all InfoSec news
Shortcuts Arising from Contrast: Effective and Covert Clean-Label Attacks in Prompt-Based Learning
April 2, 2024, 7:11 p.m. | Xiaopeng Xie, Ming Yan, Xiwen Zhou, Chenlong Zhao, Suli Wang, Yong Zhang, Joey Tianyi Zhou
cs.CR updates on arXiv.org arxiv.org
Abstract: Prompt-based learning paradigm has demonstrated remarkable efficacy in enhancing the adaptability of pretrained language models (PLMs), particularly in few-shot scenarios. However, this learning paradigm has been shown to be vulnerable to backdoor attacks. The current clean-label attack, employing a specific prompt as a trigger, can achieve success without the need for external triggers and ensure correct labeling of poisoned samples, which is more stealthy compared to the poisoned-label attack, but on the other hand, it …
adaptability arxiv attack attacks backdoor backdoor attacks covert cs.ai cs.cl cs.cr cs.lg current language language models paradigm prompt shortcuts vulnerable
More from arxiv.org / cs.CR updates on arXiv.org
Jobs in InfoSec / Cybersecurity
Technical Senior Manager, SecOps | Remote US
@ Coalfire | United States
Global Cybersecurity Governance Analyst
@ UL Solutions | United States
Security Engineer II, AWS Offensive Security
@ Amazon.com | US, WA, Virtual Location - Washington
Senior Cyber Threat Intelligence Analyst
@ Sainsbury's | Coventry, West Midlands, United Kingdom
Embedded Global Intelligence and Threat Monitoring Analyst
@ Sibylline Ltd | Austin, Texas, United States
Senior Security Engineer
@ Curai Health | Remote