基于改进粒子群的加速K均值算法在入侵检测中的研究

Research of Accelerating K-Means Algorithm Based on New Particle Swarm Optimization for Intrusion Detection

针对K均值算法在入侵检测中受初始聚类中心影响而不能获得全局最优,粒子群算法容易出现早熟收敛的问题,提出了一种基于改进粒子群的加速K均值聚类入侵检测算法(NPSO-AKM),对K均值算法和粒子群算法进行了改进和结合,使得算法具有较高的处理速度和全局搜索能力。针对NPSO-AKM是一种聚类算法的特点,设计了基于NPSO-AKM的入侵检测模型。针对实验数据集的特点,设计了交叉法用于构建高质量的训练数据集。通过实验分析和比较,该模型有较好较快的全局收敛能力,并能在入侵检测中获得令人满意的检测率和误检率。

To overcome deficiency of global search ability for K-Means algorithm impacted by initial centroids in intrusion detection and premature convergence for particle swarm optimization algorithm, an accelerating K-Means algorithm based on new particle swarm optimization(NPSO-AKM) was proposed. In the algorithm, K-Means and particle swarm optimization algorithms were improved and integrated, which provided the algorithm with relatively higher processing speed and better global convergence. To address the features of clustering algorithm for NPSO-AKM, an intrusion detection model based on NPSO-AKM was designed. For experimental dataset, the cross-method was achieved to build high-quality training dataset. The experiments show the model has relatively good and fast global convergence, and can get satisfied detection rate and false alarm rate in intrusion detection.