An Efficient Data Filtering, Feature Extraction Based Multi-NLK SVM Network Intrusion Detection on Iot Database

Abstract

As Internet connections and new technologies like IoT develop, the introduction of various devices continues to rise significantly, bringing new privacy and security threats. Numerous, highly advanced incursions push the IoT paradigm into computer networks. To detect these attacks better, businesses are investing more in research. Institutions choose smart testing and verification methods by comparing those with the highest accuracy rates. As the Internet of Things (IoT) devices proliferate in diverse networks, connected cybersecurity concerns increase. Researchers created large datasets like Bot-IoT to train machine learning algorithms to help mitigate these risks.Focusing on network-based intrusion detection for Internet of Things devices, Bot-IoT exhibits a high-class imbalance between individual attack categories, the normal category, and the aggregated attack categories when viewed from a binary classification perspective. The growing popularity of IoT and mobile devices has made them an attractive target for attackers, as they often have unpatched security vulnerabilities. In today's world, computer networks play a crucial role in the information and communication technology era, connecting heterogeneous devices for data communication and sharing. However, the large number of Internet-connected devices makes them vulnerable to massive security attacks. Most widely-used IoT devices lack security design, making them vulnerable to recent attacks that exploit these weaknesses and recruit the devices to cause severe harm. In this work, a fitler based classification approach was proposed for detecting IoT bot cyber attacks. The proposed method achieved good results in terms of accuracy, precision, recall and F1-score, compared to traditional methods. The results demonstrate that the cluster based classification approach is a promising solution for detecting IoT bot cyber attacks in real-time.