In today-s fast-paced world, the predominant internet usage in many distributed network-based applications such as mobile phones, wearable devices, and autonomous vehicles accelerates the generation of a huge amount of data. The computational power, data transmission, and network computations are progressively increased and expedited to unauthorized data exploitation due to unknown attacks. Intrusion detection is a significant network security tool to prevent unauthorized access and report attacks in distributed network applications.
Machine learning and deep learning are greatly utilized for intrusion detection systems. Nonetheless, these learning models are ineffective in privacy and security owing to the requirement of storage and communication of data in a centralized server. Conversely, federated learning is highly suitable for on-device learning and privacy conservancy. Federated learning is a decentralized learning paradigm that trains the models locally and transmits only the parameter of the models to the centralized server while persevering user data locally.
Federated learning-based intrusion detection is applied for heterogeneous anomaly detection, varied attacks detection, data privacy preservation, low-power IoT devices, and computer networks. Challenges and future scopes of federated learning for intrusion detection are federated learning on constrained IoT devices, imbalanced IoT datasets, IoT with dynamic behavior, coordinator entity constrictions, bandwidth requirements, client selection, security for data poisoning attacks, concernment of privacy for data leakage, and non-IID data handling.