基于动态特征选择的恶意网络行为检测仿真

Simulation of Malicious Network Behavior Detection Based on Dynamic Feature Selection

恶意网络行为检测中易受噪声数据干扰,影响检测效果。为了降低检测错误率,提出一种基于动态特征选择法的恶意网络行为检测方法。构建超融合架构,将网络数据输入到架构中,采用超融合框架中的自编码器对网络数据实行降维处理,运用改进的PCNN模型消除数据中存在的噪声,避免检测过程受到噪声干扰,提升检测准确率。采用动态特征选择法更新网络数据的特征权重值,利用特征加权熵完成特征选择,剔除权重值小于阈值的特征数据,动态选择重要的特征分量,降低检测时间,通过聚类算法识别出恶意行为簇,完成恶意网络行为检测。实验结果表明,所提方法的检测时间短、检测准确率高、检测错误率低,可以有效保证网络运行的安全性。

At present,malicious network behavior detection is easily disturbed by noise data.In order to reduce the detection error rate,a method of detecting malicious network behaviors was put forward based on dynamic feature selection.First of all,a hyper-converged infrastructure was constructed,and then network data was input into it.Moreover,an auto-encoder in the hyper-converged infrastructure was used to reduce the dimension of network data,and then the improved PCNN model was used to eliminate the noise from data,thus avoiding noise interference and improving detection accuracy.Furthermore,the dynamic feature selection method was adopted to update the feature weight of network data.Meanwhile,the feature-weighted entropy was used to complete the feature selection and thus remove the feature data whose weight value was less than the threshold value.In addition,important feature components were dynamically selected to reduce detection time.Finally,malicious behavior clusters were identified by the clustering algorithm.Thus,the detection was completed.Experimental results show that the proposed method has short detection time,high detection accuracy and low error rate,so it can effectively ensure the security of network operation.