FEX-IDS: A Federated Learning-Based Explainable Intrusion Detection System for IoT Security

With the increasing adoption of the Internet of Things (IoT), ensuring data privacy has become a critical challenge. Traditional Intrusion Detection Systems (IDS) often require centralized data collection, raising concerns about sensitive information exposure and potential security breaches. Additionally, the black-box nature of many machine learning-based IDS solutions makes it difficult to understand and trust their decisions. To enhance security in IoT environments, there is a pressing need for a privacy-preserving detection mechanism that not only protects data but also provides transparency into model predictions.This study proposes a Federated & Explainable Intrusion Detection System (FEX-IDS) using Federated Learning (FL) to collaboratively train a model across IoT devices without sharing the raw data which ensure privacy. It evaluates the system on the UNSW-NB15 dataset to over 96% accuracy in binary classification and 86% in multi-classification scenarios. SHAP and LIME are added to provide insights into the model predictions, hence making the model more transparent.