Web: http://arxiv.org/abs/2301.09732

Jan. 25, 2023, 2:10 a.m. | Gokberk Yar, Cristina Nita-Rotaru, Alina Oprea

cs.CR updates on arXiv.org arxiv.org

We study backdoor attacks in peer-to-peer federated learning systems on
different graph topologies and datasets. We show that only 5% attacker nodes
are sufficient to perform a backdoor attack with 42% attack success without
decreasing the accuracy on clean data by more than 2%. We also demonstrate that
the attack can be amplified by the attacker crashing a small number of nodes.
We evaluate defenses proposed in the context of centralized federated learning
and show they are ineffective in peer-to-peer …

attacks backdoor federated learning peer-to-peer

Penetration Testers - Red Team ICS/OT and Network Experience

@ Waterleaf International LLC | Florida

Chief Information Security Officer

@ Los Angeles Unified School District | Los Angeles

Cybersecurity Engineer

@ Apercen Partners LLC | Folsom, CA

IDM Sr. Security Developer

@ The Ohio State University | Columbus, OH, United States

IT Security Engineer

@ Stylitics | New York City

Information Security Engineer

@ VDA Labs | Remote

Cyber Security Consultant

@ PubMatic | Redwood City, CA, United States

Site Reliability Engineer and Security Lead (Belgrade - hybrid remote)

@ SMG Swiss Marketplace Group | Beograd, Serbia

Associate Threat Intelligence Response Analyst

@ Recorded Future, Inc. | Australia

Staff Information Security Engineer

@ ServiceNow | Dublin, Ireland

Security Engineer- bilingual Japanese/English

@ Cookpad Ltd | Bristol, England, United Kingdom - Remote

Security Consultant Team Lead

@ ProArch | Rochester, New York, United States - Remote