LEADER is a novel framework that harnesses large language models to intelligently debloat software, preserving essential functionality while reducing security vulnerabilities.

• Reduces attack surface by removing unnecessary features without breaking functionality
• Employs LLM-guided analysis to identify code that can be safely removed
• Achieves superior debloating performance compared to traditional methods
• Demonstrates tangible security improvements with minimal functionality loss

This research represents a significant advancement in security engineering by addressing the growing issue of software bloat, which creates larger attack surfaces and increases vulnerability to exploitation.

Large Language Models-Aided Program Debloating