da/sec scientific talk on memory forensics

Topic: Approxis: a first step towards fast data reduction and cross validation in the field of memory forensics

by Lorenz Liebler
D19/2.03a, October 19, 2017 (Thursday), 12.00 noon

Keywords — approximate disassembly, approximate matching, disassembly, binary analysis, memory forensics

Abstract

The discipline of detecting known and unknown code structures in large sets of data is a challenging task. An example could be the examination of memory dumps of an infected system. Memory forensic frameworks rely on system relevant information and the examination of structures which are located within a dump itself. With the constant increasing size of used memory, the creation of additional methods of data reduction (similar to those in disk forensics) are eligible. In the field of disk forensics, approximate matching algorithms are well known. However, in the field of memory forensics, the application of those algorithms is impractical. In this paper we introduce approxis: an approximate disassembler. In contrary to other disassemblers our approach does not rely on an internal disassembler engine, as the system is based on a compressed set of ground truth x86 and x86-64 assemblies. Our first prototype shows a good computational performance and is able to detect code in large sets of raw data. Additionally, our current implementation is able to differentiate between architectures while disassembling. Summarized, approxis is the first attempt to interface approximate matching with the field of memory forensics.