支持向量分类机在入侵检测中的应用研究

Research of support vector machine classifiers for intrusion detection

为解决入侵检测系统的泛化能力问题,分析了多类分类器的理论框架,并综合考虑训练集数据的预处理、交叉验证时间和入侵检测模型准确率三个因素,提出了一种改进的粗细网格参数优化算法。在基于支持向量机的入侵检测模型中,将KDD数据集映射到高维空间,并采用不同的算法对核函数相关参数进行优化。实例仿真计算表明,通过改进的网格搜索法所获得的参数相对来说有明显的时间优势,分类精度和效率得到了提高。

To enhance the approximation and generalization ability of intrusion detection system, theoretical framework of mul- tiple classifiers is analyzed, and factors such as training data pretreatment, cross-validation time and intrusion detection model accuracy is also taken into consideration. In order to get optimal parameters rapidly, a new approach based on grid search is pre- sented. The KDD dataset is mapped into a high-dimensional feature space via the method for intrusion detection based on sup- port vector machine. Different algorithms are applied to optimize the related parameters for kernel function. By using improved grid search method, the acquired parameter has relatively obvious time Superiority. The experimental results prove that the classi- fication accuracy and efficiency are imnroved.