Fuzzy kernel C-means algorithm for intrusion detection systems

Intrusion Detection Systems (IDS) are used as security management systems. There are two approaches of IDS, Misuse Detection (knowledge-based intrusion detection) and Anomaly Detection (behavior-based intrusion detection). Misuse detection is performed by monitoring activities which is suspected as an intrusion based on prior information about specific attacks. While anomaly detection is based on the observation of the activity that is incompatible with the acceptable behaviors in normal conditions and makes it possible to determine new type of attacks in the system. Some Computational Intelligence models have been developed to solve Intrusion Detection Systems problems such as Neural Network and NeuroFuzzy methods. They are chosen because IDS involves large data sets with several different features that can bring out negative effects on IDS accuracy and its computational time. Naïve Bayes, Decision Tree (C4.5) and Kernel Matrix Methods can be used to reduce the number of features at data sets. We propose Fuzzy Kernel C-Means Algorithm as another method to solve IDS problems that we claim provides better results while combined with Kernel Matrix method to reduce the number of selected data features.