SafeDecoding: Defending against Jailbreak Attacks via Safety-Aware Decoding

Zhangchen Xu, Fengqing Jiang, Luyao Niu, Jinyuan Jia, Bill Yuchen Lin, Radha Poovendran

Keywords: Large language models, jailbreak attacks, decoding strategy, safety

TL;DR: This paper introduces a computationally lightweight and effective decoding strategy, SafeDecoding, for LLMs to defend against jailbreak attacks.

Abstract:

As large language models (LLMs) become increasingly integrated into real-world applications, extensive efforts have been made to align LLM behavior with human values, including safety. Jailbreak attacks, which aim to provoke unintended and unsafe behaviors from LLMs, remain a significant LLM safety threat. We analyze tokens, which are the smallest unit of text that can be processed by LLMs and make the following observations: (1) probabilities of tokens representing harmful responses are higher than those of harmless responses, and (2) responses containing safety disclaimers appear among the top tokens when token probabilities are sorted in descending order.
In this paper, we leverage (1) and (2) to develop SafeDecoding, a safety-aware decoding strategy for LLMs, to defend against jailbreak attacks. We perform extensive experiments to evaluate SafeDecoding against six SOTA jailbreak attacks on five LLMs using four benchmark datasets. Our results show that SafeDecoding significantly reduces attack success rate and harmfulness of jailbreak attacks without compromising the helpfulness of responses to benign user queries while outperforming six defense methods.

Submission Number: 61