Task-Agnostic Detector for Insertion-Based Backdoor Attacks

arXiv:2403.17155v1 Announce Type: new
Abstract: Textual backdoor attacks pose significant security threats. Current detection approaches, typically relying on intermediate feature representation or reconstructing potential triggers, are task-specific and less effective beyond sentence classification, struggling with tasks like question answering and named entity recognition. We introduce TABDet (Task-Agnostic Backdoor Detector), a pioneering task-agnostic method for backdoor detection. TABDet leverages final layer logits combined with an efficient pooling technique, enabling unified logit representation across three prominent NLP tasks. TABDet can jointly learn …

abstract arxiv attacks backdoor beyond classification cs.cl cs.cr current detection feature intermediate question question answering recognition representation security tasks textual threats type