Enhancement of Intrusion Detection Using Back Propagation Algorithm

Abstract

Intrusion detection is an important component of secure information systems. This paper concerns the issue of identify main input features in building an intrusion detection system (IDS). Since exclusion of the unimportant and/or useless inputs leads to a simplification of the problem, faster and more accurate detection may result. Feature ranking and selection, therefore, is an important issue in intrusion detection. Since Neural network (NNs) tend to scale better and run faster than other with higher accuracy, we apply the technique of deleting one feature at a time to perform experiments on NNs to rank the importance of input features for the DARPA collected intrusion data. Important features for each of the five classes of intrusion patterns in the data set are identified. It is shown that NN-based IDSs using a reduced number of features can deliver enhanced or comparable performance.