基于 PCA 和 ELM 的网络入侵检测技术

针对基于传统 BP(Back Propagation)神经网络算法的入侵检测技术收敛速度慢和检测率不高的问题, 提 出了一种基于主成分分析(PCA: Principal Component Analysis)和极限学习机(ELM: Extreme Learning Machine ) 算法相结合的入侵检测方法。 对提取的特征矩阵采用了 PCA 降维, 并使用 ELM 算法对 4 类常见的攻击类型进 行了多分类检测。 实验结果表明, 该方法正确率高达 98. 337 5%, 检测时间仅 1. 851 7 s, 与传统方法相比缩短 了 2 ~6 倍, 同时还提高了检测率和精度, 降低了误报率和漏报率。 最终改善了正确率、 误报率、 漏报率、 检测 率、 精度和测试时间 6 项指标。

Network Intrusion Detection Based on Extreme Learning Machine and Principal Component Analysis

Because the intrusion detection based on traditional BP (Back Propagation) neural network has deficiency in the convergence speed and detection rate,a method based on ELM (Extreme Learning Machine)using PCA(Principal Component Analysis) is proposed.We use PCA to reduce the dimension of the extracted eigenmatrix and use ELM to detect four types of common attacks.The experimental results show that the accuracy of the proposed method can reach 98.337 5％,and the detection time is as fast as 1.851 7 s.This method also improves the detection rate and precision,and reduces the false positive rate and false negative rate.The proposed method achieves the improvement of these six criterions.