Using particle swarm optimization in fuzzy association rules‐based feature selection and fuzzy ARTMAP‐based attack recognition

ABSTRACT Feature selection is a classic research topic in data mining, and it has attracted much interest in many fields such as network security. In addition, data mining approaches such as fuzzy association rule mining (FARM) can improve the performance of intrusion detection systems. In this study, a FARM‐based feature selector is proposed in order to reduce the dimension of input features to the misuse detector. Furthermore, a fuzzy ARTMAP neural network is used as the classifier. The accuracy of the proposed approach depends strongly on the precision of the parameters of FARM‐based feature selector module and fuzzy ARTMAP neural classifier. Particle swarm optimization (PSO) algorithm is incorporated into the proposed method to determine optimum values of parameters. In this way, the performance of PSO algorithm is compared with genetic algorithm (GA), as well. Experimental results indicate that PSO outperforms GA both in population size and number of evolutions and can converge faster. This is very important for enhancing the mining performance in large datasets such as intrusion detection datasets. When compared with some other machine learning methods, the proposed system indicates better performance in terms of detection rate, false alarm rate, and cost per example. Copyright © 2012 John Wiley & Sons, Ltd.