A Review and Analysis of Privacy-Preserving Federated Learning-Based Applications and Services in IoT Networks

With the advent of 5G and 6G endpoints, the deployment of billions of IoT devices is anticipated to increase significantly. On the other hand, this expansion will result in the generation of enormous volumes of data that contain the personal information of users. This in turn will present difficulties for typical centralized over-the-cloud learning and processing ecosystems due to the high costs of communication and storage. Within the realm of potential solutions to this issue, Privacy-Preserving Federated Learning (PPFL) has emerged as the most promising alternative. In this article, we present a comprehensive study based on our proposed taxonomy on PPFL for IoT services and applications. Furthermore, we have summarized an analysis of existing works based on our proposed taxonomy and highlighted several research challenges. Based on our study, it has been observed that there is a need to develop more PPFL-based frameworks for IoT applications rather than IoT services.