Self-Guard: Empower the LLM to Safeguard Itself

TL;DR

Self-Guard addresses the jailbreak problem plaguing LLMs by proposing a two-stage framework that combines safety training and safeguard mechanisms. In stage one, the model's ability to assess harmful content is enhanced; in stage two, the model is instructed to consistently detect harmful content in its own outputs. Experiments indicate Self-Guard is robust against jailbreak attacks and does not degrade general model performance, with a note that some harmful prompts may still yield harmless responses. Sensitivity analyses suggest the approach avoids over-sensitivity and may mitigate it in practice. Overall, Self-Guard offers a practical path to safer LLM behavior without sacrificing capability, by enabling the model to monitor and regulate its own outputs.

Abstract

The jailbreak attack can bypass the safety measures of a Large Language Model (LLM), generating harmful content. This misuse of LLM has led to negative societal consequences. Currently, there are two main approaches to address jailbreak attacks: safety training and safeguards. Safety training focuses on further training LLM to enhance its safety. On the other hand, safeguards involve implementing external models or filters to prevent harmful outputs. However, safety training has constraints in its ability to adapt to new attack types and often leads to a drop in model performance. Safeguards have proven to be of limited help. To tackle these issues, we propose a novel approach called Self-Guard, which combines the strengths of both safety methods. Self-Guard includes two stages. In the first stage, we enhance the model's ability to assess harmful content, and in the second stage, we instruct the model to consistently perform harmful content detection on its own responses. The experiment has demonstrated that Self-Guard is robust against jailbreak attacks. In the bad case analysis, we find that LLM occasionally provides harmless responses to harmful queries. Additionally, we evaluated the general capabilities of the LLM before and after safety training, providing evidence that Self-Guard does not result in the LLM's performance degradation. In sensitivity tests, Self-Guard not only avoids inducing over-sensitivity in LLM but also can even mitigate this issue.