Feature Ranking and Selection for Intrusion Detection Systems Using Support Vector Machines

摘要: Intrusion detection is a critical component of secure information systems. This paper addresses the issue identifying important input features in building an intrusion system (IDS). Since elimination insignificant and/or useless inputs leads to simplification problem, faster and more accurate may result. Feature ranking selection, therefore, detection. support vector machines (SVMs) tend scale better run than neural networks with higher accuracy, we apply technique deleting one feature at time perform experiments on SVMs rank importance for DARPA collected data. Important each 5 classes patterns data are identified. It shown that SVM-based IDSs using reduced number can deliver enhanced or comparable performance. An IDS class-specific based five proposed.