基于ICLA-MSIS的无监督流量检测方案研究

In this paper, the problem of labelless prediction in network traffic anomaly detection based on machine learning is studied.Considering that some traffic data sets have no marked data or are difficult to mark, an unsupervised learning approach is needed to solve the problem.This paper presents an automatic data marking scheme based on main features.The accuracy of the area under the curve (AUC) index was improved from multiple angles. Compared with the index of accuracy, the prediction effect of a small number of exceptions could be better demonstrated.The grouped features are mapped to the low-dimensional feature space and marked by clustering method.Finally, several data sets are verified experimentally.The results show that the scheme can improve the AUC value.