Paper Image

Automated Code Vulnerability Repair with Language Models

Published on:

8 January 2024

Primary Category:

Software Engineering

Paper Authors:

David de-Fitero-Dominguez,

Eva Garcia-Lopez,

Antonio Garcia-Cabot,

Jose-Javier Martinez-Herraiz

Bullets

Key Details

Introduces efficient format to represent code changes for automated repair

Uses fine-tuned Code Llama and Mistral models for higher accuracy

Simplifies modifications across vulnerability types for flexibility

Underscores dataset integrity, avoiding train data in test sets

Sets new standards for automated repair, spurring more research

AI generated summary

Automated Code Vulnerability Repair with Language Models

This study introduces an efficient method to represent code changes for automated repair of vulnerabilities. Fine-tuned large language models like Code Llama and Mistral are used, significantly improving accuracy over previous techniques. The method simplifies changes across vulnerability types. The study underscores using clean test data without train samples for integrity. It contributes to digital security by setting new standards for automated repair using AI, fostering more research.

Answers from this paper

Comments

No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up