摘要
为了提高多数据流融合异常检测能力,保障网络数据的安全性,提出基于布谷鸟搜索算法(Cuckoo Search,CS)和空间间隔采样的分布式卷积神经网络的多数据流融合异常检测方法.根据多数据流融合异常检测特征进行网络用户的浏览信息特征检测,建立多数据流融合的特征提取模型,构建反映网络安全等级的多数据流融合量化特征分析模型,采用空间欠采样技术进行多数据流非线性特征重组,提取多数据流融合异常检测的统计特征量,根据多数据流融合异常分布状态实现特征检测和识别,结合布谷鸟搜索算法进行多数据流融合异常检测中的自适应寻优.仿真结果表明,采用该方法进行分布式卷积神经网络多数据流融合异常检测的准确性较高,检测概率较高,抗干扰能力较强,提高了网络的安全性.
To improve the detection capability of the multi-data stream fusion,the security of the network data is needed. A method for detecting the multi-data stream fusion of the distributed convolution neural network based on the Cuckoo search and the space interval sampling is proposed. It comprises of carrying out browsing information characteristic detection of a network user according to the multi-data stream fusion anomaly detection feature,establishing a feature extraction model of a multi-data stream fusion,and constructing a multi-data stream fusion quantitative characteristic analysis model reflecting the network security level. The method comprises of carrying out multi-data stream non-linear feature recombination by using a space under sampling technology,extracting the statistic feature quantity of the multi-data stream fusion anomaly detection,and realizing characteristic detection and identification according to the multi-data stream fusion abnormal distribution state. The self-adaptive optimization in the multi-data stream fusion anomaly detection is carried out in combination with a bird’s search algorithm. The simulation results show that the method has the advantage of high accuracy,high detection probability and strong anti-interference ability,and improves the security of the network.
作者
周素青
ZHOU Su-qing(Fujian Polytechnic of Information Technology,Fuzhou 350003,China)
出处
《内蒙古民族大学学报（自然科学版）》
2020年第3期227-232,共6页
Journal of Inner Mongolia Minzu University：Natural Sciences
基金
福建省中青年教师教育科研项目(JAT160735)。
