XD-IDS: Explainable Deep Learning-Based Intrusion Detection System for IoT

The growing use of the Internet of Things (IoT) in our daily lives necessitates focusing on its security. Conventional security techniques are ineffective due to IoT devices’ limited memory, computing capacity, and battery consumption. An Intrusion Detection System (IDS) is a key security technology used in IoT security. This work proposes a new approach to enhance network-based intrusion detection in IoT environments. We use random undersampling and Synthetic Minority Oversampling Technique (SMOTE) techniques to handle the imbalanced IoTID20 dataset problem and recursive feature elimination (RFE) to identify the pertinent features. The selected features are then utilized to train the Deep Neural Network (DNN) model. Shapley Additive Explanations are used to interpret the model’s prediction. The proposed model achieved 99% accuracy.