A Federated Learning Approach for Intrusion Detection System using Deep Neural Network

Network technology advancements have significantly increased the risk of device and data compromise, particularly during data sharing and distribution. Machine learning (ML) and deep learning (DL) based intrusion detection systems (IDS) present effective solutions for mitigating these threats. However, in networks with numerous devices, sharing data with a central server for model training poses serious privacy and security risks. Federated Learning (FL) presents a viable solution by eliminating the need to share data with the server for training. Instead, it enables local model training, with only model parameters transmitted to the server for aggregation in a continuous learning process. This paper addresses privacy and security challenges in intrusion detection by leveraging federated learning with Convolutional Neural Networks(CNN). The proposed approach is tested using the UNSW-NB15 dataset under both independent and identically distributed (IID) and non-independent and non-identically distributed (Non-IID) data distributions, providing competitive results compared to centralized architectures, while ensuring data privacy.