Explainable Deep Neural Network for a Reliable Intrusion Detection System with Shapley Additive Explanation Method

Abstract

Data security is a critical concern for many parties, as cyber threats and illegal activities, such as intrusion, can compromise both data confidentiality and integrity. Awareness of this issue has inspired many experts to develop various methods for intrusion detection. Following the discussion, previous studies showed that a neural network-based model provided superior performance when compared to others in identifying intrusion. On the other hand, a neural network-based model is difficult to interpret due to its complexity, which is a major obstacle for security analysis. Therefore, this study aimed to implement a Deep Neural Network (DNN) model incorporated with Explainable AI (XAI) methods to detect network intrusion. The analysis applied the CIC-IDS2017 dataset on an explainable DNN to develop a reliable intrusion detection system using the Shapley Additive Explanation (SHAP) method. During the process, the SHAP method was used to perform feature importance analysis, identifying features that contribute to the model prediction. The results showed that incorporating XAI methods with DNN enabled the model to achieve superior performance and also to provide a useful understanding of the features that contribute to its predictions, increasing confidence in the outcomes.