🤖 AI Summary
This work systematically evaluates GPT-3.5, GPT-4, Code Llama, and CodeGeeX2 on end-to-end secure Python code generation, focusing on vulnerability detection, self-repair, and cross-model repair capabilities. Results show that over 75% of generated code contains security vulnerabilities in SecurityEval, with widespread “self-repair blindness”—i.e., models consistently fail to detect and correct critical security flaws they themselves introduce. Method: We propose a lightweight iterative repair framework integrating semantic analysis with multi-turn prompt engineering, requiring no external tools or model fine-tuning. Contribution/Results: Our approach significantly improves repair success rates across all models—from 33.2%–59.6% to 65.9%–85.5%. This is the first study to quantitatively characterize the security repair bottlenecks of mainstream code large language models and to provide a scalable, plug-and-play enhancement paradigm for secure code synthesis.
📝 Abstract
The rapid advancement of large language models (LLMs) such as GPT-4 has revolutionized the landscape of software engineering, positioning these models at the core of modern development practices. As we anticipate these models to evolve into the primary and trustworthy tools used in software development, ensuring the security of the code they produce becomes paramount. How well can LLMs serve as end-to-end secure code producers? This paper presents a systematic investigation into LLMs' inherent potential to generate code with fewer vulnerabilities. Specifically, We studied GPT-3.5 and GPT-4's capability to identify and repair vulnerabilities in the code generated by four popular LLMs including themselves (GPT-3.5, GPT-4, Code Llama, and CodeGeeX2). By manually or automatically reviewing 4,900 pieces of code, our study reveals that: (1) large language models lack awareness of scenario-relevant security risks, which leads to the generation of over 75% vulnerable code on the SecurityEval benchmark; (2) LLMs such as GPT-3.5 and GPT-4 are unable to precisely identify vulnerabilities in the code they generated; (3) GPT-3.5 and GPT-4 can achieve 33.2%~59.6% success rates in repairing the insecure code produced by the 4 LLMs, but they both perform poorly when repairing self-produced code, indicating self-repair"blind spots". To address the limitation of a single round of repair, we developed a lightweight tool that prompts LLMs to construct safer source code through an iterative repair procedure based on the insights gained from our study. Experiments show that assisted by semantic analysis engines, our tool significantly improves the success rates of repair to 65.9%~85.5%.