Loading...

Intelligent Model for Vulnerability Detection and Firmware Binary Code Testing

Faghani, Paya | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57621 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Jahangir, Amir Hossein
  7. Abstract:
  8. The correct functioning of the systems in the industry depends on the correct functioning of their firmware. It is important to ensure the security of these firmware and guarantee that they have no vulnerabilities in order to prevent attackers from infiltrating the systems. Due to the lack of firmware source code, researchers have always tried to assess the security of firmware and other programs by analyzing their binary codes. Vulnerabilities in applications allow attackers to increase their access rights or make the system unavailable. The two main approaches for binary code analysis are static and dynamic analysis. The need for industrial emulators, which are difficult to design due to the need for simulating interrupt mechanisms caused by peripheral devices, is the basis for dynamic analysis. Therefore, static analysis is considered a more suitable method for analyzing industrial firmware. Static analysis is time-consuming due to the large volume of code. Hence, developing automated solutions for vulnerability detection is crucial for industrial firmware. Moreover, firmware developers use techniques such as obfuscation and symbol table removal to prevent reverse engineering. These methods make accurately converting binary programs into human-readable pseudocode challenging. Finding an automated method to detect vulnerabilities without the need to convert firmware to assembly language is the main goal of this study. In this research, we have developed an intelligent model successful in detecting firmware vulnerabilities, in particular, buffer overflows, despite the lack of labeled data without converting the program to assembly language. The proposed model for binary ELF files has reached 85.96% accuracy without the need to parse the file into assembly language and by converting them to images
  9. Keywords:
  10. Static Analysis ; Firmware ; Semi-supervised Generative Adversarial Networks ; Buffer Overflow

 Digital Object List

 Bookmark

  • Part1
  • ApprovalSheet
  • Part3