A network abnormal flow analysis method based on improved SOM

A network abnormal flow analysis method based on improved SOM neural network is proposed in this paper. This method uses the known characteristic flow data to train the SOM neural network, and mark normal flow data and abnormal flow data clustering neurons according to training results. According to the best matching neurons of the testing data to judge whether the abnormal flow happen when detecting. To verify the effectiveness of tests, using the KDD cup99 evaluation database as the network training and test data, the detection results of abnormal flow detecting methods based on improved SOM is compared with the detection method based on classic SOM. Simulation experimental results show that the analysis method based on improved SOM have high detecting rate, short training time and strong generality etc.