Effective Feature Selection For Hybrid Wireless Iot Network Intrusion Detection Systems Using Machine Learning Techniques

Through the ascent of online business and the Internet of Things (IoT), safety of such frameworks over wireless systems are getting even more an alarm. Hence, the job of Intrusion Detection System (IDS) is critical to direct and help the movement of actions over the wireless systems. One of the essential difficulties to IDS is the issue of confusion, misidentification and absence of continuous reaction to the assault. Thus, a hybrid IDS structure that relies upon machine learning classification and clustering procedures is anticipated to decrease the false positive rate, false negative rate, to increase the discovery rate and identify zero-day assaults. Primarily, the features are chosen utilizing feature selection techniques like Information Gain (IG), Chi-Squared statistics (CH), and Correlation-based Feature Selection (CFS), with the goal that the count of features taking part in the discovery of assaults must be given more importance. Therefore, in the assault discovery stage, first anomaly or profile-based identification model is fabricated utilizing K-means clustering algorithm and based upon the output of profile-based identification, misuse or signature-based identification model is implemented utilizing Random Forest (RF) classification algorithm. Aimed at training and testing of proposed hybrid IDS model Aegean Wi-Fi Dataset (AWID) with three classes of wireless assaults, for example, Impersonation, Injection and Flooding is utilized. Subsequent to training and testing, the outcomes have indicated that the proposed method (K-mean + RF) has accomplished high true positive rate, low false positive rate and high accuracy.