Improved Nonlinear Fuzzy Robust PCA for Anomaly-based Intrusion Detection

摘要

Among the most popular tools in security field is the anomaly based Intrusion Detection System (IDS), it detects intrusions by learning to classify the normal activities of the network. Thus if any abnormal activity or behaviour is recognized it raises an alarm to inform the users of a given network. Nevertheless, IDS is generally susceptible to high false positive rate and low detection rate as a result of the huge useless information contained in the network traffic employed to build the IDS. To deal with this issue, many researchers tried to use a feature extraction methods as a pre-processing phase. Principal Component Analysis (PCA) is the excessively popular method used in detection intrusions area. Nonetheless, classical PCA is prone to outliers, very sensitive to noise and also restricted to linear principal components. In the current paper, to overcome that we propose a new variants of the Nonlinear Fuzzy Robust PCA (NFRPCA) utilizing the popular data sets KDDcup99 and NSL-KDD. The results of the conducted experiments demonstrated that the proposed approaches is more effective and gives a promising efficiency in comparison to NFRPCA and PCA.