Wi-Fi Network Intrusion Detection: Enhanced with Feature Extraction and Machine Learning Algorithms

Wireless Fidelity (Wi-Fi) is one of the most popular technologies for Internet connection. It helps users to connect to other devices by simply typing the correct password. It contributes to a great extent to building environments powered by IoT in schools, offices, homes, and cities. However, it has vulnerabilities, causing others to suffer low quality service or personal information theft when misused. Misusing is possible in several ways, and it is often caused by a network attack. To maintain the security of a network, network intrusion detection systems are important. For this reason, this study focused on network intrusion detection and used publicly available Aegean Wi-Fi Intrusion Dataset 3 (AWID3). We have used K-Nearest Neighbors, Support Vector Machine, and Gradient Boosting Trees algorithms for attack detection. In addition, we used time-series data and extracted features using two methods: Convolutional Neural Network (CNN) and statistical measures, to observe the effect on learning and facilitate the delivery of time-series data to machine learning algorithms. To the best of our knowledge, we are the first to conduct experiments on the AWID3 dataset using a combination of machine learning algorithms, feature extraction methods, and time-series data for network intrusion detection. We conducted six experiments in total, and the results of our best-performed experiments are as follows: accuracy: 0.978, precision: 0.978, recall: 0.978, F1 score: 0.978.