A Network Intrusion Detection Based on Improved Nonlinear Fuzzy Robust PCA

It is acknowledged in security field that intrusion detection systems are a powerful way to detect intrusions in a computer network. Nevertheless, the network traffic used to construct an IDS is huge with useless and unnecessary information. To address this issue, we should retain just the relevant information using a feature extraction method. The most popular technique used in detection intrusions area is Principal Component Analysis (PCA). However, PCA is sensitive to noise and outliers and it is limited to linear principal components. In this paper, we have proposed a new variant of the Nonlinear Fuzzy Robust PCA (NFRPCA) using the two well-known datasets KDDcup99 and NSL-KDD. Experimental results demonstrated that the new NFRPCA gives a promising performance in comparison to NFRPCA and PCA.