Network intrusion clustering based on fuzzy C-Means and modified Kohonen neural network
Hongwei Ye1, Lianjiao Zhang1, Xiaozhang Liu2
1School of Electron and Information Engineering, Heyuan Polytechnic,Dong Huan Road, Heyuan, China
2School of Computer Science, Dongguan University of Technology,No.1,Daxue Rd Songshan Lake,Dongguan,China
Kohonen neural network recognizes and clarifies substantive network data, but with a long running time and a slow convergence process. To solve this problem, a network intrusion clustering method is presented in this paper. Specifically, the training data is pretreated using Fuzzy C-Means (FCM). Then some selected data will be trained with using Kohonen neural network. Meanwhile, to speed up the convergence process of Kohonen neural network and to form a better optimized network topology, a neighbourhood function is established for the competing neuron. Each neuron has neighbourhood topology collections. The data simulation results demonstrate the efficiency and effectiveness of the proposed algorithm.