Research on proactive network security detection algorithm based on machine learning

In this paper, we introduce anomaly detection on the basis of misuse detection, and propose an active network security detection algorithm based on machine learning. Firstly, the anomaly detection model based on clustering marker K-means is studied, and then the Gaussian process Bayesian optimization method is adopted for hyperparameter optimization, and two biased classifiers are trained to re-detect and classify the detection results whose confidence is less than the fixed threshold to reduce the false positive rate. Finally, the proposed anomaly detection model is used to re-detect the normal results of the misuse detection. In this way, the missed detection of misuse detection can be effectively reduced. The experimental results show that the average false positive rate of the proposed anomaly detection model is reduced. The proposed average accuracy has improved.