C-ing Clearly: Enhanced Binary Code Explanations using C code

TL;DR

This paper introduces C-ing Clearly, a synthetic data generation method that uses C code to improve large language models' understanding of assembly, leading to better binary code explanations and vulnerability detection.

Contribution

The paper presents a novel data augmentation technique leveraging C code to enhance LLM performance on low-level programming tasks.

Findings

Improved LLM performance in binary code summarization

Enhanced vulnerability detection accuracy

Consistent gains across various LLM models

Abstract

Large Language Models (LLMs) typically excel at coding tasks involving high-level programming languages, as opposed to lower-level programming languages, such as assembly. We propose a synthetic data generation method named C-ing Clearly, which leverages the corresponding C code to enhance an LLM's understanding of assembly. By fine-tuning on data generated through our method, we demonstrate improved LLM performance for binary code summarization and vulnerability detection. Our approach demonstrates consistent gains across different LLM families and model sizes.