Department of Computer Science: MSc Thesis Presentations

Utilizing large language models in reverse engineering: Annotating decompiled binary code

Author: Roni Kirla

Supervisor: Lachlan Gunn

Abstract: Software reverse engineering is a difficult and time-consuming task, where the analyst tries to recover information about the functionality and inner workings of a program without access to its source code. Typically, the main source of information is the binary code of the program. In this thesis, we try to leverage LLMs to assist in this process. We design and implement a tool to this end as an extension to the software reverse engineering framework Ghidra that improves upon the existing alternatives. Our evaluation shows that providing information about the calling context of the function under analysis to the LLM improves the results by a significant margin.