A Federated Learning Approach to Intrusion Detection in Software-Defined Networks
Santhosh Voruganti, Sairam Utukuru, Sravyasri Shambuni, Yeshwanth Valishetti, Hemalata Mote · 2025
Network Intrusion Detection Systems are essential in the fields of Software and Hardware. In Networking, the use of Software-Defined Networks is rapidly increasing within Information Technology to manage traffic in data centers while enhancing security. However, due to their centralized management, there is a significant risk of data breaches. Securing SDNs is crucial for both data centers and media. Numerous existing models implement IDS using machine learning algorithms with various datasets, but a major drawback is that the training data provided by users is often insecure, as it can be accessed by others. This is where traditional algorithms fall short. Federated Learning offers an improved approach to implementing these detection models by enhancing the security of user-provided data. It employs collaborative learning, where individual models from clients are aggregated, and the average is taken. This concept of integrating Federated Learning with the base algorithm safeguards user privacy. The proposed model stands out in terms of the number of attacks detected and in alerting users when an intrusion is identified.