Encoded Flow Features for Network Intrusion Detection in Internet of Things

In the recent years, the Internet of Things has been becoming a vulnerable target of intrusion attacks. As the academia and industry move towards bringing the Internet of Things (IoT) to every sector of our lives, much attention needs to be given to develop advanced Intrusion Detection Systems (IDS) to detect such attacks. In this work, we propose Codebook-based Encoded Flow Features (CEFFs) as an innovative method to transform the raw flow-based statistical features into more discriminative representations taking into account the different flow features and patterns of various devices. Based on the proposed CEFFs, we build and leverage Support Vector Machine (SVM)-based classifiers to discriminate the malicious flows from the benign ones. The effectiveness of the proposed CEFFs for intrusion detection is evaluated on two state-of-the-art realistic datasets, achieving high accuracies and low false positive rates across a variety of intrusion attacks.