Performance Evaluation In Internet Of Things Dataset Using Machine

Published: 4 October 2021| Version 1 | DOI: 10.17632/kzyfgkz8xw.1


Many Internet of Things (IoT) applications face significant hurdles in terms of data security. Machine Learning (ML)-based intrusion detection systems (IDS) claim to be effective and accurate at analysing network data and detecting threats. Our suggested technique, nweighted-univariate feature selection, creates a threshold value that serves as a weight, from which critical features are extracted and then used to machine learning algorithms like support vector machine (SVM) and decision tree (DT). These models were trained using the UNSWNB- 15 dataset, which was developed in the Australian Center for Cyber Security's Cyber Range Lab using an IXIA PerfectStrom tool (ACCS). It has a mix of realistic modern normal and contemporary network traffic assault characteristics. Accuracy, precision, and recall were used to evaluate the performance of our suggested model. In DT, the proposed model has a greater accuracy of 96.4 than SVM, which has an accuracy of 89.1.


Steps to reproduce

The researchers of this domain has major challenge was to created a dataset which is complicated modern real scenario-based orientation normal and attacks in network traffics. This kind of challenged dataset was created by Australian Centre for Cyber Security (ACCS) named as UNSW-NB15 dataset, it is created by using IXIA PerfectStorm device (used to test the security of devices) in Cyber Range Lab. It contains nine types of attacks scenarios, approximately 2.5 million packets are captured to form a comprehensive hybrid network contemporary attack. The IXIA traffic generator will generates all types of packets with normal spread of network traffic, it has three servers, server3 generated attacks, server1 and 3 generated normal packets. All network packets were configured to pass through firewall, tcmpdump tool which was installed in router1 to capture Pcap files and it stores in CVE site continuously.


APJ Abdul Kalam Technological University


Machine Learning, Feature Selection, Intrusion Detection, Internet of Things