Hybrid Privacy Preserving Federated Learning Against Irregular Users in Next-Generation Internet of Things

While federated learning (FL) is a well-known privacy-preserving (PP) solution, recent studies demonstrate that it still has privacy problems and vulnerabilities, particularly in the context of the Next Generation Internet-of-Things (NG-IoT). Attackers on the server can potentially retrieve sensitive information such as data tabs and memberships. Additionally, current FL studies often overlook critical FL issues such as user dropout and low-quality data, which are prevalent in NG-IoT environments. In this article, we propose a privacy-preserving federated model in hybrid terms based on synchronous + asynchronous manner. This federated model removes the fundamental issues of FL-enabled NG-IoT environments while applying Two-Trapdoor Homomorphic Encryption (TTHE) to ensure the privacy and confidentiality of all components in the proposed model. Furthermore, our server protocol removes the effect of irregular users. In this regard, to control the impact of user dropout, we developed the hybrid LEGATO algorithm, which is asynchronous. To tackle users with low-quality data, we utilize the data-sharing method. Security analysis of the proposed model guarantees correctness, auditing, and PP in a federated setup. Also, in the performance evaluation, we achieved a high level of functionality and accuracy compared to peer works, while system overheads are shown to be lower.