SEAS: Self-Evolving Adversarial Safety Optimization for Large Language Models
This addresses the critical issue of evolving vulnerabilities in LLMs for developers and users, offering an automated approach to enhance safety, though it appears incremental as it builds on existing adversarial methods.
The paper tackles the problem of ensuring security and preventing harmful outputs in large language models (LLMs) by introducing the SEAS framework, which uses self-evolving adversarial optimization to automatically generate adversarial prompts, resulting in a target model achieving security comparable to GPT-4 and a red team model with increased attack success rate after three iterations.
As large language models (LLMs) continue to advance in capability and influence, ensuring their security and preventing harmful outputs has become crucial. A promising approach to address these concerns involves training models to automatically generate adversarial prompts for red teaming. However, the evolving subtlety of vulnerabilities in LLMs challenges the effectiveness of current adversarial methods, which struggle to specifically target and explore the weaknesses of these models. To tackle these challenges, we introduce the $\mathbf{S}\text{elf-}\mathbf{E}\text{volving }\mathbf{A}\text{dversarial }\mathbf{S}\text{afety }\mathbf{(SEAS)}$ optimization framework, which enhances security by leveraging data generated by the model itself. SEAS operates through three iterative stages: Initialization, Attack, and Adversarial Optimization, refining both the Red Team and Target models to improve robustness and safety. This framework reduces reliance on manual testing and significantly enhances the security capabilities of LLMs. Our contributions include a novel adversarial framework, a comprehensive safety dataset, and after three iterations, the Target model achieves a security level comparable to GPT-4, while the Red Team model shows a marked increase in attack success rate (ASR) against advanced models. Our code and datasets are released at https://SEAS-LLM.github.io/.