Adaptive Defense Mechanism Against Jailbreak Attacks for Secure Deployments
A novel defense strategy, MirrorGuard, has been proposed to enhance the security of large language models (LLMs) against jailbreak attacks. This approach introduces a dynamic…