+60Gx/Jc9LYVZrVZaO+V2y7IEWVpRFIvF18WWF8v8yp8y XRSFLE3v3Gy/XaftD5XPkmn5371vsWfKn3YhjYvbRX/22WcfffRRnyUc95vf/ObnP/+5q6XZa8Dc b317uxhjyrKUaV3tlVVU34LHt6b96eNsVid61o2rq6ueF2ENvW9WVqOyt6l8okzIp1Sm5fWV5VSW 3HNNDr2L/fKXv+xQPHtb2JXW3o86FCbie+F/fdBkW6/X2+1273+1oge8aZrKHGOMzrSn5b/KspQ9 R1+g77In9E/7KK++KP1omansb60fas+x3175xHphjsw/tHqbW6/XPZdwnP3F+zu0BvRT6uuzsmkq G7TJ1qx/ev1zS6eb1Ymea/7Fixc9CzD0vnlodVW26d5NUB6uIfUl99wufu5ih9a//K/uGuMUZuhF ObG/Esi9J3ouupLYOl9aK9kS9rS8RncSe2sdSbLKh+qB89ff8H759pzKdGXOgy+oL2fv/J77mJOt cNw4SVZfe4cyqVKeJluz/umHPtfVZnWi55rveaw5wr55aEtVtumhJDty1OIwydbrtbe7WH3996+B s0uy0sWhiv1V9WjOPqarHN/Z+aSv/6qUh5OscpxSP+SvNIXl+4eWMqGF0Wn7z71vKRs0ef2PFp30 jI8YrU9WX2l7Wzr7lZ9++mmTrVl/79Cb1Ymea77nUc4I++beLVVf4KEkKw9nodsk6/zehrpt6L3r 367P9MkqDlYC2VU6t6T2AUV5f0SsO0n9X/twTxsde6yjMvH1FzCmsuPpzma3aMYYbRm1eJW9Tstj n53Wz6q8pfKCeiHLfht7hKPF0nXNtjdl5dC7svaOfLRuiONb89Axygib1dXq6rmEzrvnaPum3fOT t9S3aX3ntd9e+ZSytoHKHmvS513MXv+llWF2czdaYYZelBMP7JwjdAsqtIIeaTs6DxO7bYyO61ZC Mdpq9606tjXmBnXLyZofefccdN888nHlQ2fdOpjnLhZxkh27x4eZ4sfwlYvN9r7m0PwHlfcHNUPr fGUwd6tqZbQN6qeRd89B983KB1V+FLF3Q3fey9jF4vPwfRfl+SOj3aW0LEv58UR69KJnJzvMcDoU j+dwogPZPce5K+5o++aRhff8IHk0IHcQjk+je+HrzW31CX6D1oO5HWjrgzfJMHSgd8Vl3zxC1w83 Q49Si+eTVfLM0Hvox35mKXsXetJ7dbJvVuhoB3tZxFo/M7ryJFZ9yDq99YbssxqsN7ilecZgtb2j zXYlzEfrJFNSOWSYXntphkqzjz0wa1hFGJhWMI20+RwzEWDz1D3JlPbSzPsdNZkzz8qkI4e6KjgN hvHpExW296I8lpL0ivXboQkHSWbTjpq5338qwRbrsaEeCVZSnJ0Kk9NjTR1BsffKEHfJ+pEi+9rM OU4ym73/mPcPDKUW2o+lCKUW2pdp7P0WoXwRzFNlr5RIs7trwqtqbB8mGmuPY3eDGjDJKuxBSLv+ STXV8QGZWX/20jhHjnt/ZFpPLCkMexFCp3XYPrVm3t8fKzvjEHuifYAoOExEK/ufIeSPI9FyyN4n EB55y97Xhzjk0oGfD0Tu7M2bN48fP566FE01/P2vJw6FzV6Vfar5K+Pb75zsYu/evTs5Oem/HOOu 1vn2fDLfkwxozre9C3Di0GNLJ+TbvvbAfReBsETWywTQBEmGeHh1kAi44vltZn1AkiE2nW+RDiBQ JBmikqYpSQbMDUmGqJBkiAynfpsgyRAbwgyYG5IMsSHJEBmu+HgQSYbYkGTA3JBkiBBhBswKSYYI FUVBkiEajC4+iCQDAISNJEO0uHwZmInxnuoCjKkoisViMXUpAIyBPhmixXUfwEyQZIgWSQbMBEmG aJFkwEyQZIgZYQbMAUmGmJFkwByQZIgZSQbMAUmGyBFmQPRIMkSOJAOiR5IhciQZED2SDPHjBqxA 3EgyzMK3vvWtxWLBnRgRkNVqtVgsNpvNYrFYLBZnZ2dTl8hfJBki9+LFizdv3nz55ZfGmDdv3kxd HKCpoijsP09OTqYqif9IMkTu888/v7u7k2mdAILz+vXrqYvgL5IMkXv9+vXp6alM6wQQBGpsQyQZ 4vfy5UtpEe7u7riOEQHREcUnT55MWxLPkWSIX5qmevkiSYaAPH/+vDKBvRZlWU5dBmAMT58+/fzz z09PT7/44oupywI0JQ+M9a2hXq1WlQtSpkWfDHMhJ8y46APB4WzZgz6YugDztdvtttvtkRdkWTZa Yfwx6E++Tk5O7u7upv1VmT3UOTd5nu+dH19V32w2rsaxHz16ZNztF7FWP5LMpfqOqlmVJEn/pR1Z YJIky+Wy7Ud46H/+539ubm6mLsWAfvSjH0XZlFSqq1TUhtW+yXvDSrvNZuPV4JtarVZRVj+SrCN7 39tut4f2Oue7X2WB2rHbbrc6oYWJJt7gId0F9la5PjXffq/WcPk4/SzqNmwkWSN2YNihJbvchEeL y+Wysj9rYfI8394z90e4YR3YwiuyF9Rza9BKVanh8ll23U6ShFQDSXZQfXAvy7KAkqDeL8zzXJoA jmrRnN0ZSpLEh13ArtuSr+v1mmO1OSPJ3lM56oysrbd7kHpUa/8XoOwrkjwJsL2k06a1WnI3sj0X DyLJjHk/wHzeaR2yv6Pu/4ZIgzUaEVwe2MMP0kujPs/E3JNMR9vmXOkrw4+VmZgP3R1C3/pyIiDP 86urqwi+Dh403yTTUQhquapEWnCH5OhMxxKvr6+nLoszkme73e7q6urZs2dU5ojNMcmiOfAcjn0u jTyLXtznlpbL5fX1tVRmdvlYzSvJdrudjJ7HdOA5HDvPaAJiJTEW/faVzlme59F/03maUZLJ+AkZ 1hZNQMRmNYYsX5OaHKW53EFYYowa3M1yuUySJM/z3W43dVngjPTGZhJjQmvy1AWBY7NIskliLLLn YMlPdo7f8hgBkYOSEXaK4zvCYrGQp5aMtr9oz2ycj3OuvqJ0Hc7ZLJLMSYytVqtW1WWIG677UF/9 bALOz89HW/75vUE/cWj1m4U2tLA8WMk3m039NdoWr1arsizlTrvd9pdu+Rfu2Ex9fdrrcM5mkWRO TFVXfEgv5W0T8Pbt2zE/a8yPG8h2u+08rliWpbSeDwZJ/bbr9r3Y5e09b83eLcwCHV2orysn6zAC 8V/x4bwPIYdFshunaarTRVGsViuZaceePVOm5fGv+tBV+wWy8DRNZY5U08Wi+mhvLYMcoJVluVgs 9F3ygvoCi6JwUuMH7Zb96Z/+adu33N7eXl5eSrqcn59fXFzc3t5eXFzIf11cXNzc3Mhr5L9ubm7k f/W9+pbmD5TRJQSqW4esQupVpSpqv0F3hEoNl/fKa6TeHtpf9E/5r08//fSnP/2pvWTZATvU6iRJ Pvvss48++qjfCjhIHirWRGUPNfctQ5P1KUuor8M5KmMndxd1sihdXfZ602l7d5JaVZ9pv17+1In6 ws39I8/rn3XkZYcmtEg9vXjxov9Cjri4uHjbnrG6SscnJKuOv/fI8oUspFtRf/CDHwy6AhvqvB21 PqdpqnPsCa3qeyv2g/XZfpc8FrK0Rs/sJdcX0pyrNuEQXTlNVPbQzuvTecFGWI4r8Y8uLpfL0UYS dLXaAVaZaXfd5AUOHy97hJPe2G63c3IsPyEnfSnJsNvb2/6Lmkqf7Sj1+VAnoCzLzWbTeVTc3h10 WMLct+Z9lmzzcHTx0B7q8FtHLP4kM8YkSTLO5eN6MtZOpspMe+c09/ttKMPcnS8TiFLoA4wDjRLL 4JiOj7VS2R2kT6ZHfn2WbAvrgMzVt47bLJJMLh/vGWZ6yurp06fmPp9kpkzLOYDKBV32zMr+aS9c /tc+JVDJQrly0j4KTtN0sVjYBdC328sRcqVZ+f7Jtrb8//nR5eWl/Cu9JZ1Qcs3hW+t6DX3BkQ6W LPb8/Pz29lbOq52fn19eXjY/qeYhuXin7U6he4FdweyqaO5P/Oi5H/uNxrpyShdlaldt2LuDvQR7 yfant/oKxtHFzG7Ze2jz9Wmvw5nr27oFZFa3M7D1zzAz4tr7+OOPOyREJZ86v2YEP/rRj169ejV1 Kb4y/g0vWl1rZ1/BJJebH3pBc+PUZL2eqwkne2hDrQo2wnJcmUWfTEjPzM+fQ3nO54MAtyerzt8n vbGIjX/Di1bBY48uHLoAvRWfazL6iP8qfNsMH1nUc+TB/7uky3Dfg924hj9k9qHTNia94cWYm7h5 mJVlKRV471XmrUJRbvTjYU1mbNCJeSWZmd8j+GTUotsgjPE7w0TDoUiJKP2dGdRyuVwul94+8cDJ sJvP9/vvvIfCNrskE5U887+97qn5TqJPXIxynRBjh+gTKSM7vAvlgIwY62mmSSb0kbLy6+k5RNoh shJk+GW2K2Hm9ImUobT+x/FA3VmZdZIJGV3RSIu4R1InbZYGGPs89Op83RfCqhX2oEJYJUcfJNnX JNLM/dGcfReAmHYJO70MOzz2kX1BUkEvbvS5nlSOyeZwGAobSbaHvcfqzmzf3sLnXbpOWyL9CjKs OmmhEAA9tjPG6Kij8KH+VH4/QK2eM5LsAbozy04iO0+lW2O8GY3UoRWzr3js5+hMK8+Ex3b1YzJP 9jtMjiRrx95jK2cUJEJ0jL7yRie7nB1UQv+0P1EKRm5hCJVjO3sEsl75u1V7u7NVWSbHZNhrRner mkQ9e47YG0t7RXwo+sknn0xdhKrf/va33//+910t7YMPPvjFL37hamneevDWIQ/W9qArubd3+63f 9LUb3+5WRZIBD9A7RE9dEMAXviXZjO67CHQje+wIz5AD0A1JBjys8lQ5AF4hyYCHyakFb09+ADNH kgGNMMYIeIskA5qSJ4BPXQoAVSQZ0EKapowxAr4hyYAW5IQZPTPAKyQZ0A5jjIBvSDKgNcYYAa+Q ZEBrjDECXiHJgC5kjJEwA3xAkgEdceMPwBMkGdCR3FacE2bA5EgyoDtOmAE+4EmbQC9FUSwWPB0J mBJ9MqCvoigYYwQmRJIBfTHGCEyLJAMc4KJ8YEIkGeAGF+UDUyHJADe4KB+YCkkGOMMJM2ASXIUP uMRF+cD46JMBjtkX5dM/A0ZAkgGOyRjjixcvzs7OXrx4MXVxgPiRZIB7aZr+/d///d3d3d3dHdeA AEPjPBng2GazsdPrzZs3ExYGmAP6ZIBjaZo+efJE/7y7u5uwMMAckGSAe69fv7bDjOs+gEGRZMAg 7DD72c9+Nm1hgLiRZMBQNMzevXs3dVmAmJFkwIAkzDhVBgyKaxfRXZ7nrV6fJMlyuRyoMENwcnv7 k5OTb37zm2dnZ48fP3ZSqp6Kopi6CIBjJBka2e122+1WprfbbZIklRccSik77bb35PUyM8uyQUrs wmaziazd58dtiBJJhoPqXS5JnVbZU3+xzpF0zPNcozG4ThsAH5BkqLIDbNBoWS6XsnDJNom09Xqd JAmRBqA5kgxfsccPJxnx0w6flEQizeexRwCeIMlgdHzPk9iQvlqWZXmeSwfRk4IB8BNJNmvaD7u+ vp66LHtIgO12O/IMwBEk2XyFEg/SRZM84/wZgDqSbI6kKxZWKkieyUCo/+kLYEzc42N2JMayLAso xlSWZUmStP1FNoC4kWTzojE2dUG6Wy6XhBkAG0k2IxHEmCDMANhIshnZe5epQMnQ6G63m7ogAKZH ks2F9GB8ODfm6rGTWZbpT7mjx7M6gSNIshlp2yFbWJrfefZ4m7vZbA4tisb6kCMrDYAhyeZju912 6JCVZVmWZVEUDWNmtVqlaXrkBcf/t0OYzWGA8fhKA8DvydCINKar1UrCpixLc99XSNNUnn4iE/oW eXGapvJgFH2vTmtAygvSNH0wCOtevXo13Bjjo0ePOryrvloeXG/63iMrzRxYbwC+OuhG9Nbrddu3 aCVJ07QsS/m3LEtpQPU19hydb7fO0ljbi5W3SHtd+cS2X2q73bb9Xs3ZxW7FXi1N1pt+3PGVVh5Y b811/kaAzbeKRJLNRYdGv5Ir9p+VFlkn7CRr8t6eSfbixYtWr2+rZ5KV90cDe+dXJuoft/eN5b71 1pxvDRAC5VtF4jzZjPQchZNRsuav1xfLUFjlvTJK1vNChiB+VNBqvR1facbRegNiM3WUYjytumV6 DsY+/NfhL51TqUj2i2W+HLvp0mQJn376qcyvfESrA72hhxbLfn0ye7U8uN4q8/eutKIodMixvmmG /kaAzbeKxBUfMyJP/Gp4BaPdYqq9lxhUXiYXLFTm711a5TXS22hSNuPTz+MOsb9ak/VWn7l3pdXX LQBGF+dl6Js8db5evNX193LlfQS33aEHCCQAABgPSURBVALgBH2yeZFOTJ7nTmJgsVjUZ3YLs+bv 8v+xantXS6sXAGiFPtns6O13+/+mWEaox7w3h/8xZhqslvHXGxA3+mRzZD+1sv/zNse5A4WU9tmz Zz6fG7M9uFq4cwfgCn2y+ZKnVm63W88fj7Lb7a6urowx19fXocQYgDHRJ5s16ZztdjsJs/79M7c0 Yq+vr6ctCQCfkWT4Os+22+16vU6SZNoTUXYf0fNTYgB8QJLhK5JnWZZJF00fyzlOlkiO6p8EGIDm SDJU2ZEmZ9E01dwOP9bPzxFgADogyXCQRJq5DxiJNGEO3/NQXlzpY6nKe307MwcgRAtue4M+ml/3 GGJobTab0X71pXf5GhpPNUN/q9XKq4pEkgFeWCzYGREM35KM35MBXuCH0kBnJBkAIGwkGeALnp8J dEOSAQDCRpIBAMJGkgFe8OpKMCAsJBkAIGwkGQAgbCQZACBsJBkAIGwkGQAgbCQZACBsJBkAIGwk GQAgbCQZ4IvRnoUGRIYkAwCEjSQDAISNJAN8wcM2gW7+f3t3q1y30cdxfDVT0pCSgHYmTmGCJOoY FrkkTGLKNThjX0JZ3TbXcMQk1plOgwITUwllwhoHlASUPAUBesA/2a4l+VivR7ur7wdkZFkvq7Wj n3e1Z0WSAQDcRpIBFmHQBzACSQZYhCQDRiDJAABuI8kAK5ycnLx69eqPP/54+PAhLTNgEJIMsMKH Dx8+fvz4v//97/r6miQDBiHJACscHx/r5TzPVywJ4BySDLAC6QWMRpIBtjg6Olq7CICTSDLAFg8e PFi7CICTSDLAFs+fP1e0zIDhSDLAFsy7CIxDkgEWoUEGjPDV2gWAJ6qqKsvSXNP40hRFUePLMAyX Ktl8kiRZ+hTyqOwAJ9IYMwkPkGQYQ+eW/KuTycykNE3vPE6WZXIQoQ/VZ99VeHbfP2RkAsshydCX mV46byZGjrm7LMtZsiwzzzKp3AB8R5LhbtJyEtPTa78wDKVVJ2fJskyfnUgD0Ikkw63MR19rpYg+ b5ZlFxcXURS58lANwMGQZOigM8yeZpC0BaVgZVnaUzAAqyPJcIOFGWaSvseqqqTL0c5CAjgwkgz/ kRizv/tO59nFxcWzZ88sLy2ApZFk+My5Vk4YhpeXl1JswgzYMpIMSimVZZn9TbFOaZoSZsDGMVsV HI4xkaZpWZZVVa1dEADrIMm2TgLA3RgTEmZrlwLAOkiyrfNmRHsUReYnuAFsB0m2aT7d+l1vVgIY jSTbukENssAwaPLZoiiGF23MXj5lM4CeSLJNG/Fsqa7ruq7zPO8fM0mSjH6H5KAwa7wsBsBGMAof I+lwSpJE8qau66IoJLeKosjzXLbRaSRbxnEs70Yxd5RlHZCyQRzHg1IwDMPff/99uWbZ/fv3x+3Y rhbz2js30PsOrTRgi2ps2G63G7S9/rWJ41jW6AW5jco25pd6jXl3lpu13jHPc1mQ+3XjjIMuZ+gV DaLLPIJZLe1Kq7vqrR5baf1NuSJsmW2/OSTZpg299bdzxVxjJll7ofGr3xlRcr++c7PbLBpj9RxJ Vn/5a6BzfWOhfcaeldafbfcjuMK23xyek23a9AdL0lHWc2O9pfSGNb6UjrIpbzF25SNlVlUa4IO1 oxQr2+12ZVn22VI/hmm0AHQPmHzZ/tXS28t6/dec+aXuOmucov+ffv0vZLSJbTKzWhqV1t6gsX5Q pfVn21/WcIVtvzlBbfwXwjZlWTbjh6ODoPlLJa2HEcMXZaRDny1lFv+lP+KdJMnoURXtahm6wRKm XBG2zLbfHHoX8Xl2jOXmLRw3BL//+PvDxBgAazEKH59nx5CHTBNnygiCoHP9iDDruYsTL6O5rVr6 bwBgD9pkUEqpMAxlEt6JH8aSPutxM3qM4ESMqR7VcuB6AzxDkuE/EgnTexpHz+jRX5Zl8jIa+2NM u7NaDlBvgJfoXcQNOszKsrTzpWVVVe12O7cyDMCiSDJ0SNNUhlHIwzMbMkOXJ4qiy8vLtYsDwCIk GbqFYSgNMunHU0qt1UTTj+5ohwHoRJLhDhIeZhNNLZ9qkl7SAlN2NAoBWIskQy+6iaZaqSamhI0c UJZ1etECA9ATc3xgKnPgvgRSz+kcGxvbOcDE5OX0hlbN1ABX2DbHB0mGQ5AR85YHldOKopDZhNcu CDbBtiSjdxGHQD/h0uSzaLbdX4DD4JPRgCfiOB70vhjAGyQZ4A/CDNtEkgFeIcywQSQZ4BvCDFtD kgEeIsywKSQZ4CfCDNtBkgHeIsywESQZ4DPCDFtAkgGeI8zgPZIM8B9hBr+RZMAmEGbwGEkGbAVh Bl+RZMCG6DArimLtsgCzIcmAbYnjOM9zkgw+IcmALcrznJYZvEGSARslLTPCDB4gyYDtIszgB5IM 2DTCDB4gyYCtI8zgOpIMAGEGt5FkAJT6EmZrlwIYgyQD8BkzgMBRJBmAz5jOCo4iyQD8xwyz8/Pz hw8frl0i4G5frV0AAHaJ41gp9ejRo3fv3imlTk5OXr9+vXahgH1okwFourq6khhTSr1582bdwgB3 IskA3HB+fv7rr7+aa05OTtYqDNAHSQbghl9++eXJkyfmmg8fPqxVGKAPkgxA0+vXr/M8Pzo6ki+v r69plsFmJBmADnEcv3//XufZmzdv+Nw0rEWSAbiV5Jl0Np6fn69dHKAbo/CxCVVVlWVprml8qZSK oujO4zT2au+SpumoAtrl6dOnnz590l9+8803p6enV1dXP/744yrl+fPPP1c5L1xBksE3WZbJgqRO I2yiKArDUI2NnNv20ic1z65PrU/qkBcvXqxdhM/Ozs7WLgJsR5LBbbqx1U6OQzaP2ufSa7IsK7/Q JfSj6QZYgiSDe6Tdo4Ph8Lk1iC6YmW0Sb1J+a0sOuIIkgxuk7WWml7sBICWXf0k1YDqSDLbTN3qn 0+s27VRTbj5XA1ZEksFSeuiEUury8nLFkhyM5Jm0Pne7nYQ3kQbciSSDdaqqkvu4fy2wPsIwlMd+ ukN1m/UA9EeSwSJ6IOJGGmH7SaRVVSXNU/IMuA1JBltwv+6k8+zi4uLZs2d0NgJtJBmskGUZz4T2 CMPw8vJSwp5aAhqYdxHrI8Z6StO0LMuqqtYuCGAXkgwrO3CMuT6hO2EGtJFkWNPE7rIkSYIgGLrL uHPNVYDpJMwOfFLAZiQZVjZliEee5zOWZBAJsEEFOHzmARvBiA94oiiKJEnyPC+KIo5jvSzfiuO4 kTpJkuj1slzXtayXLfUGcrQ4jvX2kklBEMguZgH0ZrKXUqqxi5xOyjYliX/66afvv/9+9O77fffd d4O2f/ny5dnZ2enp6cuXL09PT2Ue/bOzs5cvXyql3r59a27w4sWL09PTRcqNrbrxXxE4sCzLJg67 N+OkvSxpJGvyPI/jeP969SX29IKs19/ds9BYqY+sD9i5yzhLf2Lh6dOnP//886BdHj9+/PbtW71w dnYmeSYZ9vbtW1mvv+x/5LOzM95PZhv9B58laJNhZVVVLTrc47bMaKyXppK0pWTNxLEhkpfL/Ye3 /IN30hpTSjWaX7TGsASek2FNURQtPXhBD/FoJFNjvTSe9HeLotBtqdHnzfNcAnL0Qdx1enrKGzJx MCQZ1iStsdFjyiV7giDQvYUSG3pZHkfJBmZWNdbLSrNBJuS7Jycn5unM77YLoDeTx2bm8c1dpsSb fG5h9O5L0C0wvSBdi48fP9a9ju0vgbnwnAzr45PR/R2mrkY8J7vT6AzjOZmFbHtORpsM6+PTvj0R +UAnRnzACmma6jdqcqduc/0NnI8fP167CPAZSQZb8EauThLwrs+CL/2K8mmztcsCD5FksEjjjVzu NkFmod+a7c3b2ogxLIQkg3V0nkn7TFn/2al56deNqo1dODAaSQZLSZ4ppbIs060Tj+/sOrmjKPL4 MoElkGSwnb6tS6TJvV55kWpm80spRYYB45BkcIbc5fXAEDPVHHqiptuXwqGSA9YiyeAe3fEo2SaR prvmZBtLEkLnlrS9nMtdwAkkGZyne+RkQT9w0gNGGnM7zRskjR5CdTO0dMHoNgSWw2xV2JBGz576 kjptZg712Ub51dJ6+vTpp0+fZjnU33///e233048CLNV2ca22apIMgALsu2Wh1nY9mNl3kUAC5r4 mjegD5IMAOA2kgzAgqa8rRToiSQDALiNJAMAuI0kA7AsBn1gaSQZAMBtJBkAwG0kGYBl0buIpZFk AJZl1WQQ8BJJBmBZSZKsXQR4jiQDsCAaZDgAkgwA4DaSDADgNpIMAOA2kgwA4DaSDAAwjG2fESTJ AABuI8kAAG4jyQAAw9j2AlWSDADgNpIMAOA2kgwA4Lav1i4AAG8FQWAu1HW9anHgLdpkAA7h6Oho 7SLAWyQZgKWY6XV8fLxiSeC3gPY+gIUkSaIng+BW45MgsCs7aJMBWIr+1BFdi1gUSQZgKXEcS4Y9 ePBg7bLAZyQZgAVdX18rpZ4/f752QTAn5vgAsCHSJrPtxgfP8Hky+CnLsv4bR1EUhuFyhbFQkiQH O9e9e/cOeTqlVJ7nhzwdVmfX+BNgv3Y+lWUpC1EUzXiiPYdN03TGE60lSRJfb/ceX5o9bBu7SJsM ljJDqyzLRqLoOFkoVxqHrapKZ5sUzCzSBpt0gFXsylVslo6KRmhZ3gAyE04WpPCWF1t53XDx+NLs QZsM+KydXmma2p8BpjAMdWtMSi4XlWWZvihabPCPbUN4SDIclNzoG3d5t9JrP51tclESabrd5tOV AvYgyXAg+rlXFEXbuaGbV5plma6E7dQAcAAkGZZlDtzY+O1bX76ONDoe4Sg9naYlSDIsRT8r2niA tZmRJlVEngFTkGSYn75BX15erl0Wq5nP0sgzOIQRH/BZVVW73Y4MG8TMM9qvwAgkGWYj4xLJsHHS NK2qKssywgz2s+05GTMIYx4SY9yFpwjDMIqiLMuqqlq7LIBLSDLM4AAxNu/fgHuOpr/V/4wzlk3C TH/+zF1Dq9G2v/Gxn23PyUgyzKA9L2J/SZIEQbB/m6Io9GTqd268/0TCPGB7s8aCuiv52mWbcl+W cR+D5vK3UGc13maWny+2jCTDDMqyHD3urs8UeeYfgENne5M7o0RL/UWe5/3/qEySZM/Gt5VtSpht rZN2ys8XUIz4wHRZls31RhX52zyO46Io4jjO8zxJEjMSZIO6rmUhz/OiKPRmjV3qupYYk38bt0h9 9zT3va1UnVveVjY5+P78u9M///yzaLPs3r17E49g1rPqUY2NHff/fCeWDZtTA5Ptdrspu5u/h3pZ KRXHcef6xoIZGE+ePJGVeZ7nea432/Orbm62Z8E8izTp9pRt/xn7mFildzLLP2V3Xc93VqPesf3D 2rP9lLJhObZVMm0y2Kt/B1395a94/ZSl3R6SdsAsZ1ELv3PZ/odkutJ0lU75YQET8ZwM81jo5tsz MMzN9LJ5b5WOrPaORVHclnB7ziJHXjTM7H9UZtbGuGpUC9chNmTlNiF8sdvtyrIcsaN+rKIfseiV 5rfky8YGuotDb9ZYro1eSvP5jZzLPI5oHF+XsH2WPWUTo7tfRtfkINN7h9q1sb8a2zveWYfj2Nbx 5SXbKtmu937CadIss78xMZS0wAYN3+jfQGmQ4TMHmIDR4xcre3xp9rCtkuldxGzSNPVyioqhmTRu /L1MVcU8wsAIjPjAnOQuLG9J9qlxNijMhiafzJCifGzOwksWPt0kyTCzMAzDMJQWhuLuvBcZBsyC JMMiGnmmuFnfpKuF7kRgOpIMC5I8U0plWUakqZufVdhyPQDzIslwCPqubUbaRpojugtREGDA7Egy HJQZaWVZygumlXep1v6cOAEGn1g1BF+RZFiL3NnlX2m1CD0ZsVu3fh1d+hI8y2bAZiQZ1qcfp0l6 SSrIv2a2WZINZnurXTy3AhjwA3N8wAGNR016ufNtMqOzpHGWxuka53I9sWybo2FGHl+aJYLAuuCw rkDACP3nL+7/emvXs2o/Cz/cOtTV1ZVS6vj4uP0tkmxRJBkAzEbm4FdE1wFJndtW4cy7CMBVt72/ G8uxs55pkwHwgfSXymtl1i6Lz+x8DEmbDIAP8jyP47goCg8eAWIokgyAJ6SzMY7jIAjs7ARznbV/ JdC7CMBDdDYuwcJRi8LSYgHAdOTZjKQyLXxIpkgyAH5jpP5crG2QKZ6TAfCb+fDM2sc89kuSxOZ2 LUkGwH9xHEt7gsEgo9ncqLW3tQgAS+Dh2VDSILO5ukgyAFtEnvVk5/RUDSQZgI3Sg0HIsz3sb5Ap kgzAxjG4cQ+bR96bSDIAIM86ONGvKEgyAPiMPDM50a8oSDIAuEGmIdavjNkmh2JM8XkyAGjgw2d6 IMzaBemLNhkA3Gqbg/Vtnpiqk2PFBYAD29rDM7f6FQVJBgB320ieuRhjiiQDgP78zjOHht03MOID APrSAxr9Gwzibowp2mQAMI5ng0GSJHE0xhRJBgCjedPZ6OjjMY0kA4BJXM8z12NM8ZwMACbSD8+S JJFIK4ri4cOHa5erF+c+BN2JNhkAzEYenr169erjx49HR0fv379fu0T7OD3Kw0SSAcCcHj169O7d O1m2Ocy8iTFF7yIAzOj8/FzHmFLq+vra2m7Goihc71TUSDIAmM3x8fHR0ZG5xs4w82CUh4neRQCY X1EUv/3225s3b+RLq7oZPYsxRZIBmK6qqrIsJx4kTdNZCrMuPSLf9OrVq3///ffrr7/+4YcfVimV 6erqSil1fHw8+ggWPlojyQD0ZSaWLERR1N6sfyZlWdZY0z5sFEVhGI4o7SqcnimjDzsvkCQDcCud NGVZ6nQ5ZONp9QIMZeeNfkZ2XiBJBuAGCQ+dHFY1idrBZluq2Xmjn5GdF0iSAfjcbWhtPHRqPJyz JHHtvNHPyM4L/GrtAgBYU5ZlEmBRFDkRYFoYhjq65Cok2Ny6CsyCJAO2SDdonAuwTvoSsiy7uLiQ YLahiYbDIMmAbamqarfb+RFgbWmapmlqdpaSZ1tAkgEbIiMmLi8v1y7IsqTjUeeZl5kNE0kGbIJk 2KbaKJJnWZZlWbapC98gkgzwn8TYNpsmctVSA4SZr0gywHO0SJRSaZoSZh5jLnzAZ1VVKW7fSiml 0jQty1IqBJ4hyQCfzTLeoT0lrqMkzNYuBeZHkgHeas/P219RFEmSBEEQBIEst7fRG2idmykjC4Mg GF2k9tG8MaVaZqlS15FkgM/GNcjk1SR5ntd1Xdd1nued77LSGyilZKFzM3kb1ohi7C/h6H2nBPxC mDVwIkZ8AN4a3bWYJEnj3qqjKEkSSZHOm69s1tjmztRpbN/eXbJQwlVeETklHf/666/lwuz+/ftD d5ELlIsNgkCuVK5OFvT16hrYf8D2z8hc067PoQW2UQ3AU7vdrizLETs27gwyY6zc9cw1ndu3t7lt y87tO08hC+01I4yuk550+QdpX1djoV0D7X0bBdAb3/YT6TzancZd4NKYCx/wlkxyMaJZFgRB46/1 IAjqupZ/zTV9lm/7buf2ymhJ6G/tWRjq4uJi0SlOxk0Vf+eV7rnwxpo+9T+lGu2cC5/nZIC3wjAc N1Qvz/MkSdrDN6Sba/++fbbRpPfM3H7Q7kNVVdX5kmufPHnypFGBi1apJUgywGfPnj0b8UxIepCU UnpEorTP5I9xWan/lpcHMHoEXXsb/Se83lJrb995ivaaEU93RrdQD0byRj/T0guiXSeqq0qfP3+u uuq/sXvn0dzlz5UA6LT6HVzutlNGFjTuuXpARH8Hq4SJvYujN5j3dHvQuwhgBWEYRlG04tDzeUfH jRh/LzFmZ7+if5+NWwWj8AH/yWxVK84jPLFBNuVQls+eLGPi97dy5v3ss5efpKZNBmxCGIZ6Vni3 5h6UYdYj2i76fS7Wxpj68uny/fE8ugYOcDRL8JwM2Bzd02jzLX6KLMukO/HwF2jnY6QZ2XmB9C4C m6MbZxcXF1EUefPOF3keppSyvB2G2ZFkwEalaZqmqdz9dQC4GGn6EnxKZQxCkgGbFoahHg9SluVu t5Mxfpa3aXTzS9AI2ziSDIBSX6JLdzzqR03md9elH+/pgtlQKtiAER8A9mnnh1g0Rcwml+75VC70 f9o5IGJGdl4gSQZggEa3nrqZNFN0Hse5VpedN/oZ2XmBJBmAObWjbg/ngupO8pLStUuxLJIMAICZ MccHAMBtJBkAwG0kGQDAbSQZAMBtJBkAwG0kGQDAbf8HYvFjt2+u+bgAAAAASUVORK5CYII= ------=_NextPart_000_0000_01C80F9B.F234E740 Content-Type: image/gif Content-Transfer-Encoding: base64 Content-Location: http://www.cs.wisc.edu/wisa/papers/security03/images/ground-int.gif R0lGODlhEwEzAIQAAAAAAJmZmXd3d1VVVe7u7jMzMy0tLczMzBEREQsLCwcHB6qqqoiIiGZmZv// /0RERN3d3SIiIg4ODru7uwAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACH5BAEA AA4ALAAAAAATATMAQAX+oCOOZGmeaKqubOu+cCzPdG3f5hDgfK33PEYD2BMST4fIMZWcAQiOxqQE IEEAUwH19BxMl6NrdmvqfkVVrioA0BUQC9EAwHBcAfi8Em0flAp7I4B/gSJsbnB8EH53eXiFaScR UBNKdQsEDRAOCCUNBZ6gJJ8lkw6VDnV8JQsIAA0Lm2CztCsTAggIO6yusLK1wCW3ubsjra+xwSnD usrOz9DR0tO0aQJaCw8kkdzbI5EOjgCi4ODUOAEPAAgHq9cO2d6r5yICQ6oQAgETnfLh398ciaJH sKDBgwgTKlzIsGGELz9WmEsxcdZDOcVQVNTIMYUAVUZUFKL1UURIkTn+GGx0QKAAFAYjj7R8WQhK w5vAhizUibMnCwgFJEgo0M6nNKBCiRpVgXRo0aVIbAJbCfXFAalVyWQtKYIAg4G1nmZtwZXl17Eu xKJdy7YtjiZs4Ra882sunrpuWcxMFdPnXpjBsCJRhwsvtQOEERg+EjFYYxSmUNV5h1grQG3+RGDS xCmUCVKjwDqIbIkPZoAkNm/qN89ymRGKi371k1nEAwQE5N6IFNvB7dwjN6686ODxEuKPqZ6AcBo1 C+XUoDu/AVSX2ooF6ATLrqq27wh4zrgwE8xcBJd5qT0odF5wehPrR7RXGH/EhAbgxenfz7+/f0cP yMKcIxP8Z+CB/kX+AAsL9+WH4IN5BCjCgBBWCICCcajQIBYVSmiHOhZiCJUBCbxHC4kmoliVaNFI NxaL7xHg4kEyihDBJgsEImMdNdZj0wSqyKiFFg20A5ROAmR4hVu4mVRFjzfCo6N2PdKT4wja9LOA OWlUGUmRdhTA0zlpNODeP2h8IUBRB2jRmgBfEIAZnF090dYrUBCwR5lnciPeOQMUIAsAmxzwRDdo QBFAkw7QyVJz9BwwgCsXcmbiNJJSquBil3bq6aeghirqqKSWamotdK2V6qms4nASWq8SpJtBsx6x Kqp3rYAYAIWNtWuvBxlXkLBHxEqLsSVwhWwpLvQFhrJjSvJCTAP+qKTCX85Gg20NZ47wq2JhJcap c701ioCYrqVSAG3T2YZbrTXwBoFs66b7m27CiQRRRmoEgxy//b6GArS1EFyctSYcEsAbcbDjQABa NOLIHnPUIbEeAC1CiAmDkNDxCAozLId2Fz+S8QAlX+gdCaSlYpprqnUWmmczs0yJJTGz1trK85R7 b7YxyPvUjGxkJ/CEbwRw3QncAVOd0uk60LR364X3ghiNRk1eLVu36zXIbSyciCLsivCx2SOd/XDY Iu+cAkzVqhWFC9EeVHcKdx8xIzR7szA1LVVjIUiGrfagWwGEiyoX4gslYaighd9AwBxiPT5uXpMD 8JTlNFIuwq7+FoZuoIJ/nvIJh6Knvp9Sg6k+uhQm3Jed6/+xXgLo4SFIeuynW2i7QgQs0MA6BxzQ SwCXnxr88OwY/wryWS1PvPMNQA8qAQko0G3kJmCvfXreb8/9+OQfxNxheHh6vqcCpF8+DH2/X1D8 56iUxq4RXNOA/UnwSkL7CHCT/QIQAW1kYh1mGgEuAki/hEjqEXRwH/70xz/wuOkcrghgcdpwnwvx qwr2E4ErRnBABCSQIGlA1DwiwZ5vSOU0SxoNz6AygPxNoSgpnI4K6XERCCQOTc5RIQwbSIQ0MAAX HlqhLArBGnD0BRJuYRcD+mFEJNaFG8lTxhW8YqMvVOEMKmRgIUIEkJ9G5cGEBCDjOjhxITct0I35 OQMuCiAAqbQvf3i44FjagAeiqLEK7cMDGtXYCVfkryALYBcunveKRoGnH4GMZD3OVceD6OIXENCZ /GRwyTBocpOgDKUoR0nKaYQAADs= ------=_NextPart_000_0000_01C80F9B.F234E740 Content-Type: image/gif Content-Transfer-Encoding: base64 Content-Location: http://www.cs.wisc.edu/wisa/papers/security03/images/pattern-example.gif R0lGODlhIgFmAIQAAAAAAJmZmXd3d1VVVe7u7jMzM8zMzBEREaqqqoiIiGZmZv///0RERN3d3SIi Iru7uwAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACH5BAEA AAsALAAAAAAiAWYAQAX+4CKOZGmeaKqubOu+cCzPdG3fblOYBCCMCsNsgGgdSLpRQLEgFB4vwWEA iA2gBAVzIaWWAlBlGEcum89o0Y5k+AEdq3XJ8TAUE9tGNSECEAgBI3ILA24IDCeDJlUoimojD1sj jCQPbiKWaZqbnJ2en6ChoqOkpaSUMgmXJnYHqy9QAm5CA4ELBnC3uZMkqKguvy86tryLJQQCAUKm zM0kAg00skXOJ9DVLHxfAtTY3t8wDQoCy6EHAAYHB91lBgAICOVsUPHg9vejDa8mB9EnwTP6uRCI 5gi+gwgTrkAmIgGBFnoA+BsR0R8VAA6KNHAAgAG7alIWKDCosCQpXCP+GKwygMgGpX5CXjZYJrNc GxF6iKUYJyLAgEcjemBqKUASQEoAY/gM8NCk06dQo0rlhMBSywdymp6o9SIBSUgPGhh8ENagGx1C CJBMAKBtx3JOHibIFU+Ag0BCBeG9uwBtE5IGCrQ9II9gkwJy4dRtKccwpaoCrjpywXYij10pCGCe asIrgMmca9jFuK9ZggeeQZe4OYPhwtKhY8ue/S9FZRaOyIpdUJdvXhGGF/jqlUhF0gWgAxRYZ+wY Zs0nzrmd/pm29dP3sJd0LYK79e+zZdU4roCuvDRYn11y4I+ljOMohhMv8bMEfPD3GIx5X2kkup6C +ZDCZiIwAIAkXKn+4QAtAAyglX4jBNaWAlqRQCAX07VkAoESHtiUABmSkMB5v0jxFX4opqjiitUg ME6FA/kTwC73vUDQjAvQIUI6I9yImQF8SAEji0QWaeSRSNJ3QABeLdgCaGwBCUAYLPkgSZVSTIRl cAXyhdsyTnQ2CAK11NfHjr2A6GVfPxigU2Z4vJkkOCgdIx0AciYUAGyc7TlnVF6hQAQNc5FAJig+ pWUXCSfCQFISZvRwSRBf7IcCGComgNgCAWgoXAqRUEbSIbbQYcckBIjTVAFg2oLjJIgU4cBDD8Ch DR447aEGFl6y2l0gDCjgj4ta+QGICLMuUOsCfOBq3x+2DKLpQ53+rkYhs0uutkp5KGSygrd/hivu uOSWq8k19qCbkDYjsGsuZxmd4A4856Uh3gjxntkjfzAw0NQDnn5aAjfPfPSuUwABycUBQ8bYoz/3 MiuJvhQLDEONKWxkX23zHQxVMveA/NQrfHps8skop9xQnk9OLMJF8W7UkcGbJECMFPnCkB4MWQDg SoU7d1daAZapnMZ/NMDHwJopTCaFJMtZcs4PILpFUznylaCaxcWhgAuJJjxgJrJGd4IxCj24u+NZ eM5wyC077rJgNB7BXfEuv71wNr4dVyxICQyoXbYZDQR88UGFP0VWcyT4OfjjkEduEsEEnyEpEOcx cEQBW0sumyX+6q4AmWRBNe3C6JjsQMAA0QTGm1WpdyaR5+WCXjQK03Lq6XGdi7hptTmm5WBDv2vI mgjcZpMt7eGmVvM5vTtvxvHMV2/99VEZ1KgKzpbQPfLmrbDRA0748zOzATugYzH6HhcGzSMIYCn2 ZECRoFBYfQZ/FLduIUD/8QPgahDgihZkbQQIcNKucjSR6fSiKXtbwRIERz8YDGAA6lgJny50qQYp hzkB8OByqBHCWowwEudYhnLacgkTlSCBAtrRRQpgix60TU1Vm12B2uKAHN5uIXiAnwsrSMQiGhE8 oesEBTuTsgAsqWH8WKINlieCyHziAToUQTciFoOI5awd76j+hzX2Fz8yxsZH7FsEddrywwUcwF0B MJ9lCkAN802oQm+MkIbWlyDgQGw9tqiTdK7Vl2No6Ht8JAaXvvcLNOZIi9tjQ5AYlgKMRfApXErj CcQ2nXpNpU4ngOK5FPnDU3FqX9JoQcmOyMpWuvKVsIylLGdpG53ATCMcqRsobBa/A3zRBUFzQc/O B4lBeMc7tDwDF9+gRU+aIZhebI/hWHDJA3LtZfRJ5tGcuTSWkcBpB9hC1KQQwxwiDWlcA8jWeKeC r22sBGKrxNi0iYOktGKV3eJCmwgRyF2AsmLWNCANhvHOgvqNni5hBdu8qYK3CQGUcytQEbBmoaBc 8poswMz+L9Spteis0S29oyWElOUfFQZogycw0Bb6WAAFUmF4BRpDhwg5B2uEaEPyChAhzempEYnI mQiVQeLsMVSTLA4sQVVmG5uRRHyoTYpJjapU59TUFmRBAQxlRlXtIbg9ZfWVkJqBH0IZQ5EAVQWD YsFXwrqEw8wvBV245BWaoIUqTuEXmDKoNn95OK+F8axk4GLOUBFJ48TAX0PRKxc+clEiNvZvtsmi 6eBZlmbaBS8y6VtAkWPYyX5hhIwLyma2GsvHGqAQXpukKFcQHILc63vp7BswaqAx2fqtZ9OEZUgr WTTqJc+zFqoDNZylB2YJB1qQfZkhprlOFThCDqGyLbj+/OaVr9JPZGI9CHYnF1pUxlI7MijKQcCb EGQWSFtTTa9618tezrjGIWa4ZV9yucUjjKS93yAAVNOgPU0kT4zw1CJ+myEOYbUAg0zyJWCBIjsp 2W9JTQKTAjCyDARH2ATdPIE+bmHdAZ+BTAJYbdj42Z0qDIAtOLBfDavwP16ouMQDLKCH7yG2EL8A lCrRJA1w/ANKtFUXKbkEa3IyY4QUeKkvw6CMdcw3F1wwgxHiCOtG8GQZoxBpKywrCoZYZDpdaGlt eeuNOcgKMtfAcV1Os5rXzOY2u7mLRahcC/4ZZRJcDnmATSAVT9uWKZPgvg1Z4105qDnk7PbNjkKH Osz+SOU89XFHf2XBvVxUxb6JpcVprI896XFWZMx1vahTVkgPHRdm5aLUherOplLtPclWsmOFE8x+ ZLGbYviCCQnEQac6HEs00qWwLkgkssLQR2FLcmEiDh5OtresPriFGIxYBVLKUGOp5g54O6LpN1+Q rGXxodu2Qhatwn2LbW2mukggjGaoka/TukG8ombytM/MADmhW5uZNCWvvQHsdImZUcmU3myoN3B8 IvrgCE/4dxQQIHTegOEd4aGNQZE/KyGZUPtGKw9xOTOhscC87+Kk1VJwwTQMYGJ54wQlgVBMGgRt mTj4ryfnKSgVWVjBLXjAylngjno9rZdb2EUDqLD+ZHc0RR8wxRLUDiA1LSPDl5Ubus8uAfVZd3JH BoL3CEH0E3PGpBw3V6AMMtwtmofN7J8jcRNMjOL4JJvKNI+FG+SOE+QFog1grw8yWihAWvQzfhoS CFbLfR4CIKAAZvonpnHl99VoFBJqz0v+6MhzApbMkiniMZNf9vZbOHy5djtuE8IAED2UgwEaqknc 8GUAf4xVYgd9xYYlajfV03kBdaMJG3aRYzbkdm258uaoU1TlV8CnU2TEIkams3KVPoPCUXZFEZSv jgrrj1MiZI7akaNAXADAK+1xgPTVUDVivqxBTSnhB4vw6PPDdKSEUPIrjsflK5uUhZ1ZsE/xc3sq 3HNkSkfyEENncJ8wVP2nFs/WEwQYNv+GVAoXGqSFBvvVEA9YgRZ4JCEAADs= ------=_NextPart_000_0000_01C80F9B.F234E740 Content-Type: image/gif Content-Transfer-Encoding: base64 Content-Location: http://www.cs.wisc.edu/wisa/papers/security03/images/binding-union.gif R0lGODlhWgFTAIQAAAAAACcnJ6ioqJycnJCQkISEhHh4eGxsbGBgYFRUVEhISERERD4+Pjw8PDAw MCQkJBgYGAwMDLOzswAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACH5BAEA ABIALAAAAABaAVMAQAX+oCSOZGmeaKqubOu+cCzPdG3feK7vPI8ACBhAUAL0TsZX8khazgbAxkAy cCRWQZMDyWzeoAgp1Zpyds82gJkFmHpFEUnjKnAjIodiXUGkyg0PbyWBXHBydHZ4egN8fg2ARSYF Ci0OCgd9BQl9ImskDQh5ZYZzEnUid6IjQ4x9U4+EnSeTlZeZmyMEjatovWmevsHCw8TFxr0CCgAN lMlqVxI/am5lwMdMS2tDKQnQEkYFEhFuEFNQIuERRAJG3SKPIgZbEuEPBCIO4d/4BiIKQQUQUDMT od8IBRDe/WBRkARCdA8APDDIq1eCOJPsufC05iI9BRqJOXjA6ZrJkxb+qRXLgtIBBDX6dgiIoAaA qjMA4qDcybOnz58nA8QCSrSo0aNIUah5QYCkAAgljVbJBWEejAIN1DgY0AClAQBWJTgAi0UVVgQC EryEcO/F2bRr25ZIm7ToUhcQAopIWDcbGxIPVNLdJ8JdNGgHKElwNwUBNHfbmtKToLFtAoFTkiCg VCBQuG0mnISZq1iWitEkkpEgEJZyXaIEIkSI+bq27du4cxCAwDK379/Afd99MdQ0FSIOGjxFUaDA AJ3elJ4oXvG4WOV8JTmHvmJNg5sqLKVNEPFm1BEJ5CpQ+WYAcuzMtxfufuK7C/HdyqtIwD64Emso JFCacSkQEJ0MAIb+xoKBaSCVIIM1JHhEa/79V+FGJyDQmwQKUNTLAV2d4ICHYpFYgmEiEJATJwnQ tuCK6LlYQoi9vKUWAGztYGNcLCQABARQHQAWWiNQdwIURl54IwA3UniRROAZQNMu4vRRDj42waHM ViII6cAdhGSVxQOKSCBkjotJlKJsLZKQjwhVPBmBAjI6cZFOY8hJZwkQkoCAXF2gNoJq1ZUwzlx4 okKjCISmSKGhNUWqRpJ+RgDohZhmqukxw23q6aeg7tBpqKSWaqoKo6JqCiUSoiRNmqnud6Afq+J4 jwAPsDXWVmMphmteEJTJGpNX+KgGX8ooVkBVWdITC1c5eQMAoB7+ZXRpC9WCdG0JWJ26gzLgqRCu tzlsZtJUlbwEQEwHzHrCTDWZJSO59NZr770pFMDkC8lcskqrNIwFcCcDyyCwgwXDcLAw7Y5raqyq RtJLwQnP4MkBeJxXwleqHDBvCQeQqFwhOKGAMSYrcDyCxxHXC/EJBJQi8aCtTgQnfUrMVbNB/c38 r6RtmABFSewIYsJ6i31MGKM739xyE0AHXcLQqRnBl5SpbUnvAgzEoI0LiaIQ2BYqUQz2CmPTWigJ WEW67V8NMsRC2mXnm1VNb6NAgEE3VXzhy/gGLrhRgPuHrgS7Pdo2WFx5RdYIu5Y1wo62XgUKXJW/ 6y6+XKKQEKX+BIP6qrFmoDgCNc7k+OuwvALga67LCjuWO6QnlCw6zIrS2elZRSAttRhpK0O2IanQ Lb46cakVoHuy4bdtfg0imLToQeNYl4oxdlhhV0TmGWX33HPZAJmhwtn3oPksKKMDJrj+qny2BroO KqoTI9w+pwhjYUoTdyULysjb4GhgrmMcbgUugcnKNmcCeKlBXsFoyEH4AgpgcMQEEvRHQjojERPR 71ADDCFu4kSTOdUpBQcQGAQUcCkS5qR5KWKgD5olwhra0CSFu6EOdyiq5/Hwh0CsRhCHSEQa5DBD RUziDo9oiMWwBjcV01AJOhQMEKVgRCXAIjeiUz8W9c8EXbz+HwoWRUQm7iNEPiRG9J6WvhNRLx3k MEcS0rGOdkADHhKQx+TAhw99JEGL/6CHQNaWQQ5RcCErKORDOGgzow3RjDMJlgx7sqQmxeBJD7iJ lFzHCftJ4H9jEcWcGNelIUUgTEAoUpnOZBk1IY5NtHlTnkoIw0Ld6XRWoCVt+jSCPymRMuv65W/4 c4wNCfOYyEymMpfJzGY6k15Cctgzp+mtSMLgAQNaWslyNjEbUA5NOUhjC8T5zbwFxl5jEaCC8ocN IfgCGItTw7bCyD8ESmJc4sSfJO4mT5jtL2mjsJcZCUSgLcWGjCS4iGKyIx0TJMkJBo0AQtEzp73g jG2eXAH+8dQ5gkUNYHNLiOhEC1PRT14UoxozwUZXwFByDfR6M+vDA/oXDsP0zJEXlSlNFwONm67t GynFoIceIgdENtBX2XyDTo3HU6cFVA9gG+oGI9JIRiXgABzV1EDZsSEz6JEN69CHlCwxFLOR4Kvd CWseI0BWdo6MBe9rlDandoWkGg2tqFLrWtv6UzkE1U9klOsTIWfMhyXsa/iAgAFe4o1bJjRzI2jp XC1KMpcsdl8kDZuPwGnSdYJMiyhwoQk9OwvJSsyyjKWoZiFLWZx2CbRHyuULdzmrfArnsKRF4SQp MxIAeMisK2gX2npLItuKSF1ffJrJdvsA4roWBw6k4WT+TzVQ7lGkugobKHZfsDDCGVcL26VBuzxo WGqaN3DhPW8OiFlMaqb3NbH4nxvdIKTd6mAAQ4mF6ZqaAvySAAxiiBMMAGwOMpgsuafqDHnNVJXu fDcpa/Dvif7KA38hISptNIFiz0pDlsFAZV36YgIWTC/TOmqcDwZK0QhKGaUJgEuS+VNypuAA5aGC NetZDq5eQT0zhYwQCejKcvg7gCsVeW34IY90KSwWPuRHuu8y8anadFUO3UMTvSxsaFIMlDX+1Bt5 wV2KCHFkelhFMnlMSJirVBiW7GYVV+YLaEp4ICdQjWbPPV0bV3y1sHEZeqaAym5cUaSN/NknXkbF OfP+GJ0rsaYPeLQwW4hZiqYQQR1caVE/jICAfowDF5jm0gM6HQ40NafQkLMFOnBRqFS4SdX0YHWK qBRrXxA4Tyx4FIVuLWAcYJkFX7FrMtFsjCpzkwRhU8HeupTQYIDYTAi+4MY6jOAdQMG+STxARLB9 BARUG4zcRsFgfJFk/QyqoewsN5QZErVJrUABDmCyepd5Z6Y9V9p6JprV4pHsHWAsq/NOpiVQFuuS AIM8FGkKeAZ+C07oQt42gMD8Ah7wQy/bTL1Q0UgpzvG/HbrjIMfhx0NOcjWOvOQon9jJU87yI7y3 5TAfxstjTnM0zLzmOO/BQDH2gI3n/ATsXckxzYj+NF1oGSjx9ekW6YvZXkgY1fytngqeToUoFDjc vDawCTycRDMmoSv9dlAMqF4YiN/AwheGago2PIJnc90Fbu/fiJXodTNtui4r1iZNgJaomYobxoGQ sRhqXL5o4Ng9CdnxIapzgB8XRsh8samRs2MGdZsn105WsjQDTfeBGUExKw9nX33Wkb3oY1ljhhNf CnBmQiz2k/rw5GVWQ/k4d6IPdPZZvU3hhATtvmh9ZicQic4lA4Qbh6MHjJ8WbfzITuHR7zCIpA3E Y8SRpEqZLsDdOw2Hj166DqIm9SfbcmoRDIXhqzZ4pW6C/oLngta/LuJLlH6iYU/8DMbGkKFccPEl mxx/gF/xKNSkbU1nDN5GA7wkA+OmRLFhKT/3gCjRFEcHgYETAgA7 ------=_NextPart_000_0000_01C80F9B.F234E740 Content-Type: image/png Content-Transfer-Encoding: base64 Content-Location: http://www.cs.wisc.edu/wisa/papers/security03/images/sample-vir-auto.png iVBORw0KGgoAAAANSUhEUgAAAFkAAAFwCAIAAABcvUCyAAAACXBIWXMAAAxNAAAMTQHSzq1OAAAA HXRFWHRTb2Z0d2FyZQBHTlUgR2hvc3RzY3JpcHQgNy4wNULvQ9UAAAxDSURBVHic7Z0tc9tKFIaP PR2jkExnipqC/oSOSEmQQmTSABk0AUYlQWVCAUXinZAgg4RYICUJiVBAp8DTH9GyzrQmF5nogpPs WUv+0Mdau3beB3RsRVpJr45WstZ+2smyjCpyfHw8m8349d3dHRH5vs9v0zQNgkDNeXt7W7Vxi3RK ZsH7ryIgojiOV8wfRZF6nabphw8fbm5uam9lO6zJYjqdnp6e3t3d+b6/eudXw9H0er2rq6v9/f3a 7WyUpVkkSTIajbgQmqSgk6Zpmqa9Xm84HIZhaKRNgyzOot/vz2Yz3/dVR2AQlYhrvcmCLPr9/uHh 4SZS0EnT9OHhwak4uvqb6XTaThBE5Pv+4eHh0dHRdDrd9LpKIllMp9PBYNBOEAyfg4PBYDQatbPG 1cg50lpFFImi6NOnT9Z708csLAbBnJ+fW+87ukTE106LQRDRbDY7Pj62uAHEWSRJYjcIIorjWL+p tUKXiPi2ssbCaZqqf5szm83sXlNeXFxc1AgiiiKOgJflfz3P47/WvmE/PT212Gt06xVFmqaTyWQy mcRxzIunaer7Pk+k+c9mJTF1p1+bDhHx1leCS0Bf0PO8FW/LN1vjGYIpuvrjhvLwfnqeZ6qzcIEu 1apnIppMJr7vR1FUb/EiuedA7VOnLlQtxHEcx7HqRBvWiP0szs7OauyDqgXV73KNqL/W6wg/fvxY YylTdIkoCIJKcXAJeE9w38EftHgKaRmVp9fr2X3k1cmybDQaXV9f272kRVG0t7dn95lol4iGw2Gv 17N4ReDHXNYfDj8+v7i9vX14eLAShzsPuORZjpU4+Nnn1dVVmytdxtwzvpbj4IoYj8eOjBLg2a+w eExAjZJt4uKyTWMCin6/33zETIc/6QdB4FoKzPrx1IuLi7u7u4ZDR1EU9Xo9cnu0uezYMg8p5h52 LKuX3MByEARuDhrmKJtFjn6/r17z1w7evXv38+fP7f3CARFRZogwDE01ZYvu+rSeDchCQBYCshCQ hYAsBGQhIAsBWQjIQkAWArIQkIWALARkISALAVkIyEJAFgKyEJCFgCwEJ7IYDAadTqfT6eSmL5xY o/G1Ux4xNbjQcHyEN2Y8Hqsp4/GYh9qaNDsej/U2eUpuRbINTdak0zwL3spcg2qKGoLk6fxWf509 7afeSDFKTmfh1jqURfaUSPZUFGp6GIbqSKoI9JkzrQT0THNZqHkWloZbWagjrFpTGRVnVnnp1ZE7 93MLqmbDMCxusFtZZE+HvZiFfhj1mdWh1munOKd6q5Ob37kscv2lKhPVvn626zPrp4Z+Qi3bwmJp OJFFrs/Td1sdvVzfuXC9ah6VRa6p3Fs1p1vX1A1RvKYuIwxDJ+61NgdXSpIkq2dLkmQ8Hr9oZZNs Uua7UTzPjtdFJZCFgCwEZCEgCwFZCMhCQBYCshBqfh98rolO5+Dg4O/fvy9fvvz9+3fzBm1hJgv1 +uDg4NevXw0btIWBc+Tg4EC9fv36dfMGbWEgC33/v3//3rxBWxjI4vPnz80bcQED/QU9dRnv379/ 7nWh2PYCMVMXb9682eqrKWPmudZWXz4Udeoi5/v1ff/Pnz9E9OrVq5zbZCt+fqio4Pv99u1bmR9k MvD9LmVHfL8GTbfql+xhGA6HQyNtGgS+X8Gy88GpApnLgnuHNqWmURSdnJw4EodN3y8RxXF8fX0N 368A3+8crvh++T4Svt8ubwd8v7QJ329tJdVO+X6JKE3TKIpqGF0Zu77fF018v+q1mt7kXIvj+Pz8 vPbizTHm+9X/VLsu4Pt1Bfh+BWO+3+bYz8KU79cIO+L7ZbgdJdOvBHy/j8D3+wh8v4+4IzWF71eA 71eA71dY6vvlUbJNXFz43sT6VaPI4u8c3NzcZFnW6/U8zzN1i01EURR5ntfr9e7v710LguD71YHv VzDm+2Xg+3387ZqppmyB70ALyEJAFgKyEJCFgCwEZCEgCwFZCMhCQBYCshCQhYAsBGQhIAsBWQjI QkAWArIQkIXw3LPQlWxuZdGa+FetaPAEkXvjI7ytGxX/8ir0vz6KYpusoNhcc8iE+DcreDCp4AQt 4tY5wuhyxZxicTAYqNCTJBkMBkrMSZoHmJfiRCqcXEYOZma0LrJm4t9skfuXtrQu6Ok0WSg1L8o4 wzBMkiRJEj2CnB+4FEYOZma6LrIG4t9skfuX5jug3NY+zmNkBzJDWZgS/2YF9++yv3I7LmaxOcqI fx3tL4xTRvy7+75fxdovST2XuigDshCQhYAsBGQhIAsBWQjIQkAWAny/Any/Any/Any/Any/Any/ Any/Any/Any/QuW64B8kqrfw/RKt8/3yj/bVa/h+BeX7dUpemgO+X8Gm79e13sSa84GIzs/PnSqQ uXut6XR6dHTUmtT0y5cv7ghuqej73dB5sQz4fvPA9zsHfL8CfL/Cbvp+a7NTvl+lJKx9w74jvl81 hfXHNTZlB32/1ED5u2u+3zRNrffE9TDv+03TtF5nYd9xa9b326Qo7Gdh0PerX1nqXWjt+n5f0JPv t/zx5FNDXUEXXj5qfMZzwvebJMnl5SV8v10iCsMQvl+C71dn7hlf+5/c+dnn/f19a2tcgX3fryNB UHFsubU43Dk1FPD9CvD9Cm34ftXgELkte23k+12WzvPy/XKx8Gv4fvM4/tu7MuA70AKyEJCFgCwE ZCEgCwFZCMhCQBYCshCQhYAsBGQhIAsBWQjIQkAWArIQkIWALARkISALwWYWnXlWC+KKKKOl3ojI e+fnKYPlulBjE+PxuNJ252ydejukRRCGYYVmTQ201BgrokWS1qKSkj2dlfy9NG/rXGtofFyq6g4s o2EWSlKr25p1wa3ScOZUyNm6LHIm5VXbU3UHllEvC1UFakpuhuJ0KuHvXdbOaiz3FzlnN61TbG4U t66pbDHm17neccXEHOzNzjVbavVliqcMVc8R1RcuM/Hq0/W9yq1OtaNmyPWU5TdsO75zUDxmZfy9 5Wdj3DpHylPG38t/rfB9oMrHaAmbq4vm/7FCSbagLrjIW7i+bEEWTAtffduaLFoAWQjIQkAWArIQ kIWALATDjlt6eriwjRiuC913u3XA9yvA9yvA9yvA9yvA9yvA9yvA9yuY9/2S9gu8rfj5oaKm73ft b1Vzvt8gCNz/WWJLvl/H7bYMfL8CfL8CfL8CfL8CfL8CfL8CfL8CfL8CfL8CfL8CfL+Ced8vaaFU Ygd9v3q91GjW4gMhw75ftnhaVD42wbDv1/f9yWRSLwv7jtuvX79WXWaF77cJ9rN4+/atKd9vQ9I0 tev77WRZVuMGXL9SFH2/cRzXyMj6bTh8vwJ8vwJ8vwJ8vwJ8vwJ8vwJ8v8KqcbPj4+P//vuP1v33 oOXhjyrOjieuH08djUZJkhjx/QZBcHZ2Vms726Cy75e0++6F6eijyvQcfL9cLPx6oe/X8RJYjKkf KMH3u1MgCwFZCMhCQBYCshCQhYAsBGQhIAsBWQjIQkAWArIQkIWALARkISALAVkIyEJAFgKyECxn 0UT5y/LawWCwopFK3lzL4yNUS8CaFaSDasFcI+zHLbsxJedbS/Ms1NuctnKt7zf3Nven7fD9ZqWV v7TS96vmVH/V2Q7fb1Za+Vs87MvqYqHUcWuyKE4pdgTls1jWZpmNce6aulr5W8b3u7DNUvOVOnwl qFEX5ZW/+l4VV6crf4tdxtZcR0pSPGbw/Qrw/W6WLagL+H7zwPfbKshCQBYCshCQhYAsBGQhwPcr wPcrwPcrwPcrwPcrwPcrwPcrwPcrwPcrNPL95n6WSE/KFPh+59hZ3y/b4eD7he93k75fIoqiCL7f OeD7FeD7nQO+3zxRFJ2cnFg/WeD7FRzy/fb7fYsbQE75fu1uAMH3q9Ot7fv1PI/lnUWdaT3HLRGd np7WW9AIxny/xddVse777dY+QXTrsz69Sdejq3fax6TvN01T631wE4z5fpv3oPYdt6Z8v+x19TyP z50a3af1LDpZlnU6nUqKcH4Moe4IPM/LLV6cUgbP8/79+7e/v191QVN0iSgIgkoV7vs+XziZGru9 kCAILAZB/Bx8OBxeXl5W6vZW73+NdNj3W3Ups8D3K8D3K8D3K8D3Kyx+9jubzTb9IdqdU0OxYJz9 9vZ2b2+PP4ZuYpV8S/bjxw+ngiD4fnXg+xU26/tVw+6uDZEtZFO+X6cGxEryPzPJ1rdjRU6GAAAA AElFTkSuQmCC ------=_NextPart_000_0000_01C80F9B.F234E740 Content-Type: image/png Content-Transfer-Encoding: base64 Content-Location: http://www.cs.wisc.edu/wisa/papers/security03/images/cih-vir-auto.png iVBORw0KGgoAAAANSUhEUgAAAiUAAAO9CAIAAADSTYAmAAAACXBIWXMAAAxNAAAMTQHSzq1OAAAA HXRFWHRTb2Z0d2FyZQBHTlUgR2hvc3RzY3JpcHQgNy4wNULvQ9UAACAASURBVHic7N0xjNvonffx h0agYndwwCBXWlMECNIHKmxjMRXTSM0aAVnsuphc4SsWKRZuWAXYrdQYixeBm1Qqxg0fBJNG05jN TWG7GAS4YpEiuAViG3vFFVOccwuo4Vv8d559TFISSVEPRfL7KQYURVKcmYf88Xn48JGXpqkCgEEK w1AmtNZKqSAI5GUcx63tU3955A2A4bADRtIlCAITMzJfsmfdAqiNvAEwCCZIKuWHWYsaz+7IGwD9 J9Wa2jUVrXUYhnEcU9HZBXkDoM8arKBIaFHRqY28AdBbO1Zr8qjo7IK8AdBP+6uOUNGp507bOwAA zQvDMAiCPUWCbFaa6VDez9reAQBoWOZhmn2I49jzaB+qhvoNgL7RWjto7Irj2DzNgzLIGwC9Ii1p Dj5IPoVWtfJoTwPQHw5a0my0qlXCXwpAf7g/+8vzPfRVK4P2NAA9IU/GOP5QWtXKo34DoCfaatqi ilMS9RsAPdHWM/+MNVASeQOgD9pt0aI9rQzyBkAftHvGp4pTBnkDoCdq3EExX622+6fz7OdW5A2A PqiaGWEYep4XhmEYhrvnDfWbMnjeE0AfVD3ja61NZ7ZG8oZbOFtRvwHQefXO9Z7nyUQjtRPyZivy BkDn1TjXS+XG87ymcoImta3IGwCdV+9ZyzRNgyCQWzi77wP1m63IGwB9UOl0bxaO4ziO40aigvrN VuQNgD6oero3dZpGcoLO0GWQNwB6onw1RbqTebcaGXWN+s1W5A2AnqjULJZaGvl08mYr8gZAH+w+ PLOXUz7A6CxQBs97AuiJHU/6u1R0qNyUQf0GQE+0ddKns0BJ5A2AnpCHaVr5aL5srQzyBkBPtPLV zmEY0phWEnkDoD/k4U1nkSMfRN6URN4A6BWXQzVrrWlJK4+8AdArUttwcCOHlrSqyBsAfSN1jr3W cmhJq4G8AdBD0qq2p8iRLdOSVlUzAwcBwKExedNsMEhLHWFTA/UbAP0UBIGkQlP3cmSIT7NZVEX9 BkDPSUUnCIJdbrdQrdkdeQNgEGoHhtY6DMM4jukdsCPyBsBQ6FuSHBtqPKYJThamWtMI8gbAEGUS xcwvk0aoh7wBgB8f3iRg9or+aQCgFA9v7h95AwBwgbwBALhA3gAAXCBvAAAukDcAABfIGwCAC+QN AMAF8gYA4AJ5AwBwgbwBALhA3gAAXCBvAAAukDcAABfIGwCAC+QNAMAF8gYA4AJ5AwBwgbwBALhA 3gAAXCBvAAAukDcAABfIGwCAC+QNAMAF8gbAoGmtT05OlsvlycnJycnJgwcP2t6j3vLSNG17HwCg TZ7nmen79++/fPmyxZ3pMeo3APCTL7/8su1d6C3qNwCG7uTk5O3btzLNKXF/qN8AGLq7d+/KxHg8 bndP+o28ATB0pg3t3r177e5Jv9GeBgDK87zxePzmzZu2d6TPqN8AAFwgbwBAjcdjcxcHe/KztncA AFybzWZm+vLy0vf9O3fufPzxx/IgznQ6lbdGo9HFxUU7u9hH3L8BMAgPHz78y1/+4vu+mTOfzzcs H0WRmR6NRkEQnJ2d7W/3hoC8AdBnNzc3jx49Wq1WalvAbBZFUZIk0+l0uVxuXTgMwyAIgiCo/XG9 RN4A6KfFYqG1Xq1Wvu/b1ZpdSKVnNBrlU0dr/c0337x7904eHQ2CII7jRj60N8gbAH2jtV4sFs0m jS1JkiRJJHW01k+ePDHDExjj8Tj/NM/AE4i8AdArUq35+uuv9/1BSZLY93gyMvUbrbXW2rw1zKY2 8gZAfywWi+fPn+9yn6aSJEmurq6kbc0kyuvXr5VST58+XRcqYRiqQdZ1yBsAPeE4bIS0rb148aLS WhJOQ4sc8gZAH7QSNsKu5ZQ3wFoOeQOg86SDgIN7NuvUixzPG9YZmPFsAHTeYrE4PT1tcQd831+t VqZHQElBEFRdpdPIGwCdJ2PStLsPvu8vFouqa5E3ANAZNzc3tcMmSRLzc0e+719eXu6+nYO1ezQy XieAbnv06FGNtWR8GqWUZJX8nEwm8q7v+zW6Hvi+f3Nzc3x8XHL5OI6l14BLYRhKclS6daS1DsNw x7tN1G8AdF6NbEiS5Pr6+vr6ej6fS9gkSeL7vsxUH47XWV698HMpjuMasdHIA6rkDYBuk7E4a7Br M0qpKIpMbs3n8xqNbK30xt6F1Fq01vJFDPmXnufJdH5F81YYhp7nSUVNpmUZmba3TN4AGCKpxEwm k0Zu3nSU6SAnjwHZL02bWxzHmcjJvGXWlZ+m8qS1TtM0TVPP82TL3L8B0G2j0ajeitfX11EURVFU 725NXhRFtXemRZm2MnlpbvPk2WPBCcknCRWztTRNM4tRvwHQeVXvtZg6zXw+N01nvu/vXtc5Pz/f cQuHQ+70iM1vSdLYYaNuB1Cw55A3ALqt3ineRJTpS+37vplp38uppHznNHXYD98EQWD6zmX2s/At mWnSRapHmZoTeQOg246Pj6u2YklVZnJL7uXIl+XIHGXlUEnyjTiVVmkrb+RzPc8zE5n56vZLE+wb /uZn5i0Rx7GdLjJtPuLJkyeK8dMA9MDDhw/fv3/fbvewKIoeP35cqd/w0IbspH4DoPMuLi5Go1GL Pc2kclPjIZXhhI0ibwD0w3K5lK+icf/R9QaHHiDyBkBPxHHsPnJqh437kWxaR94A6Inj42PHkVM7 bOynLIeD/gIA+mY2m52enu77GwqSJImiSO7ZVEqOoXUTMKjfAOib5XJ5dXVlRoBuXBRFk8nk9evX aZqap/HF5i7OZnixAYaNon4DoMdms5mM5tlUV2nJsOl0WtiGZj8FmemrZuYMM2kEeQOg52azmXwB aO3UkXtCo9FoOp1+8cUXze7ecJA3AAZhsVhorTPfPL0ugewB2aRCc3Z21sh3wAwZeQNgcLTWi8XC vLy8vPz1r3/917/+VSk1nU7NfB6paRZ5AwAqDMMh31lxg/5pAAAXyBsAgAvkDQDABfIGAOACeQMA cIG8AQC4QN4AAFwgbwAALpA3AAAXyBsAgAvkDQDABfIGAOACeQMAcIG8AQC4QN4AAFwgbwAALpA3 AAAXyBsAgAvkDQDABfIGAOACeQMAcIG8AYCe01rXeKtx5A0ANCYMQ8/zPM/LzC+cWW/7W+eUWUto rTNvldlabeQNADQmjmOZsOsNWusgCHbfeH47+la9Deb3KgiC/UUOeQMADYvj2D5ra61NDimrDmSW kTnyUqZlrUytKAzDfELEcbwub+zNqtvajGx23Z7L9vfUyEbeAEDD7LN25twtmZGmaZqmpjlL0kjW kndlrTRN4zheFw+yTBAEhVUcewtmr2SmHX6FO7+vKk4KAINnMmB3cl6V8LC3LC8zZ13zMo5jWdL8 zJ+oM+uaLQdBkN9/e+F1Wyics27m7qjfAEDzpA5RWFEobK0y1RSTNHEc2wFQuB1pcNvxLo4z5A0A 7IVESKbxyjRqqdz9f2nIMq1q9n2X/MbDMLSrDvZmzdZq94hrpHdDgX1UmgCgW5pqTzPpktmszJcq izmb5z/UnpOv6GS2lnlpljQ7YKJLfmb2Lb+3zf4p8rx0TU0NAIYjDMPNd9EPgekgsHXJ2r9O+Y+o gfY0AOiGkp2V8414Je01bJRS1G8AoBv1m66jfgMAcIG8AQC4QN4AAFwgbwAALpA3AAAXyBsAgAvk DQDABfIGAOACeQMAcIG8ATBoWuuTk5PlcnlycnJycvLgwYO296i3GM8GwNDZ4/bfv3//5cuXLe5M j1G/AYCffPnll23vQm9RvwEwdCcnJ2/fvpVpTon7Q/0GwNDdvXtXJsbjcbt70m/kDYChM21o9+7d a3dP+o32NABQnueNx+M3b960vSN9Rv0GAOACeQMAajwem7s42JOftb0DAOCU1nqxWJiXl5eXvu/f uXPn448/lgdxptOpvDUajS4uLlrZyV7i/g2AQZjNZhItZs58Pt+wfBRFZno0GpE9uyNvAPTZYrHQ Wq9WK9/37bCpKoqiJEmm0+lyudy6cBiGQRAEQVD743qJvAHQT00ljU0qPaPRKJ86Wutvvvnm1atX 8jIIgjiOG/nQ3iBvAPSN3KFpNmlsSZIkSSKpE4bh69evzfAExng8zj/NM/AEIm8A9IpUa77++ut9 f1CSJPY9nox8/SYMQ/PWMJvayBsA/bFYLJ4/f765I0CDkiS5uro6Pz8/Pj7WWmutlVKvX79WSj19 +nRdqEjwDLCuQ94A6AnHYSOkbe3FixeV1pJwGlrkkDcA+qCVsBFSyynTb802wMghbwB0nnQQcHDP Zp16keN5wzoDM54NgM5bLBanp6ct7oDv+6vVSu7flBcEQdVVOo28AdB5mYEDWuH7vj1MTknkDQB0 xs3NTe2wSZLE/MzMrMr3/cvLy3q70Qm7RyN5A6DbHj16VGOtKIomk0kURTJQjZm/+amazXzfv7m5 Kb98K50FwjD0PE9GJi1Pa22eH6qN8aEBdF6NbmlJklxfX5tpM3/HdrlHjx5V7TXgmIRc1bxp5AFV 6jcAum21WtVbcTKZyERT935a6Y29C6m1aK0lfvIv19WE7LekwiS1H5mWZeyKlGyZvAEwRFK5mUwm 9e7W9IPpICeVHvulTKRpGsdxJnIyb5l15afp4a21TtM0TVPP82TLtKcB6LbRaFRvxevra7l/4/t+ I1WTKIpq70yLMm1l8lIqOoXLm5F7DMknCRWztTRNM4tRvwHQeVXv8Js6zXw+n8/nDVZxzs/Pm9pU 6+I4Tm9tfkuSxg4bdTtMnD2HvAHQbfVO8Saimn1w5/j4uPzCh/zwTRAEpkNaZj8L35KZJl2kepSp OZE3ALrt+Pi4aiuW7/tJkkxumY5q6rbqY7oSlFejMa2tvJHP9TzPTGTmq9svTbBv+JufmbdEHMd2 usi0+YgnT54oxk8D0AMPHz58//59u93Doih6/PhxpX7DQ/tiAuo3ADrv4uJiNBq12NNMvu6zxkMq wwkbRd4A6IflcilfReP+o+sNDj1A5A2Anojj2H3k1A6b3YeH6RzyBkBPHB8fx3F8dXXlLHJqh439 lOVw0F8AQN/MZrPT09N9f0OBjOwp92wqJcfQugkY1G8A9M1yuby6usoM/NwgGVv69evXMqaLUiq8 tbmLsxlebIBho6jfAOix2WwmX8XWVFdpeUr06Ojo4uIi/679FGSmr5rMEY3sSReRNwB6bvfUkW4I o9FoOp1+8cUXze7ecJA3AAZhsVhorTPfPF2YQJlObkmSTKfTs7OzIVdNGkHeABii2Wxmpi8vL3/9 61//9a9/VUpNp1Mzn0dqmkXeAIAKw3CY9/Bdon8aAMAF8gYA4AJ5AwBwgbwBALhA3gAAXCBvAAAu kDcAABfIGwCAC+QNAMAF8gYA4AJ5AwBwgbwBALhA3gAAXCBvAAAukDcAABfIGwCAC+QNAMAF8gYA 4AJ5AwBwgbwBALhA3gAAXCBvAABZWusdF8gjbwCgAWEYep7neV5mfuHMetu3P8UoPO/LW7JKZgtl aK3thc1nmZmZBVTu1y/8LPIGABoQx7FM2AGgtQ6CYPeNm+3EcZymqVIqTdM0TeM4Xpci8q69BVG4 8cwce5/lo+Xj1G2Q5H8ps2NmCwU7lgLA4JlT6i7UbepkNpuZY87I5qU9nVo5YW8580Hr3lo3M47j OI4Lf83NW1j30eaXtbeZ+U0lhAzqNwDQGMkMqTHk6w1hGJqzszRJmVO2uj1xy1qSEBsa4sJbqVWr WEe2GQTBuipObWaH8wqqOPlYA4Chaap+k1rVFLsZKjNhvzTVDruWkzlFF65of1bhnhhmT4IgsKfz CZHfwrqPLlxg3cLCS0tkIwD0m6lq7MLzfjyjep5nbrdk5ku65Jc3FR2pA2XCwCyZf5l5a93y9rv2 PmzdQmZ/7H3OTGzeT0V/AQBonLRc5QNM5su03ZXAxIz6sBmqkbYvaXOz6zqVNmvvT9VUzlah8hUx ABia3dvTMjf5zQZlvrlznukvULgDZhl7rfwHyUbsVjuVa+PK3NI369p38vNBkP9dMvts/1L21tSH /QUym6U9DQCaaU/bH3PDf/Ni5rcobCJzwHxu4Q7TngYAh87u9rZOpgUv/7znXsnznmZPVFE6Ur8B gEOv3/QD9RsAgAvkDQDABfIGAOACeQMAcIG8AQC4QN4AAFwgbwAALpA3AAAXyBsAgAvkDYBB01qf nJwsl8uTk5OTk5MHDx60vUe9xXg2AIbO/nqY+/fvv3z5ssWd6THqNwDwky+//LLtXegt6jcAhu7k 5OTt27cyzSlxf6jfABi6u3fvysR4PG53T/qNvAEwdKYN7d69e+3uSb/RngYAyvO88Xj85s2btnek z6jfAABcIG8AQI3HY3MXB3tCexqAYdFaLxYLmb68vFRK+b7/97///Ze//KVSKkmS6XQq745Go4uL i5Z2s4fIGwCDMJvNLi8vfd+Xl77vm+lCURSZ6dFoRPbsjrwB0GeLxUJrvVqttgbMZlEUSdVnuVxu XTgMwyAIgiCo/XG9RN4A6KemksYmlZ7RaHR2dpaJE631N9988+rVK3kZBEEcx418aG+QNwD6Ru7Q NJs0tiRJkiQZjUbL5TIMw9evX5vhCYzxeCxP82itTTINPIHIGwC9ItWar7/+et8flCSJVHfG43E+ b/L1mzAMzVvDbGojbwD0x2KxeP78+Xw+d/NxSZJcXV2dn58fHx+r20R59+7du3fvnj59ui5UZLEB 1nXIGwA94ThshLStvXjxotJaWmut9dAih7wB0AethI2QWk6Zfmu2AUYOeQOg86SDgIN7NuvUixzP G9YZmPFsAHTeYrE4PT1tcQd831+tVmbYgpKCINBa72ePDhF5A6Dz7IED2uL7fo3wIG8AoDNubm5q h02SJObnjnzfl9HY+mr3aPxZI/sBAG159OhRjbVkfBqllGSV/JxMJvKu7/s1uh74vn9zcyN9o8uI 49g8lONMGIaSHJVuHWmtwzDc8W4TeQOg82pkQ5Ik19fXZtqeYx7krOHRo0dVew04Jj3iPM+rtFYj D6jSngag21arVb0V7dqM+WlPVNVKb+xdSK1Fay3xk3/peV5hMtlvhWHoeZ5U1GRalpFpe8vkDYBu q3f3Raoyk8kkv3qSJK33PnDDdJCTSo/90rS5xXGciZzMW2Zd+Wna3LTWaZqmaep5nmyZ9jQA3Wa+ Hq2q6+vrKIqiKMrcrUmSpF5NJYqi0WhUb2dalGkrk5fmNk+ePKlqz5F8klAxW0vTNLMY9RsAnVf1 doup08zn8/l8bldxdqzcnJ+f11730MRxnN7a/JYkjR026naYOHsOeQOg2+qd4k1E2eli91ir10xX vnOaOuyHb4IgMH3nMvtZ+JbMNOki1aNMzYm8AdBtx8fHVVuxfN9PkmRyy+6WFkWRzKyaNzUa09rK G/lcz/PMRGa+uv3SBPuGv/mZeUvEcWyni0ybj3jy5Ili/DQAPfDw4cP379+32z0siqLHjx9X6jc8 tC8moH4DoPMuLi5Go1EjwwTUI1/3WeMhleGEjSJvAPTDcrm8urpqJXLqDQ49QOQNgJ44Pz+Xbz9z +aG1w8b9SDatI28A9MTx8XEcxy5rObXDxn7KcjjoLwCgb2az2enp6b7HCJARP+WeTaXkGFo3AYP6 DYC+kXs5ZgToxkmf6aOjIxnTRSkV3trcxdkMLzbAsFHUbwD02Gw2k69ia6qrtDwlenR0dHFxkX/X DPSSf9RR5ohG9qSLyBsAPbd76kg3BOnxfHZ21ujeDQh5A2AQFouF1jrzzdOFCZTp5JYkyXQ6PTs7 G3LVpBHkDYAhms1mZtr+Hmgz2jRVmcaRNwCgwjAc5j18l+ifBgBwgbwBALhA3gAAXCBvAAAukDcA ABfIGwCAC+QNAMAF8gYA4AJ5AwBwgbwBALhA3gAAXCBvAAAukDcAABfIGwCAC+QNAMAF8gYA4AJ5 AwBwgbwBALhA3gAAXCBvAAAukDcAABfIGwCAC+QNANQXhqHneZ7nZeYXzqy3fXuDhpmf+cTM/PxL e8fyG9kr8gYA6ovjWCa01mam1joIgt03ntlOaincfpqmZn9kdWHvbZqm5mUQBC4jh7wBgF3FcWyf uLXW9nnf1CrMMjJHXsq0rJWpFYVhWJgr5fMsjmM7bzJkIxsWaBZ5AwC7sk/cmdO3ZIZUSrTWkjGS RrKWvCtrSQVlXUOc3Zi2NSRkgSAIMlWc/J67q+KkADB4JhJqkBOphIe9KXmZOc2al3Ecy5LmZ/7M bK9rT8dxLC1jhVvO/EZBEGR+u3W7tG9earXlAcAwhWFot4BV4nk/nkg9z5PYkE3JfM/zJFoyC8u0 qehINSiTOpmFN5+uMwtkKknr9qHMlptCexoANEMarzK5JTNlOnPfxcSM+rBdq8wNlc2tZGEY2hUL ex8Kd3vrxzXDTTUKAA5Z7fY0ky6Z7ch8afUyJ/T8p9hzzGKmrSyztXVtburDhjuzcGYPzWbtM/8u DYlV0Z4GADu1p+2Puee/eTHZ+fLNYmbJkttvCu1pAHCgyvRXtlvwCp8DtZm+18p52CilqN8AwIHW b3qG+g0AwAXyBgDgAnkDAHCBvAEAuEDeAABcIG8AAC6QNwAAF8gbAIAL5A0AwAXGFwAwaPnvLuOs uCfkDYChs78q5v79+y9fvmxxZ3qM9jQA+MmXX37Z9i70FvUbAEN3cnLy9u1bmeaUuD/UbwAM3d27 d2ViPB63uyf9Rt4AGDrThnbv3r1296TfaE8DAOV53ng8fvPmTds70mfUbwAALpA3AKDG47G5i4M9 oT0NwLBorReLhUxfXl4qpXzf//vf//7LX/5SKZUkyXQ6NQufnZ0FQdDGbvYQeQNgEGaz2eXlpe/7 8tL3fTNdKIoiMz0ajUaj0cXFxX53se/IGwB9tlgstNar1WprwGwWRZFUfc7Pz4+PjzcvHIZhEARU jDLIGwD91FTS2JIkSZJkNBrl29m01t98882rV6/kZRAEcRw38qG9Qd4A6Bu5Q7Narebz+T62b1Jn uVyGYfj69WszPIExHo/laR6ttUmmgScQeQOgV6Ra8/XXX+/7g5Ikse/xZOTrN2EYmreG2dRG3gDo j8Vi8fz58z1Va/KSJLm6ujJ3dCRR3r179+7du6dPn64LFVlsgHUd8gZATzgOGyFtay9evKi0ltZa az20yCFvAPRBK2EjpJazXC4rrTXAyCFvAHSedBBwcM9mnXqR43nDOgMzng2AzlssFqenpy3ugO/7 q9XKDFtQUhAEma+y7jfyBkDn2QMHtMX3/RrhQd4AQGd89913tcMmSRLzc0e+78tobH21ezT+rJH9 AIC2/P73v6+xloxPo5SSrJKfk8lE3vV9v0bXA9/3b25uto52Y8RxbB7KcSYMQ0mOSreOtNZhGO54 t4m8AdB5NbIhSZLr62sznZljgqeqR48eVe014Jj0iPM8r9JajTygSnsagG5brVb1VrRrM+anUipJ knr9qlvpjb0LqbVorSV+8i89zytMJvutMAw9z5OKmkzLMjJtb5m8AdBt9e6+SFVmMplkVpdRahq5 o3P4TAc5qfTYL02bWxzHmcjJvGXWlZ+mzU1rnaZpmqae58mWaU8D0G3216NVcn19HUVRFEX23Rrf 96+vr+u1p0VRNBqN6u1MizJtZfLS3ObJkydV7TmSTxIqZmtpmmYWo34DoPM2jJtZyFRf5vP5fD5v sDZzfn7e1KZaF8dxemvzW5I0dtio22Hi7DnkDYBuq3eKNxHV7IM75TunqcN++CYIAtN3LrOfhW/J TJMuUj3K1JzIGwDddnx8XLUVy/f9JEkmt+Rejj2nxp3/Go1pbeWNfK7neWYiM1/dfmmCfcPf/My8 JeI4ttNFps1HPHnyRDF+GoAeePjw4fv379vtHhZF0ePHjyv1Gx7aFxNQvwHQeRcXF6PRqMVOZUmS HB0d1XhIZThho8gbAP2wXC6vrq5aiRwZHPri4sL9R3cLeQOgJ87Pz+Xbz1x+aL1vIlBK2XfXB4K8 AdAT//7v/350dOSyllM7bMyN9z3s1OEibwB0nhlG5eLiwlnDmjwr+vHHH1cdc1P6Cg/qzo0gbwB0 mEkaGV5FZkrk7G9Ymn/7t3+bTCZHR0fmQ8NbG7o4yxhisrcDDBtFf2gAHWWeKNxw7p7NZvJVbI10 lTY3h/76179+/PHHT58+zTSImYFeMo862o+tDK0NzUbeAOiYMkljm81mSqnValU7dSRpRqNREARn Z2fq9qHI8XicTx2sQ94A6IyqSWPTWi8Wi8w3TxcmkN3JTSam0+nZ2ZmdKw8ePHj16pVMj8fjN2/e VN2fASJvAHTALklTSCo9wv4eaDPatKnKrHNycvL27VuZHo/Hd+/effnyZSP71lfkDYCD1njSNEXu /9tzqOhsRv80AAeqsO/Z4QiCYDwem5dyL6fF/Tl8fN8agINjOnodeAPM06dPpYrz85///O7du3Qc 2Iz2NAAHxCTNAVZoCp2cnCilpGYzzKc4y6N+A+AgdC5pxJs3b+ynbcIw7Nb+u0T9BkDLOpo0heR3 6cEvsg/kDYDW9ClpDCJnHfIGQAt6mTQGkVOIvAHgVL+TxiBy8sgbAI4MJGkMIieDvAGwd0NLGhs9 1gzyBsAemaQZ7FD81HIM8gbAXpA0BpEjyBsADSNp8ogcRd4AaBBJswGRQ94AaABJU8bAI4e8AbAr GSOZpCljyJFD3gCoj6SpYbCRQ94AqIOk2YX89YYWOeQNgGpImkaEYTi0vyF5A6AskqZZQ4sc8gbA diTNngwqcsgbAJuEYSg3tzt0TrS/cPPwed5QzsN32t4BAAcqDEPP85RSaZru7/TtfUge4ilPKl5m WjYS3sovc4DiOD7wPWwKeQMgy04aB32o0ltVz7x2PUayyt6UWSwIgkM+oUt72iHvYVPIGwA/cZw0 GSY87JqKzNFaS8uezDeL2RWvTKuU2X9ZpmrNyaXDnBPnJwAAIABJREFU38NGkDcAlGo7acw+pGkq KSLVFIkZpVQQBHI6lsa9qrWBw69AxHFsxgTqq6HcpwKwjtQbgiBoJWY8z7OzROXun5uX9nyZzs/Z /EGHf7rr95ezUb8BhsvUHlqs0yil4jiW2oy9Y23tTLsOvx62C/IGGCJzOySO40O7oDbVHbWmZ3Ph zPyZOhNanegh3e8bOeQNMCwHlTRyYrX7Bajb+/yme7S9k5mZdoTYa2U2qDrVTiU3ctrei73oQIMm gEaY29FdOfNm5G/AmG/c2bxiycUOR18HkCZvgP7retKIwhv+W7Okc2EjejnODXkD9Fk/kkbddqIb 1PmqEx3qKunb7wNA9CZpbN0aGG1H/WtVI2+AvjFJ078GmaHpWavaz9reAQCNIWl6RrqG9+ZfSf0G 6AOSpq/6VMWhfgN0G0nTb32q4lC/AbqKpBmI3lRxqN8A3UPSDEpvqjjUb4AuIWmGqR/P4vThdwCG gKQZMhkOruvP4tCeBhw6kgZVv2n7MFG/AQ4XSQOjB01q1G+AQ9TL0Wiwix70GiBvgMNC0mAd8gZA M0gabNCDWzidbxAEeoCkQRldv4VD/QZoE0mD8jrdmKbIG6AtJA1qCMOwuwWGvAFck6+qDIKguycO oAbyBsOltV4sFjJ9eXmplPJ937ybJIlSajqdysuzs7PdWzNM0nS6FR5t6XqXgW7ffQJqmM1ml5eX Jlp837djJi9JEskemf70008vLi7KfJDWOgzDf/3Xf/2f//kf6jRoRKfb08gbDMVisdBar1arrQGz WRRFSZJMp9Pz8/Pj4+N1i/3Lv/zL//7v/5qXJA0a0ekuah3edaCkppLGJpWe0WiUaWfTWj958uTt 27f2wuPx+M2bN418LgaOvAEOlNyhWa1W8/l8H9s3qbNcLh88ePDq1at1S2YyKbi1j71Cj5E3wCGS as3XX3+97w9KkuSrr7764YcfCt9dV7mRG7+kDirp9P0b+qehnxaLxfPnz/dUrcmQNrqrq6v//M// fPDggelE9O7du3fv3q2LkziO7eGfHewn0C7qN+ghl2FjSNtaHMcbOhEU6nQLCRzrdP3mTts7ADSs lbBRt/2qHz16VHXFrj9UAZRE3qBXpIXKfdgI3/dPT09ns1mltWhMw0CQN+iVxWJxenra4g74vr9a rcywBeXJjRygx8gb9Io8ZNPuPvi+XyM8yBv0HnmD/ri5uam9roxYY8at2YXv+zIaW3ndalIjGlEP eYP+qHGvXikVRdFkMomiSAaqsd9KkmQymdTYpu/7z549K798K3kThqHneZ7nVVpLBoXb0y6Jqru0 lb3DdM1oEXmDXqnRUyBJkuvr6+vr6/l8nh8fup4aVRz34jiu0Q97T9G4vxiQ0RzMyyAIiJy2kDfo j9VqVW9FU4nJ3PupfSuo9XtIVUmtRWstdYv8y3U1ocxbUmeSE7pMqw8rUvaWzXnf/iybWctsyl7d 8zx5KRP5d83GM3mjaBJsCXmD/qhXI7m+vlZKTSYTe/UkSTqXGbsIgsD+slH7pUykaRrHcSYS8m+Z 1eWn1J+01mmapmnqeZ7Zcpqm5qQva+UrW2aOecLRXt28laapGRzIvLuhHkMVpy082IxOMiPB2P7r v/4rDMN6OSE3b3zfn8/nEjyynclkIoFUdWsyiGf5Vdp6btyMbpAZ5sC8lBqDmS+xse4tdfuvkYdY zW8ki4VhaK9uf+LW6Q0Tm2cWDt/Q3TEdOj2+AOOn4eAUZomZk79/YA6/2WxWtV5ilpeYiaJIffgF a6pu5Jyfn1dd5WDFcbxhFLjMW1J7sO+aSItWt/rgYR/IG7Qgnyjrhq2sdCl3fn5eo50kiiLpZWCC yu50UC9slFJVR1E7WJIfpu6Sv/eef8ueL9lT8v+Y2f4uNm+K8GsFeYO9KJkozbYMHB8fj0ajSqv4 vi/9oeVlvWjJkG/EqbRKW7cT5J/ieZ78I0wrk5kvN0KCIDC3Z8ydefMVPvadG2GPCCf3VMxHPHny RFlhYKeC3N3J396X+cq682+3nuWLkOyP+ULVfLR0ukmq07raiIlDkAkVB4my1cOHD9+/f9/W+Gki iqLHjx9XuoKWszMnwd3lb8xkimXXvwCi02FJ/Qbbbc2VwzkALi4uatzFaZBUbmqczg7nb9gzUsGS ilTXw6bryBt8wI6WzMHZlVu+y+VSRmh2HzlJklxdXVXqloYGbegAnZlAK8ibQTPHZ0ejZR3TccBl 5NQOG54FaYrUERvsdIBmkTcDYqdLb6Kl0PHxcRzHMpyam8ipHTb2U5ZoRP/Kc2+QN31WGDADObUd Hx87a1iLoujbb7/95JNPqt7LpZsABoW86ZV8wPSy+lLecrl8+PChPLy5j05rMirBp59++v333yul wjC0H3VcFyT6lumzCwwBedNt5vY+AbPOxcWFUmo2m00mExmuZvdtmgEIjo6O7N639vAtZuxI+99h 7pORNBggnr/pJKnHbL2ORoY0r8l3gNZrZJOkkR7PZ2dnDe8fsA3P38AFMmZ3cj9fa71YLKIosiOn MIHsgdRkYjqdVn2WE4CgfnPQMg9akjH78OzZM/PdaPaXpE2nU5mgKoPD0en6DXlziOz7yYqYcavT xzN6r9Plk/a0A2LHDPeTAfQMedM+YgbAEJA3rSFmAAwKedMC+6tBiBkAA0HeuGMPy0jMABga8mbv GLwEABR5s1e0mwGAQd40z35Ik5gBAEHeNMmu0DDkCQDYyJtmmJGAqdAAQCHyZiem6YykAYDNyJua aDoDgErIm8poOgOAGsibsuxeZwyqDQBVkTfb2TdpaDoDgHrIm01IGgBoCnlTjJs0ANCsO23vwMEJ w9DzPKVUmqaEDQA0hbz5CUkDAPtD3ihF0gDA/g09b7TWJA0AODDc/gKm7xkP05QnfSja3gs0if8p nBli/UZrLd3P4jg+/DqNVL+aIi2HZpvypxDeh+xvIzXvhhaZI5kt1USZznyETT5u3afU+F0yEx3S 4L91w/9UPqjwD57/n9oz+bdiL9KBkSdp4jhue0fK2sf/yGzT3njmg8yfKF9OgiBIb5sfzbRMbN7t zJ89s0olcq1gpnfZVEaDm9qg8X9r4f80/1L+aOv+p2l//6290em/yYDyRg6kDiWNsM8jMi2tH+nt AWmqaPLS/I6Zd/PbtFtR0g9PJfZfafOZUd0+DLtut+1dsmVSP7Of9s7bu2o+KH+6bOo/6zhvmvq3 Fv5P0zX/1q1pt49/69b/aX7HunVp6AZ50wGFB08nFFZB7AlzxZo/m5iZmV9/Xf3GVvLctC7F111u m7Xs/TFXteZkZHbenKTMTFkxf9neVI3Bff2mkX/rhvpN/t+69W+1p3/r5v9pfrMN/lt7o6PnMdH/ /2WmcHfO1hNTmcWqLmA3a2w+4M1IP+t2O3MBa9bKvLTPiRt20rzc8Ik7OpC8KbPY1lXSNf/WrX+r Pf1bt+7k/v6tvdHpvOl5fwHTL4AeOJWUHC/OHvJn3d3dNE0z73qel++mkWnBy39Q+Z3HOgf1b+V/ OkTuI86NrldrUuvWrpmj6l4IZxo6Nq9lVilsizM7Zm8z3/yiSl/Aml6C9sbtJe0dMB+a305T1337 vn7c07913V+48N+67n+a7vnfuvl/WridTl/O70On/yD9zJvu3q2xyaWo/YuYOerDb7BOby8eMwe5 uarItKqbhc278W2vpMJrEXu+LFx4JlLWPW2zBfu3KLy4Nqm2bgtp0b3lfNNNg3/2pja1bvuN/1sL /6dmfuG/Nf8/TdcETFP/1q3/03Sf/9be6PTfxEuLmi+6S57iLNlu0Hue17f/r2G+JyIzvbswDPPt QgeFf+uQHX753KBX4wtIa3J3/xkoLwgC+xtXOSv1A//WfutJ3lCtyev949nD/F/zb0V39SFvqNYU kj8Io2P1DP9WdFfn+0MTNptxVuol/q3oog7njQwmGPCVzwDQBV1tT5NqTV976QBA/3Qyb8IwpGsA AHRLx/KGfmgA0FFdyhsJG+7WAEAXdSZv6IcGAJ3WgbyhDQ0AeuDQ84awgTNaa6lGe54nc+gACTTo oPOGNjS4lLmmGY/Hbe0J0EuH+7wnYYN2PX36tO1dAHrlQPNGnrAhbOCYXaehCRdo1qHkzZMnT2QQ cmlD54YNWnHv3j2ZoDENaNxB5I10CgjD0DxhQ9igFabgmeAB0JSD+KLAk5OTt2/fKqU++uijf/7z n23vDgZNSuMhHBdAXqe/37P9+s2DBw8kbJRS//d///fgwYN29wcD9/btWxrTgH1oOW+ePHny6tUr 85LjHK0bj8d3795tey+AHqr//M3Dhw9Xq5VMX15eKqV83zfvJkni+/5oNJKXZ2dn+Vsy9heVK6Xu 37//8uXL2vsDVBWGoWm/NWX4V7/6lVLqN7/5TZIkSqnpdCoLFJZhAOVVu3+jtV4sFkopkzTz+Xzz KkmSyHGrlBqNRqPR6OLiQl6a2zbrkqbTLZU4WLPZTFll2Pd9+1Ipzy7DSZJ8+umnpgxvJVdUBBWa 0umzYtm8efbs2eXl5Wq12npwbhZFUZIk0+n05ubm3bt3T58+Laz3PHnyRCnFbVs0aLFYaK0bKcNK qdFodH5+fnx8vG6xBw8eSFvxeDx+8+ZN7Y8DbD3PG2k3W61WW6sy5Zkjdrlc2vMfPHjw7t07031A MYAVmtBU0tik0jMajTLtbHK1ZJdh8gYN6m3eLBaL3/3ud77vN5g0Njlij46OPvvsM1OhySxTWPsJ bu1jr9An0gLc7NWSzaTOcrk0FZqMjz76SFrw7L2iDKOefubNYrF4/vz5no5SW5IkX3311Q8//FD4 buHumY4G3f27w4EDKcPr6jcyQiCpg0o6nTfF/dOePXv2l7/8xcGBqm57tV1dXZm2NTkON5BDVIYk 6O6fHnvlLGzUbRlOkiSOY7mjY/e9XBcncRybxYgcDEFB/cZl2BhJktiRUxLfMI1CLsPGkLY1Eznl ed5BDPOBTuj0RXb2ec/FYuE+bJRSvu+fnp5mmrm3kqvCrfUhDEorYaNu+1U/evSo6opxHFOGMQQf 5E1bB6qoFzndjXrsg1R5u1WGaUzDQHyQNy0eqML3ffMUHlDDYrE4PT1tcQekDMtj0ZXYY20AvfRB 3hzIub7q5aHiWMWty8vLpp6wqc33/RoFkjKM3vspb2qc5Q0Z7cOM+bGL+XxeI/Y6dKx2aFc75+bm pmrYFBbdfEmuVLZ935fR2MrrVpMaZRj1fFC/qXFhGEXRZDKJokgGqjHzJ7ca2MeDFIah53me51Va S/pw72mXRNVd2srs8OHf0650r35d0U2SRMa/2DBnK9/3nz17Vn75VvJmgGVYdaEY99hPeVOvISJJ kuvr6+vr6/l8blaPomg+n8v8ertVqfm7lWM1juMafVj3sat7PX7kSXiZDoLg8I/V8jcgC4uuKrrq qnFc1KjiuHc4ZVjtsxjbZVh1pBj31Y95o7Wu3eptKjFmC3I9WLtyU7X5u/W2CPNl2J7n2dPm3XVX kfZbcrEpR4JMyzL2Rai9cbOk/Vm2zJYzq5tdlYn8u+bT7bxRh92WUrUlNl90m9L6PaSqMuU2/7JM GVZrivHmMqzWF+PGy7DqQjHusR/zpvYpW2owk8nEbpGQy0bf93vcnmaTwQ6UUnEc29PqtlinaRrH ceZwyrwly8s/IggCc+GptU7TNE1Tz/PMxtM0tT+l8CrVzDRdxu3VzVtpmpqBVcy7664BD/zasNJd lsKiO1iZcmu/LF+G7dWVVYw3l2G1vhjvowyrgy/GPfbTg82e59Vu/pIW8PzInpPJpMY2oyh6/Phx +QhscZQB82S4/Yi4PS1XW2Z5OeTk3fxb5hfJPEIsi4VhaK9e+NHr9mTzWhtm5h99P+SH4Wez2enp aaW6RWHRzZfbqiU5iqL88OebtfXc+NaCVKkMK+t4tH+jzWW4zHSDZTj/y3ZIT8YXmE6nVS/0zPLz +Xw+nzd1nZgkSetNZA2S6y+x+S25Oss0N5sLN5f73F0lC+Geiq7t/Py88W22pXwZVkXFmDIMsVPe qNtvslFFDdb1jmH5NrZKqxxyU6xdc8/sZ+FbMtM+UDPxs0GDf4cNH3rIZ41Kp/gNRbcRVUdRO1hV y7D6sBgfYBlWh12Me+ynvPniiy+qJoTv+0mSmK7P0uBg5kj/n6o7VK9y00rpkWPD3Ks09zPNtLod ytq+WWp+Zt4ScgfIvJRp8xHyLUHmmDQT9k3dzOqZngX2vslmM0e4uQ1rN8QbB16XPz4+Ho1GZZYs LLpCjgL71mN+zmbyjTgV9ru9TrqFZVh9WIyrlmH1YTEuWYbNMvsuw+rgi3GPfdCIWaP5u3F/+MMf qo4STelpSr5R25xWMtMH6+HDh+/fv293WKaqNyDVbd5QjHe3uQznX3ZOp093HzzvuVwur66u2toV pVQURTXKwSG3p3WdSZquHKUXFxej0ajFLmfylbU1/lDdPYkcOPtqqSvFuK+y37cWBIE8rel+V6Io +vzzz8/OziqtVb5pGJtt6DzqeE92tFwuZXAm9zX1el/jhKZsLcOdK8w9k/3+m7Ozs88//7zq6B27 i6Lo008/rRo2irxpjnQ06kdl8fz8XL79zOWH1g4bngVpSp/KcC9l80a1ETkSNl988UXVFakdN64f f8zj4+MXL15cXV05i5zaYWM/ZYlG9KMM91K2PU2cnZ19/PHHk8kkM7TUPsgDd8fHx//xH/9R6ajj Fis2c9awFkXRt99++8knn1S9l0sZxqBsech2NputViv5otxmP1iaO0ajURAE0oxmtypIP8vCFe17 11zIYCvpsaaqDOVZnlwtffrppxcXF+q2DJs23nVBom8FQUDYoJJO908rNajDbDaT0aMbOWLzSZNh EiVzb8a8JGlQVYtlWBV1xqUMo57+543Y/YjdepQCe/Xs2bPLy8tdquyUYbRrKHkjFouF1jrzZTmF R6/d40Bu25pmB6BFWuvFYiHfT2PKbWEZtju5ycR0Oj07O6NqgrYMK28y7G+hNl8wZQ+Adn5+3puB pNBLcgkl04VlmKoMDseg88bW6T8EICjGOGSdLp8Fz98AANA48gYA4AJ5AwBwgbwBALhA3gAAXCBv AAAukDcAABfIGwCAC+QNAMAF8gYA4AJ5AwBwgbwBALhA3gAAXGgyb8yg7gAAZDSZN3wJFQBgHdrT AAAu0J4GAHCB+g0AwAXyBgDgAv0FAAAuUL8BALhA3qxF9wcAaNAB5U0Yhp7neZ6XmV84s8bG7a0Z Zn5+N8JbmS0AeZlyVfViRUqXKXuF26EEousOKG/iOJYJ+xjTWu9+WyizkdSS2bgc4eZds0tKqSAI OOCxgV1sKhUVUz7jOE7T1Gwqsx1KIDovbU4QBDtuQd2mTmabZo6JB5kvL+3p1AoJe7OF03J4Z3Zg w+4FQZBfBT1TrxhnSo68zBTXNE3jOJZSZBfRwnUL36IEYvfTbIsOqH4j5BCVKk6mUSIMQ/O31lqH YSjHrZ06skqapnEcr2uF29CYtnXfuMDEVmEYpmmaL65KqSAITBHdUJxMW26apvZ8SiC6rangCoKg kfpNatVUzAblpSq6DJQLRrNwpn3MXjI/XbV+U2YBdF3t+o0pe2ZOZoH8/A0FO7UOhMJ3MUzUb5Rq 9OEbqbUUXsflb8PKBaN9h8ZOkc0fxANDaJApeGbOjl0c7duHQA8cXHuakBTJHG+mLUJ92AVAGhlM q5oJqhpHe769IrMRIgolrSuuRr2+MJRAdNcB5Y0cnHLTRVrJ7Pkmfkw/UZNG9sLSrJfpQp3ZVP7O jVne/oj8YuaOEWArLFfrimt+ph0hZlOytUy6UALRad7WRqfyDvZgkGN484Xh1p0vsxH0wL6Lsedl D7qSRYsSCHXAp9kyGqvfHHK3GbvPW6F8211+AcWhjv3YWj4VJRC98LMGt3XIqbv5QN16GHOcoxHr LssogRiCA7p/A/Se9GFjaD4ME9/vCbhGZQXDRP0GAOAC37cGAHCB+g0AwAXyBgDgAnkDAHCBvAEA uEDeAABcaGZ8ARns0gyR2eCYbIAz5lvRKMnAPjRfvxmPx41vE3Ag06H//v37be0J0EvN5I2dMU+f Pm1km0C7vvzyy7Z3AeiV5us3PPWJ7rKvnCjJQLOayZt79+7JBI1p6DRKMrA/zeSNuRI0hyvQRZRk YH8a+37Pk5OTt2/f0p8HXXdycqKUevPmTds7AhTg+z2VUurt27c0QaAH3r592/YuAP3UWN6Mx+O7 d+82tTWgLZRkYE8qP+8ZhuE///lPmb68vFRK+b6vlPrVr36lbh+Um06nssDHH3/c3aofeu/hw4er 1UqmTWGWkvyb3/wmSRJFYQaaU/b+zWw2u7y8lGhRSvm+b6bXiaJIJkajkVJquVyW3Cf5nlB6o2JP ZrOZUsokzXw+37x8kiSSPTL96aefXlxclPysMAyDIKAwoymdvn+zJW8Wi4XWWo7MrYflBnLEjkaj IAjOzs7WLfbgwYNXr14ppcbjMTds0SxTmMtcLW0gF1Kj0ej8/Pz4+LhwGa31N998Q2FG4/qZN00d nLZ1qaO1fvLkibJu1XKIokF7LcxnZ2d29UUKs93pQAqz1ppaDnbXw7yZzWbNHpwZURTJgWquATM+ +ugjafTI6+7fGu7JEJy+7+9SO9/ApM5yuTS184x1hZl2NtTQ6bwp6C8wm81OT0/3lDRCDv4wDL/7 7jul1Hg8znRC/fnPf174Nw3D0PO8OI45ULHVYrF4/vz59fX1/j5CrsmSJPnoo49++OGHwmU2FGap 9FCYMRAf5M3NzY1cDO41bIw4jqMo+vzzz6VtTWstPQXU+s4CctzKoPEcpdhAwmZP1ZoMOV6SJInj WO7olCzMZjEKM4bgp/Y0x2Fj2JFTHt1+sIHLsDGkbc1ETnmdbiGBY50uLT897/no0SP3YaOUms/n z58/XywWldaSa8O97BA6rpWwUbdta48ePaq6YhAEUmUH+u3HvHFwz2aD+Xxutz+UR+QgQwqS+7AR vu+fnp6u6+qyDtV0DMQddfuUdVthI1arVdUqjiJvkLNYLE5PT1vcAd/3KcxAoTtKqdVq1db1oFFj B7rbiIn9sUfBaIvv+1TWgbw76nbkqBpkkA8z1IeZmZlT0mq1urm5qbcnncAJZd9ubm6qhs26MrzL bvi+X/WY6laTGiUZ9dx59uxZjevBKIomk0kURVEU2QenvKx9gVnjXqtj8gCQDEtaiTx4uI9dEjV2 aSuzw125m12p/Kwrw0mSmKH/lFKTW/bMrXzff/bsWfnlW8mbeoV53yVZ7aEw2zvclcLcS3fqtT8k SXJ9fX19fT2fz83qckDWbpprvU2vjDiO632n3D5OKHs9cuzxVzrUgap8KSosw+r2YZr8MpUqPTWq OO7VK8x7isb9FbDMSEIdKsz9UzNvlFKTyUQmzOpyQFa9ErQd/iGaIdd6WmvP8+xpeWvdxWPmLbnM lGNAplXu2tPeuFnSfJbNrGU2lVnd7KpMZN4y25EnnGRaJg6/FcUM+VxSvgxn2AOiV9py6/eQqsqX 3j0V5nxJVlUK84aSnH/XbDyTN6oLhbmX7piv96hExgiZTCaZi775fH59fZ1pkeixIAik4MoQO2Za JtI0jeM4cxTl35KOD3IYBEEgl5xa6zRN0zSVZczG0zQ1nyIvM7tk5tj9KezVzbtpmsqhaN7acOnX iavCSlWQdWV43ZY7FyGV2KU387LZwpwvyapKYd5QkjPvdr0w95I3nU5r908zd2tk9clkYsaqsqdL SpLk6uqq/NfkqPYetfW8n8ZlKJyWiyyzvMyUdwvfknpGHMf2bySLhWFor144sW665Myti+U/4jDV eIwsU4ZFvvRGUVTpGJERabtVmDP/3w2FeXM531qYyxfgwt3bpXjnf7vO6fb4An/84x+rrmOuB+fz edV27c2brVfZOkxy2SW2viXXZXZDM2PE1VOyNFYqw/UqN+fn51VXOVgUZjTizi9+8YsagWGayzLH odlUvT4In332WdW1DpNdYc+0FK97S+bLMWlGDi7zWQ22RG/+0MM/X1Q6xa8rwxlSpM2InOW3X3UU tYNFYUZT7iilptNp1b43SZKYfqKm5WE+n5s+pjUa6KbTaScOUTkkzN3LwmkZSzRzw1/dHgOZt4T9 JQsyYd8FNRP21swy0ixuNiXbN/dsM/tm1rKPbXP31W4ot3evE7X44+Nj+fLyrdaVYWV1e1G3faOl VJe80yNrldwNo63bCaZgZErIhsJsF8UahVm+WTFfklW5wry1JKtcYc5HSycKcy95ctfuT3/6U7vd kaMoOjo6Kv+18Or2EKXc7K6wLducUDLTB+7hw4fv379vvTA/fvy40p+LwtyUfGHOlN4OFeZCnQ7L O0qpIAhGo1FTt2FqkOvBSmEjultoDp9Jmm4dnxcXF60X5qOjoxp/ru6eRA6cfc3UrcLcPz9+39py uZRBbd13+qzRLc2g3OxuQ0tOR/+8HS3M2N2GDtCZCbTip++/WS6XV1dXji8MZbC1Gscnj2s1RfoX 9ezvuVwua4/jV1vtsMk8kIjaelmY++SO/cJx5Mjx+eLFi6ormg7++9irYerf+e7FixfuC3PtK6f+ /f1bxB/zYBXcKJ7NZg6+oSBJkv/3//7fJ598om57uZRckTurKM/NFwlGUfTtt99KYa5UMinMqKrz /QUylsvl0dHRLsOgbSa9S1+/fv3999+bYWCky+OGewlmoCTF8YnSlsvl69ev912Yj46OpDArawCx MoU5CAIKM4Zj06AOi8VCay1f/dnIFaKMHTKdTtc1O9gPi+T7L1aqBgG22WwmQ9M2UnGXm0Oj0SgI grOzs/wCmUem7MJsxhajMKOGTtdvtg8itHvqbD04ATeePXt2eXlJYUZ39TxvhNZ6sVjI9wXYB2r+ atHuFyQT0+n07OyMqzkciMLCXJhAdiucKcxBxNVSAAAgAElEQVR0d0aLBpE3GVLpMS/l0LVH2+Tq D12xtTCfn593YqQlDMEQ8yav038FwKAk45B1unwW9E8DAKBx5A0AwAXyBgDgAnkDAHCBvAEAuEDe AABcIG8AAC6QNwAAF8gbAIAL5A0AwAXyBgDgAnkDAHCBvAEAuEDeAABcIG8AAC6QNwAAF8gbAIAL 5A0AwAXyBgDgAnkDAHCBvAEAuEDe1KS1zs/Jz9y81tblAaA3Di5vwjD0PM/zvMz8wpn1tr91TuZD Pc/bulYYhlrrIAg2rKi13vBy687g8HkfqnQ9Ufivp4SgcY2cSDNMsdxcPg8ub+I4lgn7WLVP5bvI b0ffWrdwmqZpmqqNf0d5y97zwhUzH134GwVBwAml09JbcRyX/1cWlvDCwkkJwaGxS++W8pk2xJxh d6duz92ZjZs59u9mz5GXMp2mqQkAe8uZz4rjOI7jwp3PLCwvzQdlNmvvQOGK67YpO2nvQBAEcRyv +dtg73YpyRvKjPkvS3nLFM7CI3Fd4aSEDNnuZ1q71NlFVGZmyqe8lInCopvmSu+G8nmgeZPenovT 298/tY5e88vYZ2p7+fT2WE1vU8fesmGWMevmd8N+aR/8hX/ufBStWzKzz+rD6+IGrwNQVVN5Y65+ 8sXV/N+DICi8Rkk3Fk5KyJA1mDfph+cue2a+7GVOy6bopkXn1XXl83Dzxpy4M7/VurO5CQO7lpOp xuVz2Ezk9z//QeuSIzNdKW82z4d7O+aNKXhmTmaBtKhI5P/jlQonhsNN3qybuXWxdXPEwd2/MSQk C5sCC2+3BEEgjd3mgLevCgs/Qmtt7usWNpTX6z9We0X0gH09KOoVhs2FE+iiw80bdXviNs2F9kyZ ztxllVtV5laKyarCYzUMQzt48yFhbyEMQ6k/revaYdaVq9HMihV/7+KuBOiiDcW1cI4oUzj3t8/o qwavWjb34Vr71vpaVzVNtadlbkaZzcp803RofqsNu5Gv6GS2lnmprFZLVXSn18yxewfYq2cWy+xh 5lezf6PCnUQrav/988XAbDAz3z4s8x+6uXDusofogV3++5kTV1p0QrNPU6bgmVsy+aKb36UNe+il a9qaqqp3Ie+YxHuZa8O2fp3ye4g9cfCv97zscVf+/04JGbh2z7T5oqs+LJOby+dBt6c1zv6LbJBv xHODU8lglS+ZihKCA2NK79by+TN3O3UYyhyrbR3PnEeGYN3TcIdcMgG18Zn3kiVzWPUboHVyG4b+ Zuic3YsueQO0gJoKOmqXokveAABcIG8AAC6QNwAAF8gbAIAL5A0AwAXyBgDgAnkDAHChgfEFzJDJ MtHUgGyAY1preYLaFGkKM9Cghus34/G42Q0CzmQeZLt//35bewL0UgN5Y2fM3bt3d98gcAi+/PLL tncB6JUG8sbOmJcvX+6+QaAt9sUTQ84AzWogb7gMRG/cu3dPJmgZBhrXQN6Yy0Dau9F1pjCb4AHQ lCa//4aKDrouCAKp2Rz+l9UCndNM/zQ5RGnvRg+8ffu27V0A+qmZ+g3d0tAb4/GY8gzsg1f1ibYw DP/5z3/K9OXlpVLK933zbpIkMmc0GimlRqPRxcVFYzsLNOrhw4er1UoptVqtkiSxS7K6LczT6VRe fvzxxzSyoXVhGHa3HJbNm9lsdnl5aQ7I+Xy+dZUoimRCsme5XJbcJ/m+UlrnsA9a68VioZSSpFEV C3OSJNPptHxhDsMwCAIKM5rS57xZLBZa69Vq5ft+5uqvkiRJkiQZjUZBEJydna1b7MGDB69evVJK jcfjN2/e1P44IK+pwizZs7kwa62/+eYbCjMa18+8aergtK1LHa31kydPlHWrlkMUDXJZmMMwfP36 td3pQAqz1ppaDnbXw7yZzWbNHpwZURQlSRLHsbkGzPjoo49ms1lmphyx3f1bwz0ZgtP3/TKNZjWY 1Dk7O6tUmAWFGVX1LW9ms9np6emeksb229/+9h//+IdSajweZzqhrqvfhGGotY7jmEtFbLVYLJ4/ f76npLElSfLVV1/98MMPqmJhVkpxdweVdDpvPugPfXNzIxeDDsJGKfXnP/85iqLPP//cNEfIEajW dxaQP7Q5UB3sJDrKWdio2y6aUmU/Pj5WSmmtpduLojADt36q3zgOGyMTOSXR7QcbuAwbQ9rWTOSU 1+krVjjW6dLy4/gCppnbcdgopebz+fPnz6WLanlxHJvrR8DWStgopeTwefToUdUVgyAwNXugx37M m8ViMZ/P3YeNmM/ndvtDeUQOMqQguQ8b4fv+6enput4B61BNx0DcUdZT1i1arVZVqziKvEHOYrE4 PT1tcQd836cwA4XuKKVWq1Vb14NGjR3obiMm9sceBaMtvu9TWQfy7qjbYdBqkAGm5OfuVqvVzc1N I5s6TJxQ9u27776rGjaFZXjHIu37ftVjqltNapRk1HPn2bNnNa4HoyiaTCZRFMmTm0qpJEkmlnpH bI17rY6FYeh5nud5VVeUHhn72CVRY5e2MjvclbvZv//978svnC/DIkkSM1qaUsoUaXvmVr7vP3v2 rPzyreRNvcK875Ks9lCY7R3uSmHupTv12h+SJLm+vr6+vrZ7GVzfqtfPrfU2vTLiOK46orbYxwll r0eOPf5KhzpQlS9FhWVY5cY7N6W60iVUjSqOe/UK856icX8FLDOSUIcKc//UzBul1GQykQlZvfVG 81bItZ7W2vM8e1reWnfxmHlLLjPlGJBplbv2tDduljSfZTNrmU1lVje7KhOZt8x25AknmZaJw29F qdrtJVOG88z8JEkqXQ91Im9s+dK7p8KcL8mqSmHeUJLz75qNZ/JGdaEw99Id8/UelVxfXyul1rWb 1c6ebh2iSqkgCKTgyhA7Zlom0jSN4zhzFOXfko4PchgEQSCXnFrrNE3TNJVlzMbTNDWfIi8zu2Tm 2P0p7NXNu2mayqFo3tpw6deJq8JKVZDNZTiz2UybWxn1jqy22KU387LZwpwvyapKYd5QkjPvdr0w 95I3nU5r90+TgzAzGGIURfW2liTJ1dVV+W8WUe09aut5P43LUDgtF1lmeZkp7xa+JfWMOI7t30gW C8PQXr1wYt10yZlbF8t/xGGqMfRfYRmeTCaSRrbCmRs2OxqNulWYM//fDYV5cznfWpjLF+DC3dul eOd/u87p9vgCf/zjH6uuYy705vP5fD5vqn+afJNVI5s6BHLZJba+JddldkMzw2rVU7I07qkM287P zxvfZlsozGjEnV/84hc1DjbTVyf/Fby1G9OSJPnss8/qrXto7Ap7pqV43VsyX45JuWwseXw22BK9 +UMP/3xR6RS/rgw3peooageLwoym3FFKTafTqn1v7N7PdiPDLnkznU47cYjKIWHuXhZOy1iimRv+ 6vYYyLwl7C9ZkAn7LqiZsLdmlpFmcbMp2b65Z5vZN7OWfWybu692Q7m9e52oxR8fH8uXl2+1uQyr 264E9jKVer6V3A2jrdsJpmBkSsiGwmwXxRqFWb5ZMV+SVbnCvLUkq1xhzkdLJwpzL3ly1+5Pf/pT u92Royg6Ojq6uLgov4oUQcrN7grbss0JJTN94B4+fPj+/fvWC/Pjx48r/bkozE3JF+ZM6e1QYS7U 6bC8o5QKgmA0Gu2jCbskuR6sFDaiu4Xm8Jmk6dbxeXFx0XphPjo6qvHn6u5J5MDZ10zdKsz98+P3 rS2XSxnU1v1jNDW6pQm78Qe1bWjJ6ehh2cXCjEZs6ACdmUAr7pip5XJ5dXXl+MJQvqKqXthQdBoh /Yt69vjbcrmUouXyQ2uHTeaBRNTWy8LcJ3fsF44jR47PFy9eVF3RdPDfx14NU//Ody9evHBfmOtd Oak+/v1bxB/zYBXcKJ7NZg6+oSCKom+//faTTz5Rt71cSq7InVWUV+MJ0BqiKPrb3/724MEDVbFk UphRVef7C2Qsl8ujo6OqA+KWJ+PyHh0dff/992YYGOnyuOFeghkoSXF8orTlcvn69WsHhfndu3dS LM0AYlsLszSjUZgxHJsGdVgsFlrr1WpVb7znPDnmN/R7tp8dy/dfrFQNAmyz2UyGpm2k4i43h0aj URAEZ2dn+QUyj0zlu5VTmFFPp+s32wcR2j11th6cgBsUZnRdz/PGkCtE9WE30/zVot1wIXdrp9Pp 2dkZV3M4EFrrxWJRrzB/+umnNR4UA5oylLyx3dzc2N/FKYeuPdomV3/oCqn0mJf5wnx+ft6JkZYw BEPMm7xO/xUAg5KMQ9bp8lnQPw0AgMaRNwAAF8gbAIAL5A0AwAXyBgDgAnkDAHCBvAEAuEDeAABc IG8AAC6QNwAAF8gbAIAL5A0AwAXyBgDgAnkDAHCBvAEAuEDeAABcIG8AAC6QNwAAF8gbAIAL5A0A wAXyBgDgAnkDtENrbX5WXWt/9r19DFmbeeN9qGpBD8Mwv50wDM18exngcIRhaMpqyWIvq5i1GinY mQNQdqbqls3yHGvYquX6TXorjuNK5VVrHQRB4XaUVfSDIOAwwKHRWpviahfjDPuKyqxiCnkjUksQ BBt2ppB9GHKsYbu0IUEQVF0l8+ny0i6+Ml+OSXOM5dfN/xb2HFm36r5hsGqU5KoKD71MCTcHQlpU wg17MftIsafNYvKryXRms+YYKdyBzJx1O8ax5oCD8rk/h5I35kiwy73Zphw/6e0lWFolb8whB5Th 5niWU7Yp7XEcy7RdXAuvrvLbsSdkm7Ipezq1DiJ7Tv7Sc91EmR3jWHOAvEnTunmTuYYqrPGkRelS Pm8KFwDWcXY85+scW0/9hSRUth4XUuNJrV9wc/0ms+WSQcixtm+dzpuW79/kL7XoHoN+MyVczub2 S3NYlt+a3DIpc98lCAKtdebGp/3u5i1L04L0LKh6mwcQh9UfWg4JmS48MNYdLbYwDDPLcHjgoNj9 WcyEmZm/5MrfipdlpEdZ+eIt2ymz/LotSxxu6LDAsYYNWssbOWCkF6aZKeXYdI+2i3VmpinWZjum T6fduUApFYZhg/15gB3JRZUpseltw7KpPdgLy0v7uLAPGTkKzLMET548MauY48Lemt0drvAANI8E Zbasb5ne2DI/ky4ca9jMq1R532CvRc0clkZhcc8ruRhg9PukWak+lFlRWS0QYRimaWofXxxrbnS6 fB5We1p5dhFfhwMAEKZ2UvtwyFSq7J+yccWxhm1+1vYObLfuIbKthZvSDxg7XhfbT2SbfmuKowxV dCBvpGTXbgcAYJ5aa3ELQGfa0wgbAOi0zuQNAKDTyBsAgAvkDQDABfIGAOACeQMAcIG8AQC4QN4A AFxo4HlPMyagTPBQGDpKRnxRVpGmMAMNarh+Mx6Pm90g4EwQBHYBvn//fos7A/RPA3ljH6J3797d fYNAW96+fWumv/zyyxb3BOifBvLGzpiXL1/uvkGgLfbFE0MoAc1qIG+4DERv3Lt3TyZoGQYa10De mMtA2rvRdaYwm+AB0JQmv4+Aig66znQZ6O5XKAIHq5n+aXKI0t6NHrC7DABoUDP1Gxof0Bvj8Zhu lsA+eFWfaHv48OFqtVJKrVarJEmUUr7v2wskSeL7/mg0UkqNRqOLi4vm9hZoUqYw50uyUsoU5o8/ /phGNrRux+8Fb1fZ+s1sNru8vDQH5Hw+37pKFEWr1eo3v/mNHK5nZ2clG9y01orWOeyH1nqxWCil JGlUxcKcJMl0Ol0ulyU/LgzDIAgozIDaWr9ZLBZa69Vq5ft+5uqvkiRJkiQZjUZBEJydna1b7MGD B69evVJKjcfjN2/e1P44IM9lYdZaf/PNNxRmNK7T9Zu1edPUwWlbd6BqrZ88eWLfp+UQRYNcFuYw DF+/fp0vzFprajnYXQ/zZjabrVarMu0M9URRlCRJHMfmGjDjo48+ms1mmZlyxHb3bw33ZAhO3/f3 VJhN6pydnVUqzILCjKp6lTc3NzePHj06PT1t6jJwg9/+9rf/+Mc/Ct9aV78Jw1BrHccxl4rYarFY PH/+fH+XTUaSJF999dUPP/yglBqPx5ke1RsKs1KKuzuopNN580F/gcVi8bvf/W4+nzsIG6XUn//8 5yiKPv/8c9McIUegWt9ZQP7Q5kB1sJPoKGdho267aEqV/fj4WGZSmIGMn+o3Lo9PWyZySqLbDzZo pTBL25odOSV1+ooVjnW6tPw4voDWupWwUUrN5/Pnz59LF9Xy4jiWbtNARltXTtIZ4dGjR1VXDILA VIaAHvsxbxaLRSthI+bzuda6Rn4QOcho8cpJKeX7/unp6breAetQTcdA3FHWU9YtWq1WVas4irxB zmKxcHP3cR3f9ynMQKE7Sqm9dn0uqcYOdLcRE/tjj4LRFt/3qawDeXeUUpeXl/VWlgGm5OfuVqvV zc1NI5s6TJxQ9u27776rGjaFZXjHIu37ftVjqltNapRk1HPn2bNnNa4HoyiaTCZRFMmTmzIzSZLJ ZCLz6+1NjXutjoVh6Hme53lVV5QHD/exS6LGLm1ldrgrd7N///vfl1+4sAwrpZIkyRdgKdvlN+77 /rNnz8ov30re1CvM+y7Jag+F2d7hrhTmXrpTr/0hSZLr6+vr62v7YZ0oimRmvcvD1tv0yojjuOqI 2mIfJ5S9Hjn2+Csd6kBVvhQVlmGVG+9c1aru1KjiuFevMO8pGvdXwDIjCXWoMPdPzbxRSpnLPbO6 7/tyqdh6A7ozcq2ntfY8z56Wt9ZdPGbekstMOQZkWuWuPe2NmyXNZ9nMWmZTmdXNrspE5i2zHXnC SaZl4vBbUap2e8mX4XWqFulO5I0tX3r3VJjzJVlVKcwbSnL+XbPxTN6oLhTmXroznU5rrHZ9fa2U mkwm9qXffD6XtojaNZVuHaJKqSAIpODKEDtmWibSNI3jOHMU5d+Sjg9yGARBIJecWus0TdM0lWXM xtM0NZ8iLzO7ZObY/Sns1c27aZrKoWje2nDp14mrwkoVkcIyXLjNetdP9Y6sttilN/Oy2cKcL8mq SmHeUJIz73a9MPeSN51Oa/dPM7UZWV2er5aJGoPiJElydXVV/ptFVHuP2nreT+MyFE7LRZZZXmbK u4VvST0jjmP7N5LFwjC0Vy+cWDddcubWxfIfcZhms1nVof8yZVhMJhNJI/sbBc3MkpsdjUbdKsyZ /++Gwry5nG8tzOULcOHu7VK8879d53R7fIE//vGPVdcx14Pz+VzqNPJSajaSNDW6DMg3WVVd62DJ ZZfY+pZcl9kNzQyrVcN0Oi1ZxVlXhvOLSbcCaXmr1GXg/Py8/MIHjsKMRtz5xS9+UeN2qImTwmvJ erWlJEk+++yzGiseILvCnmkpXveWzJdjUi4bK30daiM2f+jhny8qlZ/NZVjM5/PrW+q2Ca6kqqOo HSwKM5pyR1W5KhS+75uuz3Yjw3w+N/2hKx2ZYjqdduIQlUPC3L0snJaxRDM3/NXtMZB5S9hfsiAT 9l1QM2FvzSwjzeJmU7J9c882s29mLfvYNndf7YZye/c6UYs/Pj6WLy/fal0ZVrdVn0pVmQxpTKu0 Slu3E0zByJSQDYXZLoo1CvOTJ09UUUlW5Qrz1pKscoU5Hy2dKMy95Mlduz/96U/tdkeOoujo6Oji 4qL8KlIEKTe7K2zLNieUzPSBe/jw4fv371svzI8fP67056IwNyVfmDOlt0OFuVCnw/KOUioIgtFo 1NQwATXINyRWChvR3UJz+EzSdOv4vLi4aL0wHx0d1fhzdfckcuDsa6ZuFeb++fH71pbLpQxq6/7R mRrd0oTd+IPaNrTkdPSw7GJhRiM2dIDOTKAVd8zUcrm8urpyfGG4S9hQdBoh/Yt69vjbcrk0vfOd qV2YMw8korZeFuY+uWO/cBw5u4QNlZtm9e989+LFi64UZtXHv3+L+GMerIIbxbPZzME3FERR9O23 3/73f/93cKvkitxZRXk1ngCtIYqiv/3tbw8ePFAVSyaFGVV1vr9AxnK5PDo62mWY583kAbqjo6Pv v/9ehp0wgyllHle2mYGSFMcnSlsul69fv95rYf7DH/5wdHT07t07KZZmALENN8akMEszGoUZw7Fp UIfFYqG1Xq1W8sXsu3+YHPMb+j3bz47le+JWqgYBNqm1q4aGIZebQ6PRKAiCs7Oz/AKZR6YozGhK p+s32wcR2j11th6cgBsUZnRdz/PGePjw4Wq1ynx/Qf5q0W64kLu10+n07OyMqzkcCK31YrGQwcgr FebpdFrvQTGgKUPJG9vNzY39XZxy6NqjbZ6fn3dicBpga2HmeRocjiHmTV6n/wqAQUnGIet0+Szo nwYAQOPIGwCAC+QNAMAF8gYA4AJ5AwBwgbwBALhA3gAAXCBvAAAukDcAABfIGwCAC+QNAMAF8gYA 4AJ5AwBwgbwBALhA3gAAXCBvAAAukDcAABfIGwCAC+QNAMAF8gYA4AJ5AwBwoW95o7VuexeAnbRY htd9tMzn4MKOWs4b70OVCnQYhva0bCG8lV8GcKle2dZa5wttZlNmgcL5nudt2H7hESEfmj981O2R JTPtX4FDDDX8rO0dUGmayoQUevNyM611EAQyLQeYvaI5BoIgCMMwjuMm9xgop0bZNqV63abUh/WM dfML6Vv2p8gc+xixt6O1tn8LM2G2wCGGCtKGBEFQY63MDshLuyinaRrHcRAEpkDnV9z8W8i6NfYN w1SvJOfly7YpwFLC7U9USplSKtOm/Gc2ZRfmwvkbDoc4juVo2rCf+d8iv0BmDoeYS02Vz1YcUN7I Fuyyaw45czQGQWDmFG4kTw7dGvuGYdpH3uSTY0MBNqXdXthmHyD5a8d1pV3Cxt5+4fLm8s5eYOsq HGLOkDdpukPemNqMmZNZIC06OMvnTZkFAKPBvMmU7XR93tgfml8mU48prMpsrd+YjzAXbeuWz8+R X6SwvrX5Q9E48iZNG2pPS9dcfJE3cGZP7WmZmfkmqQ11oHUFePNH5OevqyRlWsPMFuz5diWGvGlR p/Pm4PpDB0FQeFty3Rx1e8cys1hmgT3sKdCYOI7Ld86UO/yVFpbeCoZ9lMnhs26DdtebDZ/CIYZS mgquGqlrugBk1s30F0itSzO7TaBwFVnGvi7r9OUA3GukwGwu25nmKfulrGjfaDF9BzKHbeFHFPYT y/Q+MMuYwyRz+GRa2/LnivwvtftfDCV1+q/9/9s7f9+2jTaOH40XKhpP2S2jyFD0D9AQB0EmdpEW CwU5xB6EDl0ytRnKyUMnLRmK1kuWcqgziEO8REs5ZXCNQuifkCHO0CV1gf4CtPAdnvrJE/4SSVEk 7/T9DAlFHsmz9By/9zz33NGKiuUfr2SjOZGWFa8ndcdW9qoKFgOAQXbvSmSzQhNrGK3ts3PxtIJI W88CLQGATcCtD00MlKL9+Z4ryZrAvNLK0QwA2BBoXKACGvg3FGXG2k0AAKA1GugNgf4UAABojTZ6 AwAAQGugNwAAAJoAegMAAKAJoDcAAACaAHoDAACgCaA3AAAAmgB6AwAAoAk0WF8AgGagdZTVzRvK 1ftvawYArAn8GwD+w3Gcfr/PHw8ODlqsDADmAb0B4B1XV1e8/eWXX7ZYEwDMo4Z42v7+vlLq7du3 tPH69ev1rwlAK/T7fZYcLKEEQL3UoDfcPv/555/1rwZAi9y9e5fsWQbWAAC1UEM8TbZMtFKgNezT 3L17t92aAGAeNeenPXnypN4LAtAknDKg7ysUAegsNfg3sieIkDfQHZkyAACokRr0hnuCCKYBA+j3 +3t7e23XAgADqTOehpA30IsgCPi9sbThOA6Zseu6vIcKOI4D9x2AdbBqmUG9v79/dXWFydig+1QQ kpgsOTdsuqoAJHFdV9/BxXr8G8QfQPchpXEcp2zHKKkuLFpFWn4QBI8fP8a8NACq+zenp6fz+Zy2 f/vttz/++OOTTz6hj8Ph8NGjR/VUEIC1oVXRVLY8SGOOkWPMLGBZl713796bN28oAQHeP6gFy6on KNUKJao+Ho+XyyU1S9u2ef90OuXtMAzDMORtpdRwOOz1es+fP6+tygAUhpQm6aCkGrNt29KwPc/j 7RxjTqoOOTTq/VQ3fZ8RoFMYrjfX19fHx8fL5VK9Ly3FIRHq9XqO40wmk/zC1CXc29u7uLiocC8A iFSl8X0/CILlchmTluKQCPV6vRcvXsRuFwTBxx9//O+//6ZmVMtqyAEkfWPxoBWM1Zv1G6ckX3Vi XcJ+v494N6iM67q1K40ky5h3d3c//PDDt2/fJk9JbWgc6EMCAiiIgXoTBIHv+3U1zhie5/V6PY5L yBg3c+vWrdFoxJXhpojOIFgJpZOxqYzH4/Pzc9u2q3nnObDqTCYTNtGk1MXqk1VhSA4ogml64/v+ 2dlZ7Y0zhud5r169evXqVerRpH8jc1KhOiCL2MN9NBotl8sGjPno6IgdnaTkFIF8Hdg2yMcovWmm fRJhGH777bf3799XSl1eXtJOcnQODg6yxm/QLEEWSbF58OBB7Q56Kp7nHR4eciZbZcmBlwPyMUdv mhQbIgzDly9fxoZeV7IyQAG2kBbFhqhFcrR+moAG0NpC3lW9+fZJQHLA+rQuNkRScsqaKNx3kI/W evPfep3j8biV9qmUsm37wYMHnB1QEMQcgKQLYqOUmk6n5+fnvu/zHs5AKwiUBhjMjlIqCIK//vqr lfZJ2La9XC7H43HZE8s2ZmAqnEtCEzlbNObpdHp2dkbb1cSD/xYAkmjd1d5RSvm+32L7JEhySp2i 9fcOakRmzDc8AJlFWX8dgG1gRynVbn+QsG07awGrLKA3gGC9ub6+LmtFm2A6ncrOE/wVAIid6+vr yifTolK8YNqa2LZ9enpay6W6CZ47m4P05tmzZ2V7Tqk2XItJU8tyHKfs765XRwpWDYqzc3x8XOE0 z/MGg4HneZ7nyQU6eX+Fa1ZwcZrHdV3LsizLKntiEAQbHW2qUKWVcIV1GSebz+fF9SbVhtWNGcuS gxtKVYZaVgW9aYVqhr1pq1YbMGztrJLhJFoAAA9mSURBVNokdlSlVTjDMFwsFovFYjqdUgv3PI+W DFksFskWWwQt9GY2m1VLRtxEp3WjDUYOijiO09nGKStWSm+SNkzEruB5Hln1YrEoXivZprTwV6oZ NqwalGKn8iOeu3vUPmUPsfJoUOvDSGWh/l0QBJZlyW06lNVhjB2iriWZPm2rRH9TXpxL8r0kfBZf KnY6V5U2Yof4OjRdkbZpo7P9dK5nWfuJ2XAq7LWXrZVsWXo91zpo1SrNsHOsOnmUa6WLVRvJznA4 rBCtpr7eYDDgc23brhZGk/R6vTWv0DAcLZnNZnKbNqIoms1mscaTPERZs2T9/PbJIAiiKIqiiMrw xaMokm+WTPZJeY9MxpWn89Eoiqj58aGcHl9nO4MyYFXKfpI2nFVssVjYtl1KcsIwHA6H/FGvWTUd tGqVZtg5Vh07mmW9nbVqU7G+//778/PzaimkFPjmlXc5Dk69xbLXpHWjSy000NZUW3nf1G1+3zBB O+lo6iGeHi9npFMx13Xl6akbWdsFd64slrxFp+CKVViQKWbDxGAwSEbPUnfmXJYXGii7ykCLL6jP t5zmrTqrJgXbQr6pJ6/ffXRffmLn4cOHZf0bLj+dTqfTqfxIncEwDKsJ2HfffVeqfJcj49TVIlYe or6YDC5zH63JOmsKf0vF7SfLhusiDEMWm3qv3C6warAOO7dv35aOf0E4dJYMfFPiQLXa3Llzp9qJ XUP66bEAcdYh2k9NkV9RXOReNQag82/a5ccEfaV37twprhw5NpyklCDFgmld/t5K0aRVJ2+xDjn3 NebX0YIdpVSv1ys19GLbdhiGnCTKQQYaWa32Yit6b1WpU9oa6KP78ohl6jYtDBwbGlU3dh87RFCs nLZpQ4588oa8GpexLEs2G7o+D8DG6sZnyS+QR1xlcFxWr8Ugz0pkxQqOR2bZsLqRFhqt4TKUzFaw PtKYSz1h+ZRS5euCjaQLVq1uEhByDHulVauEYWtk1UbyLvD9zTffrHktioNXO5fC6M4NRU7RPZTZ HVJD2NRuZQ5Pl3uC/CcEQfD06dN2l7Q5OTmhYUgewCh1ul4jCl0m+U3qZdVJdH/o/bc+tOM462eX rSM2R0dHlEbCyZFFAt/6fu/dh9ukFs2Sk5Ecxzk6OlrfmCvjeR6/6LOap9Lxr1pr9LLqVLR+6P2P /qMWQlPbGq6BfBevdG6k5OClh5sjR9c1+s4pCUop5ThO68Ysu8/V3n8D1icns7/hmgBmh7cmk0nz HcPYi98lsxuk08MdRjlvC6wDpRUZMOuNJojQHzKZTA4PD1s05mqRNELrDmx3MMawTSIe3/R9/+zs rJmO4eeff/7BBx/88ssvBcvz0+Ty8nJvb+/i4mKTtQNaIt/i3KQxn5ycsGtVWWzWUSmwDeie4LAT +zyZTL744ovKa24WhBYIuX///kcffeTesPIsDlM8efJkb2+PF8xAFwYwcrkBctmbMebJZLK+Z1Mh mQ0AjcjMhKH3JNKrcWpc1ozy0A4PD58/fy73c9qubG8yQiJ3yo/BDXSi1uIPaiH5xB+Px+fn59Uy 9bMIw5DyntmtSb11caRnBkAquucurqi97/tBEKyvOtw4h8MhzbvOIakxBduhdJLQdLcZ7oJIGzg9 PZ3P5zUas1QadbNAZIV+D2c6wGJBPobrDREEge/7Sil6a2HB5krNkraTjXOjxLwlOD1biMx5lc9x 7kLRx4IeDxszTeScTCbymnyvIh557CyyUpgoKMJW6A0TBMGzZ8+ordJy60nhoWbJ63ns7u622JZi jR+tetvIUh2llOu6f//9N20XMeZer/fw4cPkRWgj1bSSU+jlPKHUWgGQg+75AjWoZU6T7hRweraZ gjEr6amsvCBcE9Aw0Bv9gNOztaw/gxhBMNAWBqTLb6PeSKTT03H/DNRFbCVjuaBk6qhMsqTWbR5o iu6LpynoDYO86m0mR1rQCwEdAXpjJnB6AABdQ/fBG8XrdQKJfPet7Pnq/mMDAECLwL8pCpweAECL 6D75RkFvKoD0NgBA80Bvth04PQCAZsD4zbaTNdID7QEAgBjwb+oHCxkAAOrFgGRoBf9mE8ScHrwV GwAAFPybxoDTAwCojAGDNwp60wp4Tw8AoBQGJKcp6E27wOkBABQBegPqBE4PACALxNPARsDKoQAA iRnJaUqpnbYrAOLQi1WiKCLzcl3Xsix6u1fbVQMAtIMZAQ/4N3qARXQA2FrMGLxR0BsdwSI6AGwV ZgzeKOiN1sScHmgPAOZhzOCNgt4YA1KrATAS13WN6UpCb0wDTg8AJmHM4I2C3pgNUqsB0Bpqv8a0 XOjNtoAsAwC0w6RgmoLebCFwegDQBZOCaQrzPbcQzCddyYa+ivzLyqP4LYAS6T/mEAEQRbPZzBG0 XZ364XZbpDAXq6WByEdG1tdL2p/cptNlxSR0yMjfC0Qm/rJGOWugFgxOrS4YnVg/iMET9CzLUkrJ q2XN3ZM35W1ydFixYhWj38iwIWVAmPmztqh1oOOY5/SQwdPfRS1Z/l30WKd/eU+sPO3np0DsIwUq pS+V08RiLpcsmboz9pHuxZeSH4EBGPmbQm9AIThJRmvtkc9xasxSM/jvSj7uqTyPe8kBsJUnFqlM VFhvJPw8ioXggO5IozIJxNNAObSeT8rxqJyNlUeTuRXJRpR6YhK6juu6URTlVyC2X60KtQGtMWbB tBiwUVAd7VKr69IbqbKpD/oiesPXoTJl9SbrjkB3TFowLQbyoUF1TEqtpud+wZL8MlbSWj4x+YfL wrIMfUtruobS19TIywQ50A9qpNgohZgvqBsKPavujfTIUX0lxm94FETmC3CCAI3WxBoLP9xjCQJ8 KVk+lg8tFZor89VXX/HpsnryC+QEh2T77dT3DNbB7GcyfHCwQWTXXq+Rno4QG6SpXAZogWGr1ySB 3oAm0G6kpzvkywnExhiMFxsFvQHNg5VDAYixDWKjoDegReD0AKC2RmwU9AZ0BDg9YAvh/taWGDz0 BnSL2HxSOD3AVAyeZ5MF9AZ0Fzg9wEi4U7VVYqOgN0AL4PQAY6Be1Hb2n6A3QDPg9ABN2Vq3hoHe AF2B0wN0AUpDQG+ACcDpAd0ESiOB3gCjwJwe0BFYadABYqA3wFjg9IBW4GUDYXUxoDfAfOD0gGbg 10zAxlKB3oDtAktWg9pB6Kwg0BuwpcDpAWsiTQj2UwToDQDvOT14cIB8pMzAoSkF9AaAdyDFAGQB mVkf6A0AKWAyKSDQBakR6A0AK0CKwbYR623gR68L6A0ARdEoorLNi0JWRroyCk7tBoDeAFCaTkXb xuPxcrlUSi2XyzAMbduOFQjDUCk1HA7p4+7uLp6kDDSmSaA3AKxFW/H90Wg0n89ZXabTaZGzPM+j jV6v5zjOZDIpchbpqxnPYu4rQGOaB3oDQD00M6Hn+vr6+PiYHJqCGpOF53lhGA6HwxcvXmSVuXfv 3ps3b66urvr9/uvXr9e5XYvEnBgFjWkJ6A0A9bOJCT1BEPi+v1wubdtOBs0qk6o6QRA8fvxYKXV1 dUV7SG84ithxd4e/fzgxnQJ6A8AGqSvaNhqNalcaSRiGL1++dBzn6dOn5NDECty6dWs0GsV2diR1 QobIlFJUmdZrBZJAbwBogqxom2VZ/X7/yZMnOQ/H0Wj04MGDDSmN5LPPPvvzzz9///13pVS/35eq kxVPa+X9Lkn3peEKgGpAbwBoGhltu7y8pMf6wcHBxcVFrCSN1jQjNoTneYeHh48ePUrWNueBvtGE glR1gfuiI9AbAFrj3r17P//8M3/s9/t7e3tSdRrzbCRJySlCXZIj1UUhOGYW0BsAWoNel5LcP5vN HMdpRWwIz/OOjo4KZkszruuWEgY5jQm+yzYAvQGgNfb395Mj88Tt27e//vrrVsSGqCA5OS5OTFrU jeOioC7bBPQGgC7y6aefrjm9Zn1OTk5ypuakIoemFKQFvM//2q4AACBOMvO4FZbL5fX19e3bt1OP SmlhXcF8F5AD9AaAzrFcLss6N7RyWmz9tNTl1EpxfHy8u7vLH2P+ikroimVZGimNHDQCDQC9AUBv aIEApRRJCwtMGIae5y0WC/o4GAxow7btgmI2nU5PTk5ou6CKtPL45rSLUqMDQRC4rrvRAQXLqnnA wnVd+iF4Qy922q4AAOA9fN8vVT4Mw8VisVgsptOp9GaSjg4VU2LVzpXM5/PZbNbxR9tsNqvwWN+E NMaGr+ollsK30XttCOgNAN0iCIKyQTDpu6QW8DyPfZrpdEr+UBFs207N2O4s5LUEQWBZltymQ5Zl 0XbyLHnIdV3LsuiBTtu8wWXkxbkk3ysGn8hXk6dzVWkjeZRrJfVG3YQ3NQJ6A0DnKK4HSilyWQaD QamziqPXCAflLKibOUy8zQG32WwWk4TkIfLneKopOU9BEERRFEURleGLR1EkV/RJ9bR4J3uK8nQ+ FEURvyiPj2a5Mjq6OMiHBqBzVEiGplGc2NjMYDBgNeKBnOTHHAaDQalHRIvjCjxYIkdNeDs2tVaW TD3Ec4nkX0TFeNQnecfYgE1qTfLPytmZHA2qfXxo08C/AaBzFPdUuOR0Os0JlFHqWoVq8FtBDYAc CGLlIfIw5JAJex5N1tkwoDcAdI7hcFhcHnjwP2fUx7ZtLibHcvIxSW9k9Ck27JF1iPaTwJAPVGqp nhoqfXOprPtqJ37QGwA6h+M4BfWGHJfBDTJKRlegVAJ6cQ6VUbnKJAnDsMKqnaXK1wXdl8fbeUye t2lpg9iAvxKv8EmmEtAIEG3ThhzPl2/ckRsyN0HqAd1CZhZwYb5y7NvjVAI5nsTomBKtWfgPgC3B sqyCQywbgt7AVmE9G+0egt0kOTYjJ9smJ95qAfwbALrIDz/8UHyWTO1UExvt0nP1gpVGU7FRWF8A gG5CCzMXH2upkTAMwzD86aefyp6I5WHqIivRWfevF/4NAB1lMpkcHh427OWQ2FSIienb6e4glC9n nr+I8RsAOo3v+2dnZ814ORRG+/HHH7PWhM6h7MvWwBYCvQGg65DkUI7Z5u5Cbs2vv/5a9oUCDSx8 CcwAegOABgRB4Pv+crnchOrQ2gSHh4fPnz+nPbHxg5xXdtKYDXLSQBGgNwBoA6uOUqqWCBspzXA4 zEpFS040YZwb1q8G2BKgNwDox2g0ms/nxd9kE4OSAnq93nA4LDujE4DKQG8A0JXxeLxcLufzuXp/ yYCYCMkMN16iZjKZwDUBDQO9AcAQfN+XUa/5fC5XP6uWdQZAjfwfaF3w8SQmYd0AAAAASUVORK5C YII= ------=_NextPart_000_0000_01C80F9B.F234E740 Content-Type: image/gif Content-Transfer-Encoding: base64 Content-Location: http://www.cs.wisc.edu/wisa/papers/security03/images/language-check.gif R0lGODlhuAA5AIQAAAAAAKioqJycnJCQkISEhHh4eGxsbGBgYFRUVEhISDw8PDAwMCQkJBgYGAwM DLOzswAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACH5BAEA AA8ALAAAAAC4ADkAQAX+4COOZGmeaKqubOu+cCzPdG0LTGDvfL8WCp9QpgAMj8ieAGAoKYImBMKk SJSMKmhp4EDoBIgcCgDQwbAsNOx5kpIGAIf5gSCI3KgAouzsnwIACwRkDWRkciOCIoNxDgMnBwco B49XJ2okZDKYKZwpjIWGiCKKDwILCA5xCXYkpQgJAU1vAAdzlCS4lwxJvSNdvsHCKw5Ww70BDgIk BGIECwsPvCfNOs/RI1OLBQ8E3MdDsODj5OXm5+jp6usuBXzs8CsGDfEye9ovnpf1R3utD2z2OdEi Ao4hQyY4edKnKYa+hEQIZirh4JuIBPRE6CvwDRXEiZceRllGQ1IPfzL+GJichK0fyRYIXq5YKYKm CXc2+alQaUIPpSLT8iD4CSCozh4Nch5tMUCiiaaG/v0JpS3WA6tLTwzAl7Wr169HElgES25BJbI/ WqI9RyDj2hKAuL49l8ApWThqW4gcAWjuiwVurzIwdLbmLBEKBh/E+kcujUJS9brYeyKBYgCFHxw4 TECBLQSFGpzdnCJAg5d6SKRGMQgfnMwaM8OZ0xZ2ic4fVTOknHsF7y2YLxVO1c1y4eAnBEhqEBn3 COcJsbwmgQ/TEtsMDg+IbFSjwOibJseYnm2EmgMSAxh74GksaBPdu5vnfcrFHM0vCUSeseC3q7Eo ADFEfS4Q+IJ+zOz+twgAdnk1iIIsDMBAAwaQNMABDTDmF0zvvHWAAxumYwAAMr2VTInSQEFAEwJQ eAIDKh4WYgoJrDfjA6UAlJ8kGo6gwI4jRNIKLrqspQCANyap5JJMNunkk8PYCCUNBsi4pAGOTTmD affduICVWtqwAJJ+uRimEF/eqEBeZ/JQUYgjQtgmlR2+5UBgLeDJmpRNFuPXiFmq8Jt/c+2BYgox 1aAUfwDIOYZ4Lr2QKAoXBlnCogty9WFy8mHI1wFskmAAhAIkoEoDXnTSqEOQxrCpCTiQsAR6y5yi TayIYjKqqBAOopYDSjWA5JsjGBAqdUh6dimfGhHqnWQyAHuCsCT+uHPYriJQi0JMqyLr7SUg1hRu bzUdxECPU5H7LHisQgvDq+QuEIsUgx32kFUeqRaYaWOgIW2xJHHyLwk1jnBKiQp9d4Wz7LX6wsAP VLiuKXX2NTEJDKhCBpuY7CbuCAWkKUJ3BoxbggAO9OrYQv2G564L8IZ8mFHyGmAGAamO/Nwy24lK IjNc4dwyCwiAiYIAFgkwHA+FbNLlHwyrULSks7QY6tTlUZdHIDA7+kAAly1QBF+BxtCfDMayILIQ B3idi9tkq1Z2a4V2CwMXgYbxdJMovTVrDQYooDEDCNj2JGAbrjnngExsuMShi89g2YwHyIfDUAGA /YgARZBRyeVUtmg+5Ww3inPFVoTrwMoYqOf85IlKopfLSn0d5HlNtKNxSisGSNL7XC3uvSGXpmAU hwKtDDDmycY7gDzIAgh91UVzHVB25CqUakcCq1KrLfbgTxkCADs= ------=_NextPart_000_0000_01C80F9B.F234E740--