针对入侵检测系统特征报警聚类质量低、冗余告警的不足，提出基于改进混沌自适应粒子群优化的IDS 特征 报警聚类方法。该方法结合混沌算法特性和改进粒子群算法自适应惯性权重系数以及对非线性动态学习因子进行改善，引导 粒子群在混沌与稳定之间交替波动，保证粒子运动惯性，更利于趋近最优。本方法能够克服PSO算法的过早收敛、“惰性”反 应等缺点，利于聚类中心更能趋向全局最优。实验结果表明，本文粒子群参数改进算法提高了特征报警聚类质量，具有较高的 检测率和较低的误报率。

Aiming at the the low quality of feature clustering and excessive redundant alarms in IDS, an IDS alerts clustering algorithm based on novel chaotic particle swarm optimization is proposed. It combines the characteristics of chaotic PSO algorithms, adaptive inertia weight coefficient, and non-linear dynamic learning factor, so as to make particles move between the state of chaos and stable. It guarantees the particle motion inertia, and approaches the optimal value. It also can overcome the problems of premature convergence and "inert" reaction of PSO algorithm, and help the center of cluster to find the global optimal solution. The experiment results show that the improvement of particle swarm parameters improves the quality of feature clustering in IDS alarm, and has higher detection rate and lower false detection rate.