A Review on Federated Learning with a Focus on Security and Privacy
Amala Mathew, V. Panchami · 2024
Federated Learning (FL) enables the collaborative training of a global model without centralizing data, effectively addressing privacy concerns in the field of machine learning. However, FL remains vulnerable to various security and privacy threats, such as adversarial attacks, data poisoning, and privacy inference. This paper provides a comprehensive analysis of these threats, categorizing them based on the attacking party (aggregator or participant). We review state-of-the-art mitigation techniques, highlighting blockchain and differential privacy. Finally, we identify promising research directions for securing and enhancing FL privacy. This paper serves as a reference for developing secure and privacy-preserving FL solutions.