all InfoSec news
Privacy Backdoors: Enhancing Membership Inference through Poisoning Pre-trained Models
April 2, 2024, 7:11 p.m. | Yuxin Wen, Leo Marchyok, Sanghyun Hong, Jonas Geiping, Tom Goldstein, Nicholas Carlini
cs.CR updates on arXiv.org arxiv.org
Abstract: It is commonplace to produce application-specific models by fine-tuning large pre-trained models using a small bespoke dataset. The widespread availability of foundation model checkpoints on the web poses considerable risks, including the vulnerability to backdoor attacks. In this paper, we unveil a new vulnerability: the privacy backdoor attack. This black-box privacy attack aims to amplify the privacy leakage that arises when fine-tuning a model: when a victim fine-tunes a backdoored model, their training data will …
application arxiv attacks availability backdoor backdoor attacks backdoors cs.cr cs.lg dataset fine-tuning foundation large new vulnerability poisoning privacy risks the web vulnerability web
More from arxiv.org / cs.CR updates on arXiv.org
Jobs in InfoSec / Cybersecurity
Junior Cybersecurity Analyst - 3346195
@ TCG | 725 17th St NW, Washington, DC, USA
Cyber Intelligence, Senior Advisor
@ Peraton | Chantilly, VA, United States
Consultant Cybersécurité H/F - Innovative Tech
@ Devoteam | Marseille, France
Manager, Internal Audit (GIA Cyber)
@ Standard Bank Group | Johannesburg, South Africa
Staff DevSecOps Engineer
@ Raft | San Antonio, TX (Local Remote)
Domain Leader Cybersecurity
@ Alstom | Bengaluru, KA, IN