Noisy data-based attack: A new type of untargeted attack in Federated Learning and its countermeasures

Manh Cuong Dao, Phi Le Nguyen, Hieu H. Pham, Thanh Hung Nguyen, Peng Chen, Mohamed Wahib, Truong Thao Nguyen

Abstract: Highlights•We introduce a novel untargeted noisy data-based attack.•We propose a novel defense that identifies malicious clients using confidence scores.•Our defense achieves 97% detection accuracy and improves model accuracy by 2%–3%.