A Wireless Intrusion Detection System using Feature Selection with Random Forest

Wireless networks have become integral to modern communication infrastructure, facilitating seamless connectivity in various domains. However, their presence everywhere also makes them vulnerable to various attacks. Those can cause failure in network services. Indeed, a robust Wireless Intrusion Detection System (WIDS) is essential for identifying and mitigating attacks on wireless networks. The key objective of WIDS is to detect wireless traffic and classify it as an attack or normal. In this situation, machine learning (ML) algorithms are used to detect attacks or intrusions in wireless systems. However, these algorithms need to accurately detect attacks early. Feature selection (FS) techniques are used to reduce this detection time and increase accuracy. These allow our system to identify attacks against local nodes. The proposed system is implemented, experimented and tested on AWID3 dataset. The performance of the system is determined by using a random forest (RF) algorithm and FS selection techniques. The experimentation conducted on AWID3 shows that the proposed methodology achieved an accuracy of 99.9990% for the binary model using 143 significant features and an accuracy of 99.9985% for multilabel classification using 144 significant features