skip to content
Decrypt LOL

Get Cyber-Smart in Just 5 Minutes a Week

Decrypt delivers quick and insightful updates on cybersecurity. No spam, no data sharing—just the info you need to stay secure.

Read the latest edition

Vulnerabilities in Federated Learning Identified by New Research

/ 1 min read

🧠💔 New research reveals vulnerabilities in federated learning through malicious unlearning attacks. The study introduces a novel attack method called FedMUA, which exploits the federated unlearning process designed to comply with the “right to be forgotten.” By misleading the global model into unlearning more information than intended, FedMUA can adversely affect predictions for target samples from other clients. The researchers developed a two-step approach to identify influential samples and generate malicious unlearning requests, achieving an 80% success rate with minimal requests. Additionally, they propose a robust defense mechanism to counteract these attacks, highlighting significant security concerns in federated learning systems. The findings underscore the need for improved safeguards in data privacy practices within machine learning frameworks.

Source
{entry.data.source.title}
Original