WaSA: Wavelet Scattering Autoencoders for Unsupervised Offline Detection of Malicious Clients in Federated Learning

Key points

• Federated Learning (FL) enables collaborative model training without exposing raw data, yet remains vulnerable to malicious clients injecting poisoned datasets.1
• A denoising autoencoder, trained solely on clean auxiliary data, scores clients based on their deviation from the learned benign manifold, identifying corruptions regardless of the attack type.1
• WaSA requires no attack labels, generalizes to unseen anomalies, and introduces a Boltzmann sampling strategy for client selection that smoothly interpolates between hard exclusion and standard aggregation.1
• Experiments on MNIST, CIFAR-10, and CIFAR-100 under diverse attack scenarios (up to 90 % malicious clients) demonstrate consistent improvements over supervised baselines and robust aggregation methods, with negligible communication overhead.1

Sources (1)

• [1]WaSA: Wavelet Scattering Autoencoders for Unsupervised Offline Detection of Malicious Clients in Federated Learning
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 08:10 AM

  We propose WaSA, a fully unsupervised, offline framework for malicious client detection.

  Federated Learning (FL) enables collaborative model training without exposing raw data, yet remains vulnerable to malicious clients injecting poisoned datasets.

Extractive summary: sentences quoted from the sources.