A-GHSOM: An adaptive growing hierarchical self organizing map for network anomaly detection
Journal of Parallel and Distributed Computing
| Hi-index | 0.00 |
Neural networks approach is an advanced methodology used for intrusion detection. As a type of neural network, Self-organizing Maps (SOM) is getting more attention in the field of intrusion detection.In this paper, some improvements on SOM algorithm are made in order to increase detection rate and improve the stability of intrusion detection, include: (1) Modify the strategy of “winner-take-all” to decrease underutilized or completely unutilized neurons. (2) Introduce interaction weight which describes the effect between each neuron in the output layer to enhance the relationships between the input pattern and the weights of all the nodes when adjusting weights; The improved SOM is implemented and applied to the intrusion detection. The validities and feasibilities of the improved SOM are confirmed through experiments on KDD Cup 99 datasets. The experiment result shows that the detection rate has been increased by employing the improved SOM.