OCFI: Make Function Entry Identification Hard Again

PROCEEDINGS OF THE 32ND ACM SIGSOFT INTERNATIONAL SYMPOSIUM ON SOFTWARE TESTING AND ANALYSIS, ISSTA 2023(2023)

引用 0|浏览17
暂无评分
摘要
Function entry identification is a crucial yet challenging task for binary disassemblers that has been the focus of research in the past decades. However, recent researches show that call frame information (CFI) provides accurate and almost complete function entries. With the aid of CFI, disassemblers have significant improvements in function entry detection. CFI is specifically designed for efficient stack unwinding, and every function has corresponding CFI in x64 and aarch64 architectures. Nevertheless, not every function and instruction unwinds the stack at runtime, and this observation has led to the development of techniques such as obfuscation to complicate function detection by disassemblers. We propose a prototype of OCFI to obfuscate CFI based on this observation. The goal of OCFI is to obstruct function detection of popular disassemblers that use CFI as a way to detect function entries. We evaluated OCFI on a large-scale dataset that includes real-world applications and automated generation programs, and found that the obfuscated CFI was able to correctly unwind the stack and make the detection of function entries of popular disassemblers more difficult. Furthermore, on average, OCFI incurs a size overhead of only 4% and nearly zero runtime overhead.
更多
查看译文
关键词
function entry detection,obfuscation,binary disassembly
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要