Two Heads are Better than One: Nested PoE for Robust Defense Against Multi-Backdoors

arXiv:2404.02356v1 Announce Type: new
Abstract: Data poisoning backdoor attacks can cause undesirable behaviors in large language models (LLMs), and defending against them is of increasing importance. Existing defense mechanisms often assume that only one type of trigger is adopted by the attacker, while defending against multiple simultaneous and independent trigger types necessitates general defense frameworks and is relatively unexplored. In this paper, we propose Nested Product of Experts(NPoE) defense framework, which involves a mixture of experts (MoE) as a trigger-only …

abstract arxiv attacks backdoor cs.cl data data poisoning defense importance language language models large language large language models llms multiple poe robust them type