Multi-granularity approach for enhancing the performance of network intrusion detection with supervised learning

Intrusion detection system (IDS) is essential in order to overcome the security threats in the network community. IDS examines a large number of features in the data set to detect the intrusion. The process of feature selection is required to reduce the time consumption and storage memory. The data set may contain noisy, uncertain and redundant information. Rough Set Theory (RST) is one of the mathematical tool to reduce the features in the dataset. The quick reduct and relative reduct algorithms are hybridized with the Particle Swarm Optimization (PSO)to improve the effectiveness of the feature reduction. Multi-granularity is applied for network dataset and the reduct is obtained. It is observed that the reduct obtained through the multi-granularity approach produces better result in terms of time than the reduct obtained by the direct application of rough set algorithm.