A Secure and Privacy Preserving Federated Learning Approach for IoT Intrusion Detection System

Recently, machine learning (ML) has been shown as a powerful method for outstanding capability of resolving intelligent tasks across many fields. Nevertheless, such ML-based systems require to centralize a large amount of data in the training phase that causes privacy leaks from user data. This is also true with the ML-based intrusion detection system (IDS) due to containing sensitive user and network data, especially in the context of Internet of Things (IoT) intrusion detection. To promote the collaboration between multiple parties in building an efficient IDS model to detect more attack types and cope with the privacy preservation issues, federated learning (FL) is considered as a potential approach for localized training scheme without sharing any data collection between organizations or data silos. In this paper, we investigate the feasibility of adopting FL for anomaly behavior detection in the context of large-scale IoT networks while facilitating the secure and privacy preserving aggregation using homomorphic encryption and differential privacy.