Towards Building Robust Intrusion Detection System in Wireless Sensor Networks using Machine Learning and Feature Selection

Abstract

Wireless sensor Networks (WSN) have numerous applications in today’s world. The security threat, on the other hand, is a major concern in WSN. Incorporating an intrusion detection system (IDS) in a WSN can detect and categorize attacks, allowing security threats to be combated. As a result, the proposed work aims to build an effective IDS to classify attacks in a WSN using machine learning. The work uses a publicly available WSN-DS dataset to evaluate the system performance. A baseline system is developed using features extracted from WSN DS dataset using a decision tree classifier. To reduce the time for attack detection, a feature selection using Correlation Score, Fisher Score, Statistical Analysis via Kruskal-Wallis (KW) test, Minimum redundancy maximum relevance (MRMR) algorithm and Relief algorithm are experimented. The proposed feature selection techniques are evaluated using relevant performance metrics. Using MRMR feature selection, the metric scores are 98.58%, 92.81%, 98.46% 93.86% and 15.12 secs for accuracy, sensitivity, specificity and precision and training time, respectively. Henceforth, this study could be used to build robust IDS system in a WSN.