DELMAN: Dynamic Defense Against Large Language Model Jailbreaking with Model Editing

TL;DR

DELMAN introduces a precise, dynamic model editing technique to defend large language models against jailbreak attacks, effectively neutralizing harmful behaviors while maintaining overall utility and adapting to new threats.

Contribution

The paper presents DELMAN, a novel model editing approach that offers targeted, efficient, and adaptable defense against jailbreak attacks without degrading model performance.

Findings

DELMAN outperforms baseline defenses in mitigating jailbreaks.

It maintains model utility on general tasks.

DELMAN adapts effectively to new attack instances.

Abstract

Large Language Models (LLMs) are widely applied in decision making, but their deployment is threatened by jailbreak attacks, where adversarial users manipulate model behavior to bypass safety measures. Existing defense mechanisms, such as safety fine-tuning and model editing, either require extensive parameter modifications or lack precision, leading to performance degradation on general tasks, which is unsuitable to post-deployment safety alignment. To address these challenges, we propose DELMAN (Dynamic Editing for LLMs JAilbreak DefeNse), a novel approach leveraging direct model editing for precise, dynamic protection against jailbreak attacks. DELMAN directly updates a minimal set of relevant parameters to neutralize harmful behaviors while preserving the model's utility. To avoid triggering a safe response in benign context, we incorporate KL-divergence regularization to ensure the updated model remains consistent with the original model when processing benign queries. Experimental results demonstrate that DELMAN outperforms baseline methods in mitigating jailbreak attacks while preserving the model's utility, and adapts seamlessly to new attack instances, providing a practical and efficient solution for post-deployment model protection.