Statistical feature selection based intrusion detection system for internet of things environment Available to Purchase

IoT networks have drastically increased in the past few years because of the rapid growth in networking technology and IoT devices. However, these IoT devices suffer from various attacks because of the vulnerability of the IoT network. Because of the numerous elements in network data, the machine learning model will take a long to detect an assault. The proposed model has provided a solution for reducing this time by reducing the feature count. A statistical feature selection-based IDS model is presented for securing an IoT environment. Information gain, gain ratio, and Pearson coefficient values are calculated for all the dataset features, and then the addition is performed. After this, the addition is ranked in descending order, and top-5 and top-8 features are selected as best features. These reduced features are fed to random forests, and results are calculated. For the model's evaluation, the IoT-BoT dataset is used. The accuracy obtained with top-5 features with the random forest is 99.9995% which is better than other state-of-the-art research work. The proposed model has shown a reduction in computation time with similar accuracy compared to all the features. The proposed model is also tested against other state-of-the-art methodologies and is more accurate.