HSF: Defending against Jailbreak Attacks with Hidden State Filtering

TL;DR

This paper introduces a Hidden State Filter (HSF) mechanism that preemptively detects and rejects jailbreak prompts in LLMs by analyzing hidden state patterns, significantly improving defense effectiveness with minimal impact on benign queries.

Contribution

The paper presents a novel pre-inference defense method using hidden state analysis, outperforming existing techniques in resisting diverse jailbreak attacks with low overhead.

Findings

HSF significantly reduces jailbreak success rates

Minimal impact on benign query responses

Outperforms existing defense baselines

Abstract

With the growing deployment of LLMs in daily applications like chatbots and content generation, efforts to ensure outputs align with human values and avoid harmful content have intensified. However, increasingly sophisticated jailbreak attacks threaten this alignment, aiming to induce unsafe outputs. Current defense efforts either focus on prompt rewriting or detection, which are limited in effectiveness due to the various design of jailbreak prompts, or on output control and detection, which are computationally expensive as they require LLM inference. Therefore, designing a pre-inference defense method that resists diverse jailbreak prompts is crucial for preventing LLM jailbreak attacks. We observe that jailbreak attacks, safe queries, and harmful queries exhibit different clustering patterns within the LLM's hidden state representation space. This suggests that by leveraging the LLM's hidden state representational capabilities, we can analyze the LLM's forthcoming behavior and proactively intervene for defense. In this paper, we propose a jailbreak attack defense strategy based on a Hidden State Filter (HSF), a lossless architectural defense mechanism that enables the model to preemptively identify and reject adversarial inputs before the inference process begins. We activate its defensive potential through an additional plugin module, effectively framing the defense task as a classification problem. Experimental results on two benchmark datasets, utilizing three different LLMs, show that HSF significantly enhances resilience against six cutting-edge jailbreak attacks. It significantly reduces the success rate of jailbreak attacks while minimally impacting responses to benign user queries, with negligible inference overhead, and outperforming defense baselines.Our code and data are available at https://anonymous.4open.science/r/Hidden-State-Filtering-8652/