New FUSE method enhances fairness and security in federated learning

Researchers have developed a new method called FUSE to enhance the fairness and security of federated learning systems. This approach specifically addresses Byzantine attacks in $q$-fair federated learning, where client updates are weighted to equalize utility across users. FUSE introduces novel screening tests that analyze the geometry of client updates, distinguishing legitimate fairness-induced scaling from malicious manipulation. The system demonstrated significant improvements in accuracy against various attacks across multiple benchmarks, while still protecting the influence of high-loss clients. AI

IMPACT Enhances the security and fairness of distributed machine learning models, potentially enabling more robust AI systems in sensitive applications.

RANK_REASON The cluster contains a research paper detailing a new method for federated learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Agnostic Federated Learning
• arXiv
• Byzantine attacks
• federated learning
• FUSE
• $q$-Fair Federated Learning
• tilted empirical risk minimization

• paper
• safety
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →