Network Intrusion Classification with Feature Reduction

Abstract

Nowadays, in data technology, data preservation has become a good issue. Computers and completely different security breaches are incessantly attacked by security threats. There are completely different malicious network based or host based attacks that are a massive threat to networks. For protecting computer and networks from attacks and threats, the intrusion detection system has been used that is signature based. An Intrusion detection system gathers needed data to perform analytical actions. So that, it could determine threats that would be generated from a system or organizations inner or outer atmosphere. A great amount of data that is worked by the intrusion detection system, takes varied inappropriate and unnecessary features that result in raised execution time and low detection rate. As a result, in intrusion detection, an undeniable role is played by feature selection. To cover up such aspect various literature were revealed by completely different profound authors. During this analysis, we’ve approached to select some important features based on some feature selection algorithm so that the computational cost, space complexity and intrusion detection time can be reduced. Our analysis over NSL-KDD data set shows that once feature reduction, except Naive Bayes classifier the accuracy of the foremost classifiers is sort of as same because the performance with none feature reduction.