Shingled Graph Disassembly: Finding the Undecideable Path

摘要: A probabilistic finite state machine approach to statically disassembling x86 language programs is presented and evaluated. Static disassembly a crucial prerequisite for software reverse engineering, has many applications in computer security binary analysis. The general problem provably undecidable because of the heavy use unaligned instruction encodings dynamically computed control flows architecture. Limited work learning data mining been undertaken on this subject. This paper shows that semantic meanings opcode sequences can be leveraged infer similarities between groups operand sequences. empowers learn statistically significant training corpus disassemblies. demonstrate statistical significance opcodes operands surrounding context, facilitating more accurate new binaries. Empirical results algorithm efficient effective than comparable approaches used by state-of-the-art tools.