all InfoSec news
Attention Hijacking in Trojan Transformers. (arXiv:2208.04946v1 [cs.LG])
Aug. 11, 2022, 1:20 a.m. | Weimin Lyu, Songzhu Zheng, Tengfei Ma, Haibin Ling, Chao Chen
cs.CR updates on arXiv.org arxiv.org
Trojan attacks pose a severe threat to AI systems. Recent works on
Transformer models received explosive popularity and the self-attentions are
now indisputable. This raises a central question: Can we reveal the Trojans
through attention mechanisms in BERTs and ViTs? In this paper, we investigate
the attention hijacking pattern in Trojan AIs, \ie, the trigger token
``kidnaps'' the attention weights when a specific trigger is present. We
observe the consistent attention hijacking pattern in Trojan Transformers from
both Natural Language …
More from arxiv.org / cs.CR updates on arXiv.org
Jobs in InfoSec / Cybersecurity
SOC 2 Manager, Audit and Certification
@ Deloitte | US and CA Multiple Locations
Security Engineer 2
@ Oracle | BENGALURU, KARNATAKA, India
Oracle EBS DevSecOps Developer
@ Accenture Federal Services | Arlington, VA
Information Security GRC Specialist - Risk Program Lead
@ Western Digital | Irvine, CA, United States
Senior Cyber Operations Planner (15.09)
@ OCT Consulting, LLC | Washington, District of Columbia, United States
AI Cybersecurity Architect
@ FactSet | India, Hyderabad, DVS, SEZ-1 – Orion B4; FL 7,8,9,11 (Hyderabad - Divyasree 3)