Review On Feature Selection Algorithms For Anomaly-Based Intrusion Detection System

As Internet networks expand, the amount of network threats and intrusions increased, the demand for an efficient and reliable defense system is required to detect network security vulnerabilities. Intrusion Detection Systems (IDS) are a vital constituent of security of a network to avert data illegal usage and misappropriation. IDS deal with massive amount of data movement that comprises repetitive and inappropriate features. The detection rate implementation is frequently affected by these inappropriate features which also munch up intrusion detection system resources. A significant portion in the removal of dissimilar and not used features in IDS is done by the feature selection method. Methods included are data mining techniques, machine learning, statistical analysis, support vector machine models and neural networks. In this paper, we provide review of several algorithms used for anomaly-based intrusion detection systems to improve performance of machine learning classifiers. This paper first summarizes the theoretical basis of IDS, and then discusses the feature selection techniques and their types.