Security and Privacy in Federated Learning
Abstract
Federated learning enables collaborative training while keeping data local, but exchanging gradients and model updates introduces security and privacy risks. Poisoning, backdoor, adversarial, Byzantine, and Sybil attacks threaten model integrity, while membership inference, attribute inference, and gradient leakage may expose sensitive training data. Existing defenses include robust aggregation, anomaly detection, data sanitization, adversarial training, differential privacy, homomorphic encryption, secure multi-party computation, and trusted execution environments. Blockchain can further improve auditability and accountability. Overall, effective federated learning protection requires system-level defense designs that balance robustness, privacy, and efficiency.
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.