1. Static Analysis of Anomalies and Security Vulnerabilities in Executable Files Presented by Jay-Evan Tevis Department of Computer Science Western Illinois University Macomb, Illinois

2. #include int main() { float number; float answer; printf(“Enter a value:”); scanf(“%f”, &number); answer = sqrt(number); printf(“Square root: “); printf(“%f”, answer); return 0; } Source code file (209 bytes) Object code fileExecutable file What is actually inside an executable file? Can the contents of an executable file reveal if secure programming practices were used when developing the software?

3. #include int main() { float number; float answer; printf(“Enter a value:”); scanf(“%f”, &number); answer = sqrt(number); printf(“Square root: “); printf(“%f”, answer); return 0; } Source code file (209 bytes) Object code file (760 bytes) Executable file (13,049 bytes) File Header [20 bytes] Section Table [160 bytes].text section [128 bytes].rdata section [32 bytes] COFF Relocations [110 bytes] Symbol Table [306 bytes] String Table [4 bytes] DOS Header [64 bytes] MS-DOS Stub [57 bytes] PE Signature [4 bytes] File Header [20 bytes] Optional Header [224 bytes] Section Table [200 bytes].text section [2048 bytes].data section [512 bytes].rdata section [512 bytes].idata section [439 bytes] Symbol Table [6138 bytes] String Table [1791 bytes] Answer: There is much more than just binary code in an executable file. The contents of the file contains a wealth of information on anomalies and security vulnerabilities.

4. 4 Research Thesis A methodology can be devised that uses information located in the headers, sections, and tables of an executable file, along with information derived from the overall contents of the file, as a means to detect specific software security vulnerabilities without having to disassemble the code Such a methodology can be instantiated in a software utility program that automatically detects certain software security vulnerabilities before ever installing and running the executable file

5. 5 Summary of Research Objectives 1)Do an in-depth study of the PE format of an executable file to discover any information that could be useful in detecting software security vulnerabilities 2)Create a software utility to dissect a PE file byte-by-byte from beginning to end (without disassembling) in order identify, map, and categorize its contents 3)Formulate a methodology for combining and correlating the information found in a PE file in an effort to detect indicators of certain security vulnerabilities 4)Incorporate the methodology into the PE dissecting utility in order to automatically detect software security vulnerabilities in a matter of seconds 5)Test the automated methodology on installation files, software development files, Windows XP operating system files, Microsoft application files, security-centric application files, and other application files 6)Analyze the test results and make conclusions

6. 6 Overview of Presentation The PE File Format A Software Utility for Dissecting a PE File A Methodology to Detect Indicators of Certain Security Vulnerabilities Methodology + PE Dissecting Utility = findssv Testing the Automated Methodology Conclusion and Future Work

7. The PE File Format

8. 8 PE and COFF Formats PE: Portable Executable – Common format created by Microsoft and used by all 32-bit Windows NT executable files (.exe,.dll,.ocx,.sys,.drv, etc.) –PE files are called image files in the Microsoft documentation (e.g., binary image) COFF: Common Object File Format –MS-DOS object code format created by Microsoft and used by many compilers when producing object code files –COFF is a subset of the PE format

9. 9 Typical Contents of a PE file DOS header MS-DOS stub PE signature File header Optional header Section table Symbol table String table Various sections (.text,.data, etc.) Import table Export table

10. 10 helloworld.exe (Cygwin) (1 of 2) ADDRESS DESCRIPTION 0 +-------------------------------------------------------------+ DOS Header [64 bytes] 63 +-------------------------------------------------------------+ 64 +-------------------------------------------------------------+ MS-DOS Stub [57 bytes] 120 +-------------------------------------------------------------+ 121 +-------------------------------------------------------------+ (Contents not known) [7 bytes] 127 +-------------------------------------------------------------+ 128 +-------------------------------------------------------------+ PE Signature [4 bytes] 131 +-------------------------------------------------------------+ 132 +-------------------------------------------------------------+ File Header [20 bytes] 151 +-------------------------------------------------------------+ 152 +-------------------------------------------------------------+ Optional Header [224 bytes] 375 +-------------------------------------------------------------+ 376 +-------------------------------------------------------------+ Section Table [200 bytes] 575 +-------------------------------------------------------------+ 576 +-------------------------------------------------------------+ (** Zero-filled region **) [448 bytes] 1023 +-------------------------------------------------------------+ 1024 +-------------------------------------------------------------+.text section [176128 bytes] 177151 +-------------------------------------------------------------+

11. 11 helloworld.exe (Cygwin) (2 of 2) ADDRESS DESCRIPTION 177152 +-------------------------------------------------------------+.data section [4608 bytes] 181759 +-------------------------------------------------------------+ 181760 +-------------------------------------------------------------+.rdata section [35840 bytes] 217599 +-------------------------------------------------------------+ 217600 +-------------------------------------------------------------+.idata section [1183 bytes] 218782 +-------------------------------------------------------------+ 218783 +-------------------------------------------------------------+ More of Import Table (.idata section) [241 bytes] 219023 +-------------------------------------------------------------+ 219024 +-------------------------------------------------------------+ (** Zero-filled region **) [112 bytes] 219135 +-------------------------------------------------------------+

12. A Software Utility for Dissecting a PE File

13. 13 Currently-available PE Dump Utilities First considered using –objdump from Cygwin –tdump from Borland –dumpbin from Microsoft –pedump from Pietrek All provide PE format information that has been first filtered and then text formatted; some output is extraneous data All appear to rely on the values in the header and look-up tables to locate and display section and table contents instead of looking directly at what is actually in the tables Comparing results provided by each revealed inconsistencies in the section table and import table information None provide a complete byte-for-byte account of the file contents to find "unmapped regions" or compressed file storage Some abort on PE files containing non-typical information

14. 14 A PE File Dissecting Utility Built a PE file dissecting utility Accepts program options from the command line Works on a single PE or COFF file Written in C++; 2700 source lines of code Made up of a driver module, a utility module and 16 classes –Modeled the classes after the PE and COFF formats

15. A Methodology to Detect Indicators of Certain Security Vulnerabilities

16. 16 Overview of the Methodology Derived from an in-depth search for security indicators gleaned from the PE file format Describes how to statically analyze a PE file Categorizes its findings as facts, anomalies, and vulnerabilities Consists of four steps

17. 17 Step 1: Read through the complete file and create a file fact summary ==== File Fact Summary ==== - Image file in Windows NT portable executable (PE) format - Actual file size: 7168 bytes - Created on Tue Feb 1 16:01:34 2005 - Target CPU: Intel 386 or later compatibles - Targeted for a 32-bit-word architecture - Debugging information has been removed - Designed for Windows Operating System version 4.0 - Runs in the Windows character subsystem - Lists these table names in the data directory: Import - Contains no string table - Contains no symbol table - Imports functions from -- cygwin1.dll (CYGWIN GNU base dynamic link library) -- KERNEL32.dll (WinNT base API client) ==== End of File Fact Summary ==== Sample C-1: vulnerable.exe (g++)

18. 18 Step 2: Detect anomalies when reading the headers, tables, and sections in the file **** Anomalies **** - A section entry named.bss appears in the section table, but the table doesn't contain the location of the 226272 bytes for that section - The data directory table in the optional header states that the Resource Table (.rsrc section) is 1104 bytes in size when actually it is 1536 bytes in size - The data directory table in the optional header states that the Relocation Table (.reloc section) is 43792 bytes in size when actually it is 44032 bytes in size **** End of Anomalies **** Sample H-1: cygwin1.dll (g++)

19. 19 Step 3: Detect anomalies when mapping the file contents 135168 +-------------------------------------------------------------+.rsrc section [10360352 bytes] 10495519 +-------------------------------------------------------------+ 135168 +-----------------------------------------------------+ (No additional details) [10346544 bytes] 10481711 +-----------------------------------------------------+ 10481712 +-----------------------------------------------------+ Certificate Table [5616 bytes] 10487327 +-----------------------------------------------------+ 10487328 +-----------------------------------------------------+ (No additional details) [8192 bytes] 10495519 +-----------------------------------------------------+ Sample K-3: Real Player installation file - Contains an unusual area of 10346544 bytes starting at address 135168 which may indicate a group of compressed files

20. 20 Step 4: Detect software security vulnerabilities (see next three slides)

21. 21 Step 4a: Detect sections that are both writable and executable Sample H-2: cygwin1.dll (g++) !!!! Security Vulnerabilities and Risks!!!! - Has a section named.advapi32_text whose contents can be both written to and executed - Has a section named.netapi32_text whose contents can be both written to and executed - Has a section named.ntdll_text whose contents can be both written to and executed - Has a section named.psapi_text whose contents can be both written to and executed - Has a section named.secur32_text whose contents can be both written to and executed - Has a section named.user32_text whose contents can be both written to and executed - Has a section named.wsock32_text whose contents can be both written to and executed - Has a section named.ws2_32_text whose contents can be both written to and executed - Has a section named.iphlpapi_text whose contents can be both written to and executed - Has a section named.ole32_text whose contents can be both written to and executed - Has a section named.kernel32_text whose contents can be both written to and executed - Has a section named.winmm_text whose contents can be both written to and executed !!!! End of Security Vulnerabilities and Risks!!!!

22. 22 Step 4b: Detect large unused zero- filled regions (over 50 bytes) Sample L-2: oc30.dll (Windows system file) FILE NAME: OC30.DLL !!!! Security Vulnerabilities and Risks!!!! - Contains 9228 bytes of unused zero-filled space that could be used to store malicious code or data !!!! End of Security Vulnerabilities and Risks!!!!

23. 23 Step 4c: Detect the use of vulnerable C library functions Sample L-2: oc30.dll (Windows system file) FILE NAME: OC30.DLL !!!! Security Vulnerabilities and Risks!!!! - Uses 5 standard C functions susceptible to buffer overflow attacks: fgets (Low risk), memcpy (Low risk), strcat (Very high risk), strcpy (Very high risk), vsprintf (Very high risk) !!!! End of Security Vulnerabilities and Risks!!!!

24. 24 List of Vulnerable Functions in the Standard C Library Ultra High Risk: gets Very High Risk: strcpy, strcat, sprintf, scanf, sscanf, fscanf, vfscanf, vsprintf, vscanf, vsscanf, streadd, strecpy, strtrns, realpath, syslog, getopt, getopt_long, getpass Medium Risk: getchar, fgetc, getc, read Low Risk: bcopy, fgets, memcpy, snprintf, strccpy, strcadd, strncpy, vsnprintf (Source: Viega and McGraw,"Building Secure Software", 2002, pp. 152-153)

25. 25 If the symbol or import table is missing, our methodology cannot detect the use of functions vulnerable to buffer overflow attacks Symbol table can be stripped when a file is linked Import table is not needed if function definitions are in the PE file rather than a separate DLL The Consequences of No Symbol Table or Import Table An empty list of vulnerable function names only indicates the lack of enough information to detect any function names at all; it does not indicate that no vulnerable functions are used

26. Methodology + PE Dissecting Utility = findssv “find software security vulnerabilities”

27. 27 Features of findssv Written in C++; command-line interface; 3800 SLOC Executable size –Borland: 1,150,000 Cygwin: 660,000 Microsoft: 412,000 Performs a static security analysis of object code or image files stored in the COFF or PE format Detects and reports facts and certain anomalies and software security vulnerabilities about the file Also provides the ability to search, display and change any range of bytes in the file in byte, character, string, unicode, word, or double word format Accepts command line options Can be used on a single file, a select group of files, or a whole directory Scans scores of files in only a few seconds Produces an optional trace showing each stage of the analysis and the results obtained from it

28. 28 Sample Uses of findssv –Display the file map of helloworld.exe c:\>findssv helloworld.exe –MZ –Detect anomalies and vulnerabilities in Microsoft Word c:\>findssv winword.exe –Display the contents of each PE format part in Secure CRT c:\>findssv securecrt.exe –P –Detect anomalies and vulnerabilities in the DLL files located in the Windows System32 directory c:\>findssv “C:\windows\system32\*.dll” –Create a trace output while reading through cygwin1.dll c:\>findssv cygwin1.dll -T –Find all UNICODE strings of length four or more in a file c:\>findssv file.exe 4 0 100000

29. Testing the Automated Methodology

30. 30 Testing Approach 1)Identified seven categories of files to test 2)Identified PE files for each category (2725 total files) 3)Ran findssv against a single file, a group of files, or a whole directory depending on the location and configuration of the files stored on the hard drive 4)Saved the report data generated by findssv using command line redirection of the standard output file 5)Manually analyzed the report data and summarized the results in seven separate tables corresponding to each category

31. 31 Categories of Test Files Specific Example Files Executable installation files Software development files Windows XP operating system files Microsoft application files Security-centric application files Miscellaneous application files

32. 32 Specific Example Files (Key Findings 1 of 2) The Cygwin Gnu C++ compiler and linker injected seven vulnerable function calls into the executable program of findssv A program compiled and linked using the Cygwin Gnu tools will have standard C functions in it that are susceptible to buffer overflow attacks even when these functions are not explicitly used by the software developer The cygwin1.dll file contains security vulnerabilities that allow executable code to be modified after the program is loaded into memory and executed

33. 33 Specific Example Files (Key Findings 2 of 2) The kernel32.dll file contains the use of functions that are susceptible to buffer overflow attacks An executable file will reveal less information about the functions it uses by having its symbol table stripped and by using fewer DLLs It may be possible to analyze the general layout of the sections and tables in a file map of an executable file in order to detect a pattern that indicates the compiling and linking tools used to generate the file

34. 34 Executable Installation Files (Key Findings) An image file can take advantage of the flexibility of the PE format and serve as its own storehouse for millions of bytes of data What is in the pdf995s.exe installation file?

35. 35 pdf995s.exe (1 of 2) ADDRESS DESCRIPTION 0 +-------------------------------------------------------------+ DOS Header [64 bytes] 63 +-------------------------------------------------------------+ 64 +-------------------------------------------------------------+ MS-DOS Stub [57 bytes] 120 +-------------------------------------------------------------+ 121 +-------------------------------------------------------------+ (Contents not known) [79 bytes] 199 +-------------------------------------------------------------+ 200 +-------------------------------------------------------------+ PE Signature [4 bytes] 203 +-------------------------------------------------------------+ 204 +-------------------------------------------------------------+ File Header [20 bytes] 223 +-------------------------------------------------------------+ 224 +-------------------------------------------------------------+ Optional Header [224 bytes] 447 +-------------------------------------------------------------+ 448 +-------------------------------------------------------------+ Section Table [200 bytes] 647 +-------------------------------------------------------------+ 648 +-------------------------------------------------------------+ (** Zero-filled region **) [376 bytes] 1023 +-------------------------------------------------------------+ 1024 +-------------------------------------------------------------+.text section [22016 bytes] 23039 +-------------------------------------------------------------+

36. 36 pdf995s.exe (2 of 2) ADDRESS DESCRIPTION 23040 +-------------------------------------------------------------+.rdata section [3072 bytes] 26111 +-------------------------------------------------------------+ 23040 +-----------------------------------------------------+ (No additional details) [492 bytes] 23531 +-----------------------------------------------------+ 23532 +-----------------------------------------------------+ Import Table (.idata section) [2303 bytes] 25834 +-----------------------------------------------------+ 25835 +-----------------------------------------------------+ (No additional details) [181 bytes] 26015 +-----------------------------------------------------+ 26016 +-----------------------------------------------------+ Export Table (.edata section) [50 bytes] 26065 +-----------------------------------------------------+ 26066 +-----------------------------------------------------+ (No additional details) [46 bytes] 26111 +-----------------------------------------------------+ 26112 +-------------------------------------------------------------+.data section [3584 bytes] 29695 +-------------------------------------------------------------+ 29696 +-------------------------------------------------------------+.rsrc section [1536 bytes] 31231 +-------------------------------------------------------------+ 31232 +-------------------------------------------------------------+ _winzip_ section [1329152 bytes] 1360383 +-------------------------------------------------------------+ 1.3MB of data

37. 37 Software Development Files (Key Findings 1 of 2) The Cygwin software development files and utility programs contain scores of security vulnerabilities; therefore, we do not recommend them for secure programming activities 42 of the 56 Sun Microsystems Java software development files that we tested (including the Java interpreter) contained the use of one or more functions that are susceptible to buffer overflow attacks 8 of the 32 Microsoft Visual Studio SDK files that we tested (including the MSIL assembler) contained the use of one or more functions that are susceptible to buffer overflow attacks

38. 38 Software Development Files (Key Findings 2 of 2) 13 of the 21 Microsoft Visual Studio C/C++ 7.0 files that we tested (including the compiler and linker) contained the use of one or more functions that are susceptible to buffer overflow attacks

39. 39 Windows XP Operating System Files (Key Findings) In the Windows XP Home Edition system32 directory, approximately 25% of the executable files and DLLs use one or more standard C functions that are susceptible to buffer overflow attacks

40. 40 Microsoft Application Files (Key Findings) (Good) Only 3 of the 48 Microsoft Office DLLs have four or more security vulnerabilities (Good) Of the 15 NetMeeting DLLs, only one file has a security vulnerability; it uses only one low risk standard C function susceptible to buffer overflow attacks

41. 41 Security-centric Application Files (Key Findings) In the Network Associates Common Framework software installation that accompanies the VirusScan software installation, approximately 75% of the executable files and DLLs use one or more standard C functions that are susceptible to buffer overflow attacks The SecureCRT 4.0 software contains executable files and DLLs that are highly vulnerable to buffer overflow attacks In the Zero Knowledge Freedom software, approximately 70% of the DLLs contain the use of standard C functions that are susceptible to buffer overflow attacks In the Zone Alarm Pro 4.0 software, all three of the DLLs contain the use of standard C functions that are susceptible to buffer overflow attacks

42. 42 Miscellaneous Application Files (Key Findings 1 of 2) Earthlink TotalAccess: 42 of the 54 DLLs contain security vulnerabilities MusicMatch Jukebox: 29 of the 44 DLLs contain security vulnerabilities; 11 of those contain four or more vulnerabilities OpenOffice: 50 of the 193 DLLs contain security vulnerabilities; 12 of those have four or more vulnerabilities Real One Player: 14 of the 15 DLLs contain security vulnerabilities

43. 43 Miscellaneous Application Files (Key Findings 2 of 2) “These key findings indicate a major lack of secure programming practices by the people who developed the DLLs for these miscellaneous applications. This is in sharp contrast to the very low number of security vulnerabilities detected by findssv in the DLLs of the Windows application files. However, this high number of vulnerabilities corresponds closely to the large number of vulnerabilities found in the executable files and the DLLs in the Windows System32 directory.”

44. Conclusion and Future Work

45. 45 Research Thesis This research demonstrated the following: –A methodology can be devised that uses information located in the headers, sections, and tables of an executable file, along with information derived from the overall contents of the file, as a means to detect specific software security vulnerabilities without having to disassemble the code –Such a methodology can be instantiated in a software utility program that automatically detects certain software security vulnerabilities before ever installing and running the executable file

46. 46 Immediate Practical Uses of findssv It quickly pares down a group of executable files to the ones in which secure programming was not an objective of the software developers It can do in seconds what could take a security analyst days or weeks to do using hex editors and file dump utilities –It knows what to look for and where to look for it in the PE format –It knows when to stop looking when specific security vulnerability indicators are not present Scan for Vulnerabilities…

47. 47 Future Work Determine the compiler and linker used to build an executable file Establish the relationship between DLL function use and program purpose Provide more details on unknown regions Reveal the names of files stored in compressed file regions Detect the use of standard C functions by way of function call signatures searched for in the code sections of a PE file

48. Static Analysis of Anomalies and Security Vulnerabilities in Executable Files Presented by Jay-Evan Tevis Department of Computer Science Western Illinois University Macomb, Illinois www.wiu.edu/users/jjt107 Questions?