TrajGuard: Streaming Hidden-state Trajectory Detection for Decoding-time Jailbreak Defense
This addresses a critical blind spot in jailbreak defense for LLM users by enabling real-time detection without model modification, though it is incremental as it builds on existing hidden-state analysis.
The paper tackled the problem of jailbreak attacks on large language models by proposing TrajGuard, a training-free defense framework that detects risk in real-time using hidden-state trajectories during decoding, achieving an average defense rate of 95% and reducing latency to 5.2 ms/token with a false positive rate below 1.5%.
Existing jailbreak defense paradigms primarily rely on static detection of prompts, outputs, or internal states, often neglecting the dynamic evolution of risk during decoding. This oversight leaves risk signals embedded in decoding trajectories underutilized, constituting a critical blind spot in current defense systems. In this work, we empirically demonstrate that hidden states in critical layers during the decoding phase carry stronger and more stable risk signals than input jailbreak prompts. Specifically, the hidden representations of tokens generated during jailbreak attempts progressively approach high-risk regions in the latent space. Based on this observation, we propose TrajGuard, a training-free, decoding-time defense framework. TrajGuard aggregates hidden-state trajectories via a sliding window to quantify risk in real time, triggering a lightweight semantic adjudication only when risk within a local window persistently exceeds a threshold. This mechanism enables the immediate interruption or constraint of subsequent decoding. Extensive experiments across 12 jailbreak attacks and various open-source LLMs show that TrajGuard achieves an average defense rate of 95%. Furthermore, it reduces detection latency to 5.2 ms/token while maintaining a false positive rate below 1.5%. These results confirm that hidden-state trajectories during decoding can effectively support real-time jailbreak detection, highlighting a promising direction for defenses without model modification.