An Integrated Approach to Detect Attacks in Cloud Environment using Wavelet Transforms and Cubic Kernel Classifiers

Le résumé fourni par la source

Cyber-attacks on the Internet of Things are increasingly common due to low-power devices. Effective detection algorithms are crucial for distinguishing between benign and malicious behaviour in IoT networks and video surveillance systems. These algorithms analyse network traffic patterns and detect cyber threats. However, using feature extraction-based algorithms to predict different characteristics of network traffic poses challenges in improving classifier accuracy. This paper presents a feature extraction-based cubic kernel classifier designed to detect three major attacks: Botnet, Denial of Service (DoS), and Man-in-the-Middle (MiM) that disrupt normal operations in IoT and communication networks. In our proposed method, (i) Features are extracted using three wavelet transforms: Discrete Wavelet Transform (DWT), Stationary Wavelet Transform (SWT), Transverse Quadratic Wavelet Transform (TQWT). These wavelet transforms outperform traditional methods in detecting anomalies and patterns due to their ability to analyse both time and frequency localization (ii) Highly significant features are selected using Pearson’s correlation coefficient, and (iii) The selected features are classified as benign or malicious using Cubic Support Vector Machine (SVM), Cubic K-Nearest Neighbors (KNN). The use of a cubic kernel function effectively models complex patterns and non-linear relationships within the network and improves accuracy compared to traditional classifiers. Experimental results from Kitsune attack datasets show that the proposed TQWT-based cubic SVM (TCSVM) outperforms other methods, achieving 99.5% accuracy for both Botnet and DoS attacks, and 99% for the MiM attack.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.