Enhanced Feature Optimization for Multiclasss Intrusion Detection in IOT Fog Computing Environments

Abstract

To overcome the shortcomings of traditional security measures in IoT fog computing, The Multiclass Intrusion Detection (MCID) model is put forward. The model's goal is to improve intrusion detection by identifying and classifying different attack types. The behavioral, temporal and anomaly features are fused through SVM-BFE for obtaining the best possible selection of high worth features. Finally we use a Random Forest algorithm to robustly classify them. There is also its adaptability to the ever-changing security demands of fog computing. MCID's ability to improve the cloud security of fog computing is shown by a 4-fold cross validation, which returns performances including precision rates up to 99.43%, recall about 95% and F-measures at as much as 97.17%. Moreover there are specificity rate totals coming in over this whole range that hit close or right.