Privacy‐Preserved Federated Learning
Mohamed Abdel‐Basset, Nour Moustafa, Hossam Hawash · 2022
The success of federated learning (FL) is attributed to the fact that it can train deep learning models while keeping the client's data on their machines. Though FL seems secure and privacy–privacy-preserving from the first look, it still suffers from many security vulnerabilities and privacy concerns. This chapter dives into the main challenges that might be encountered when it comes to applying FL to the Internet of Things (IoT) infrastructures. Understanding these challenges is essential for specifying and improving the design considerations to be considered to develop any federated solution. The existing challenges are broadly grouped under three main categories including statistical challenges, security challenges, and privacy challenges. The chapter describes the details of each of these challenges and the corresponding sub-challenges, and the possible solutions and design considerations that can be used to address them in the IoT infrastructure.