An improved WaveCluster algorithm based on ICA
WiCOM'09 Proceedings of the 5th International Conference on Wireless communications, networking and mobile computing
Hi-index | 0.00 |
In this paper, we introduce a clustering algorithm for Intrusion Detection based on WaveCluster algorithm and an entropy-based characteristics screening algorithm. WaveCluster algorithm has a low time complexity when the data are low-dimensional, but on the contrary, the actual network data are high-dimensional. So we reduce the dimension of the network data using characteristics screening before they are clustered. And the algorithm inherits the WaveCluster’s advantage of multi-resolution, adaptive, and not requiring specific pre-determined parameters. We can rapidly and accurately identify arbitrarily shaped clusters at different scales and degree to find intrusion effectively. Experimental results on KDD Cup 1999 data sets show that the detection rate of the algorithm is higher than the algorithm in the reference. The time complexity of the algorithm is low.