Selecting Gaussian Process Regression Kernels for IoT Intrusion Detection and Classification
Cosmas Ifeanyi Nwakanma, Love Allen Chijioke Ahakonye, Jae‐Min Lee, Dong‐Seong Kim · 2021 International Conference on Information and Communication Technology Convergence (ICTC) · 2021
Intrusion detection is a well-documented research area recently. This is due to the growing cases of attacks and vulnerabilities of the internet of things (IoT). To mitigate or reduce attacks, several classification and detection techniques have been introduced. One of the schemes with appreciable accuracy is the gaussian process classification. However, not much attention has been devoted to the impact and selection of kernels on its performance. This study investigated various Gaussian process kernels in terms of accuracy, and reduction in false alarm rate. Leveraging the CICDDoS2019 datasets, data training and prediction were conducted using MATLAB machine learning toolbox. The exponential GPR kernel outperformed other state-of-the-art kernels such as rational quadratic, squared exponential, and matern 5/2.