Enabling Sophisticated Analysis of x86 Binaries with RevGen

Current state-of-the-art static analysis tools for binary software operate on ad-hoc intermediate representations (IR) of the machine code. Therefore, even though IRs facilitate program analysis by abstracting away the source language, it is hard to reuse existing implementations of analysis tools in new endeavors. Recently, a new compiler framework — LLVM— has emerged, together with many analysis tools that use its IR. However, these tools rely on a compiler to generate the IR from source code. We propose RevGen, a tool that automatically converts existing binary programs to the standard LLVM IR, making an increasingly large number of static and dynamic analysis frameworks, as well as run-time instrumentation tools, applicable to legacy software. We show the potential of RevGen by converting several programs and device drivers to LLVM and checking the resulting code with off-the-shelf analysis tools.

Presented at:
7th Workshop on Hot Topics in System Dependability (HotDep), Hong Kong, China, 2011

 Record created 2011-05-24, last modified 2018-01-28

External link:
Download fulltext
Rate this document:

Rate this document:
(Not yet reviewed)