SIDS: A federated learning approach for intrusion detection in IoT using Social Internet of Things

Abstract

The Internet of Things (IoT) ecosystem needs Intrusion Detection Systems (IDS) to mitigate cyberattacks and exploit security vulnerabilities. Over the past years, utilizing machine learning in IDSs has gained a lot of attention. However, in many current works, the training data from different locations should be collected in a central server to be used in the learning process. This data-sharing procedure increases concerns regarding data privacy and decreases the data holders’ motivation to participate in the learning process. The use of distributed learning models has been considered a solution to overcome concerns related to privacy. However, these distributed learning models are vulnerable in the presence of untrusted nodes that can participate in the learning process and deteriorate performance. In this paper, we propose SIDS (Social Intrusion Detection System), a trust-oriented federated learning approach for intrusion detection in IoT that utilizes the Social Internet of Things (SIoT). The proposed approach leverages the social relationships among the objects in a system to provide a privacy-preserving collaborative mechanism for detecting intrusions in IoT environments. The experimental results show the proposed solution outperforms the learning models on individual servers while providing a privacy-preserving and trustable environment for collaboration.