Study on distributed network anomaly attack detection method based on machine learning

To overcome the problems of traditional methods such as low detection accuracy, high false alarm rate and long detection time, a distributed network anomaly attack detection method based on machine learning is proposed. Firstly, the local density of network operation data points is estimated by combining the Gaussian kernel and cut-off check, and the network operation data is clustered by the DPCA algorithm. Secondly, through the constructed attack model, abnormal attack characteristics are determined and important features are screened. Finally, the naive Bayes in machine learning is used to determine the attribute characteristics of each category in the clustering results. Match the category attribute feature with the important feature to get the anomaly attack detection result. The experimental results show that the maximum detection accuracy of this method is 98%, the average false alarm rate is 2.64%, and the detection time varies between 0.25 s and 0.68 s.