A system for dynamic slicing and program visualization

Abstract

Dynamic analysis, which involves gathering information about a program as it is executing, is becoming increasingly common as reverse engineers attempt to more efficiently analyze complex software systems. In this thesis, I implement a technique called dynamic slicing, which determines how values in a program's register or memory, such as strings, are computed as a function of the program's initial state. This technique is then evaluated on sets of programs of interest to reverse engineers: string generation and string encoding/decoding algorithms. A common source of these algorithms is malware, which often employs these mechanisms to obfuscate and to make human-driven reverse engineering more difficult. In particular, malware will use Domain Generation Algorithms (DGAs) to construct seemingly randomized domain names to contact their control servers. We demonstrate that metrics and graphs produced with dynamic slicing can be successfully employed on these algorithm classes, elucidating interesting features from different families of malware and reducing the manual workload of malware reverse engineers.