Network Attack Classification in IoT Using Support Vector Machines

Machine learning (ML) techniques learn a system by observing it. Events and occurrences in the network define what is expected of network’s operation. It for this reason that ML are used computer security field to detect unauthorized intervention. In event suspicious activity, result analysis deviates from definition normal activity becomes apparent. Support vector machines (SVM) have been profile classify it as or abnormal. They trained configure an optimal hyperplane classifies unknown input vectors’ values based on their positioning plane. We propose use SVM models malicious behavior within low-power, low-rate short range networks, such those Internet Things (IoT). evaluated two approaches, C-SVM OC-SVM, where former requires classes (one one abnormal activity) latter observes only activity. Both approaches were part intrusion detection (IDS) monitors detects smart node device. Actual traffic with specific network-layer attacks implemented us was create evaluate models. shown achieves up 100% classification accuracy when data taken same topology 81% operating topology. The OC-SVM created using benign at most 58% accuracy.