AI Insight
Researchers have developed FairMean, a new AI system designed to address the conflict between fairness and robustness in distributed machine learning when training data is deliberately sabotaged through label poisoning. The system uses a bounded weighting function that prioritizes clients with high losses to maintain fairness while preventing malicious clients from having excessive influence on the overall model. Theoretical analysis and experiments demonstrate that FairMean reduces performance disparities between clients and improves worst-case accuracy even under attack conditions.
Why it matters
This work addresses a critical vulnerability in distributed learning systems where adversaries could exploit fairness mechanisms to poison AI models. The approach has practical applications for federated learning scenarios in healthcare, finance, and other domains where both fairness across diverse populations and resistance to data manipulation are essential requirements.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Abstract: Fairness-aware distributed learning prioritizes clients with large losses to reduce performance disparities, but label poisoning can create large losses, thereby inducing a fairness–robustness conflict. We propose FairMean to manage this conflict. FairMean weights client gradients using a bounded, nondecreasing function of local loss. The increasing weights prioritize high-loss clients to promote fairness, while the upper bound prevents excessive loss-induced amplification of poisoned-client gradients. In the absence of label poisoning, we show that minimizing the FairMean objective is more conducive to solution fairness than minimizing the standard average-loss objective. Under label poisoning, we establish an average-stationarity bound whose attack-dependent term is proportional to the square of the poisoned-client fraction. Experiments show that FairMean promotes fairness by reducing accuracy variance while improving worst-client accuracy.
Source: FairMean: Promoting Fairness in Distributed Learning under Label Poisoning Attacks