Intrusion Detection in Wireless Sensor Networks using Machine Learning

Abstract

Attackers are continuously enhancing their techniques to attack confidential data. Thus, the classical security solutions based on cryptography and older detection systems must be revised, especially for a network dataset. Wireless sensor networks (WSN) consist of several smart devices distributed randomly in hostile areas with many restrictions, such as low operational time, short memory and minimal energy resources, and are susceptible to numerous attacks, such as Denial-of-Service (DoS). Recently, intrusion detection based on artificial intelligence models has provided an alternative and efficient solution with low computational resources. Therefore, the present study provides an innovative approach for detecting DoS attacks in the network using machine learning. Three machine learning methods, such as Support Vector Machine (SVM), K-Nearest Neighbor (KNN) and decision tree, were implemented with a well-known WSN-DS data specialized for DoS attacks. The performance of the models has been evaluated using standard metrics like accuracy, precision, F1-score and recall. The experimental results demonstrated the effectiveness of the KNN model after achieving the highest accuracy of 99.76% in DoS detection. Thus, the present study’s approach can be used in real-time intrusion detection.