Binary Division Fuzzy C-Means Clustering and Particle Swarm Optimization Based Efficient Intrusion Detection for E-Governance Systems

With the rapid rise of technology, many unusual and unwanted patterns have been observed in the communication network andrespective systems. This may be attributed to the increase of external threats that cause many security concerns. Such anomalies and unusual behavior lead to a strong need of studying and designing the Intrusion Detection Systems and Clustering. Currently,a variety of clustering methods and their combinations are used to develop an efficient intrusion detection system, but some metrics like low detection rate and high false alarm rate make these models unsatisfactory. The problem of local minima for clustering technique makes their search ability less efficient. An evolutionary technique called particle swarm optimization algorithm, that is based on swarm intelligence, shows a high global maxima search capability. In this paper, these two techniques have been combined to present a novel approach called fuzzy based particle swarm algorithm for the implementation of intrusion detection system. The experiment was conducted on a new data set called Kyoto data set with more number of anomalies. The obtained results were compared with two traditional clustering techniques based on K-Means and Fuzzy C-Means. It was observed that the proposed algorithm outperformed the other two traditional methods on the basis of the Detection Rate and False Alarm rate. In past some researchers have presented the combination of Fuzzy Based Particle Swarm Optimization algorithm to improve the intrusion detection rate,but this rate has been further improved because thealgorithm performance depends on the termination condition and the fitness function value which are new in the proposed algorithm. Moreover, cluster numbers have been considered differently in the past, whereas the proposed algorithm works only on binary clustering.