Experiences in malware binary deobfuscation presented at Virus Bulletin 2010

by Hassen Saidi (SRI International),

URL : http://www.virusbtn.com/conference/vb2010/abstracts/Saidi-etal.xml

Summary : "
Malware authors employ a myriad of evasion techniques to impede automated reverse engineering and static analysis efforts.
The most popular technologies include 'code obfuscators' which serve to rewrite the original binary code to an equivalent
form that provides identical functionality while defeating signature-based detection systems. These systems significantly
complicate static analysis, making it challenging to uncover the malware intent and the full spectrum of embedded
capabilities. While code obfuscation techniques are commonly integrated into contemporary commodity packers, from the
perspective of a reverse engineer, deobfuscation is often a necessary step that must be conducted independently after
unpacking the malware binary.
In this paper, we describe a set of techniques for automatically unrolling the impact of code obfuscators with the
objective of completely recovering the original malware logic. We have implemented a set of generic debofuscation rules
as a plug-in for the popular IDA Pro disassembler. We use sophisticated obfuscation strategies employed by two infamous
malware instances from 2009, Conficker C and Hydraq (the binary associated with the Aurora attack) as case studies.
In both instances our deobfuscator was able to enable a complete decompilation of the underlying code logic. This work
was instrumental in the comprehensive reverse engineering of the heavily obfuscated P2P protocol embedded in the Conficker
worm. The plug-in is integrated with the Hexrays decompiler to provide a complete reverse engineering of malware binaries
from binary form to C code and is available for free download on the SRI malware threat center website:
http://www.mtc.sri.com/deobfuscation/.
"