Ensemble Knowledge Distillation-based Federated Learning for Effective Intrusion Detection in Heterogeneous Networks

The rapid development of low-cost consumer electronics and cloud computing has led to the widespread adoption of Internet-of-Things (IoT) devices in various applications such as smart cities and industrial control systems. However, these devices are vulnerable to cyber attacks due to their open deployment environment and limited computing capabilities. Intrusion Detection Systems (IDS) have emerged as an effective way to secure IoT networks by monitoring and detecting abnormal activities. However, existing IDS approaches rely on centralized servers, which can cause high response time and operational costs. Additionally, sharing behaviour data in an open and distributed IoT network may violate privacy requirements. To address these challenges, the paper introduces Federated Learning (FL) as a collaborative training approach for developing a decentralized shared model of IDS. FL allows the training of a global IDS model without exposing training data to others. Local model parameters are transmitted to a central server, which aggregates them into a global model. The updated parameters are then distributed to all local clients. This approach reduces response time and communication overhead while preserving privacy.