A Novel Intrusion Detection Method Based on WOA Optimized Hybrid Kernel RVM

Abstract

In recent years, various machine learning algorithms and intelligent optimization algorithms have emerged one after another, and are widely used in intrusion detection. As a highly sparse model, the relevance vector machine (RVM) is very suitable for intrusion detection scenarios with large scale data. The selection of the parameters of the intrusion detection model directly affects the performance of intrusion detection. Therefore, the selection and determination of parameters is a very critical point to obtain better detection performance. At the same time, the classification performance of RVM obviously depends on the kernel function. To ensure the diversity of kernel function, we adopt a hybrid kernel function formed by linear combination. In addition, RVM is easy to fall into the local optimum, and it has large initial value randomness and poor convergence. Aiming at the limitations of the RVM algorithm, we propose a novel WOA-HRVM model, which optimizes the parameters of the hybrid kernel RVM by WOA algorithm to obtain better performance. The proposed WOA-HRVM is evaluated on NSL-KDD and CICIDS2017 dataset. Compared with other algorithms tested, the proposed WOA-HRVM algorithm significantly improves the accuracy and speed of intrusion detection.