网络攻击检测的门控记忆网络方法

o solve the problem of large-scale network attack detection, this paper proposed a gated memory network method, based on word vector feature representation and recurrent neural network. Firstly, the proposed method transformed the network request data into low-dimension real-value vector sequence representation. And then, it extracted the features of request data by using the memory ability of gated recurrent neural network. Finally, it adopted the logistic regression classifier to achieve automatic detection of network attack. On the CSIC2010 public data set, this method achieves 98.5% 10-fold cross-validation F1-score. Comparing with traditional methods, it can effectively improve the precision and recall rates for detecting network attack. The proposed method can detect network attacks automatically and has good detection results.