Federated Learning for Privacy-Preserving Intrusion Detection in IoT Networks
Sarmad Dheyaa Azeez, Muhammad Ilyas, Imad Matti Bako
Abstract
The expansion of IoT networks, ensuring robust intrusion detection while safeguarding user privacy is a significant challenge. Previous techniques have struggled to reconcile accuracy and privacy. In this research, we offer a unique approach employing federated learning for intrusion detection in IoT contexts. Our technique leverages the Federated Averaging (FedAvg) algorithm. by allowing IoT devices to cooperatively train a global model. To evaluate our technique, we employ the CICIDS 2017 dataset. It simulates realistic network traffic scenarios. Our data demonstrate a remarkable accuracy of 95.2%. It demonstrates the efficiency of federated learning in increasing intrusion detection capabilities. Our investigation reveals encouraging results. Future work intends to solve restrictions, such as broadening the scope to accommodate varied assault situations. Our research illustrates the transformational potential of federated learning in increasing cybersecurity in IoT networks while maintaining user privacy.