all InfoSec news
Task-Agnostic Detector for Insertion-Based Backdoor Attacks
March 27, 2024, 4:11 a.m. | Weimin Lyu, Xiao Lin, Songzhu Zheng, Lu Pang, Haibin Ling, Susmit Jha, Chao Chen
cs.CR updates on arXiv.org arxiv.org
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 …
arxiv attacks backdoor backdoor attacks beyond classification cs.cl cs.cr current detection detector feature intermediate question recognition representation security security threats task threats
More from arxiv.org / cs.CR updates on arXiv.org
Jobs in InfoSec / Cybersecurity
Social Engineer For Reverse Engineering Exploit Study
@ Independent study | Remote
Security Engineer II- Full stack Java with React
@ JPMorgan Chase & Co. | Hyderabad, Telangana, India
Cybersecurity SecOps
@ GFT Technologies | Mexico City, MX, 11850
Senior Information Security Advisor
@ Sun Life | Sun Life Toronto One York
Contract Special Security Officer (CSSO) - Top Secret Clearance
@ SpaceX | Hawthorne, CA
Early Career Cyber Security Operations Center (SOC) Analyst
@ State Street | Quincy, Massachusetts