This research introduces a Chain-of-Understanding approach that leverages Large Language Models to guide developers through hierarchical code comprehension, making code auditing more accessible to end-users with limited expertise.

• Creates a structured pathway from high-level overviews to specific functions
• Developed based on formative interviews with experienced code auditors
• Supports both novice and experienced developers in understanding complex codebases
• Addresses a critical challenge in software engineering and security

For engineering teams, this approach offers a promising framework to improve code review processes, knowledge transfer, and overall code quality—potentially reducing the expertise barrier for effective code auditing.

Chain of Understanding: Supporting Code Understanding with Large Language Models