Achieving Security and Privacy in Federated Learning Systems: Survey, Research Challenges and Future Directions

Federated learning (FL) allows a server to learn a machine learning (ML)\nmodel across multiple decentralized clients that privately store their own\ntraining data. In contrast with centralized ML approaches, FL saves computation\nto the server and does not require the clients to outsource their private data\nto the server. However, FL is not free of issues. On the one hand, the model\nupdates sent by the clients at each training epoch might leak information on\nthe clients' private data. On the other hand, the model learnt by the server\nmay be subjected to attacks by malicious clients; these security attacks might\npoison the model or prevent it from converging. In this paper, we first examine\nsecurity and privacy attacks to FL and critically survey solutions proposed in\nthe literature to mitigate each attack. Afterwards, we discuss the difficulty\nof simultaneously achieving security and privacy protection. Finally, we sketch\nways to tackle this open problem and attain both security and privacy.\n