Federated Learning for Detecting Anomaly in IoT Networks

• Published in: International Conference on Signal Processing and Communication (Online) pp. 409 - 414
• Main Authors: Tomar, Vikas, Jain, Vikas Kumar, Yadav, Anil Kumar, Shukla, Shiv Shankar Prasad
• Format: Conference Proceeding
• Language: English
• Published: IEEE 20.02.2025
• Subjects: Cybersecurity; Data models; Data preprocessing; Data privacy; Feature extraction; Feature Selection; Federated learning; Internet of Things; Random Forest; Random forests; Redundancy; Signal processing; Training
• ISSN: 2643-444X
• DOI: 10.1109/ICSC64553.2025.10967664

Federated Learning (FL) presents a novel method for collaborative model training while safeguarding data privacy, which is a primary concern in the Internet of Things (IoT) world. This research uses FL to enhance security within IoT networks, employing the CICIoT2023 dataset. The dataset was subjected to important preprocessing, including the management of missing data and the removal of redundancies, with feature selection conducted using a Random Forest classifier to determine essential attributes. These features were incorporated into a simulated FL framework, where locally trained models from decentralized datasets contributed to a collectively aggregated model. Our findings reveal the capacity of FL to improve model accuracy without sacrificing data privacy, offering scalable and practical security solutions for IoT systems.