A feature selection approach implemented with the Binary Bat Algorithm applied for intrusion detection

The large number and various technological solutions adopted by many enterprises, overwhelms security systems, which must acquire informations from all these diverse sources and interpret them. Furthermore, the proliferation of more complex cyber threats imposes a difficult task for information security assurance. Therefore, it is clear that new solutions are required. In this paper we propose a wrapper feature selection approach that combines an improved version of the Binary Bat Algorithm with two classifiers (C4.5 and SVM). We test our proposed model on the NSL-KDD dataset and empirically prove that our method can boost the performance of the classifiers and outperforms BBA and BPSO in terms of attack detection rate and false alarm rate, obtained after a fewer number of iterations. Furthermore, we reduced the number of features with almost 64% and improved the performances of the classifier, even for unknown intrusions.