Privacy-Preserving Detection of DDoS Attacks in IoT Using Federated Learning Techniques

Internet of Things devices are widely employed in various industries, cities, and institutions as with low-cost investment various benefits can be obtained. The computing power, bandwidth, memory storage and battery are low in these devices and lightweight routing protocols are used for communication. Distributed Denial Service of Attack takes benefit of Internet of Things architecture and targets vulnerable devices to take control of the device. These hacked devices can be further used by attackers to send massive data packets to the victim's device. Generally, data servers and high computing devices are targeted by attackers through Internet of Things devices. Artificial intelligence-based cybersecurity is extensively employed in various scenarios but as there is a constraint of resources in the Internet of Things, the same approach is not feasible. Further, artificial intelligence-based solutions are required to deal with network traffic-related data, and organisations are reluctant to share the data. This research work targets to provide a solution for the early detection of distributed denial of service attacks using a federated learning model by preserving the information of the users. The client and server in the proposed approach are trained using Identically and Non-identically distributed approaches. The performance evaluation of both approaches is performed using metrics like accuracy, loss, precision and recall for the detection of cyberattack by model aggregation.