Abstract
Currently, attacks that spread across the Internet have become a major threat to computer and network systems in healthcare. To overcome this threat of intrusion a powerful intrusion detection system (IDS) is necessary. The performance of an IDS can be improved by the selection of a suitable set of feature. The KPCA has different kernel for feature transformation. Thus, need to investigate different kernels and find the best one for the intrusion detection problem, which is necessary to secure systems in healthcare. The focus of this work is to explore performance of various kernels of kernel principal component analysis (KPCA). Three kernels; Gaussian kernel, Polynomial kernel and Tanh kernel are investigated on standard dataset of intrusion detection. The results indicate that Tanh kernel outperforms the other kernels.