SpecGuard: Inference-Time Backdoor Detection For Free
[Kryptografia i bezpieczeństwo] Autorzy: Rui Wen, Ahmed Salem, Andrew Paverd i 2 więcej
Large language models are often fine-tuned, shared, or downloaded from third parties, so a deployed model may carry a hidden backdoor that behaves normally on benign inputs but switches to attacker-controlled behavior when a secret trigger appears. While backdoors can be audited before deployment, runtime monitoring remains important for models that are frequently updated. The challenge is that LLM serving is latency-sensitive: existing inference-time detectors either rely on assumptions about the trigger form, which can fail on stealthy attacks, or require extra model computation, such as input perturbations or an additional generation pass. We introduce SpecGuard, an inference-time backdoor detector that repurposes speculative decoding at zero added model-computation cost. Speculative decoding speeds up inference by using a small draft model to propose tokens and a target model to ver…
- Kategoria arxiv
- cs.CR
- Kategoria
- Kryptografia i bezpieczeństwo