%0 Journal Article
%T Intrusion Detection with Machine Learning and Feature Selection Methods
%A Halil ARSLAN
%A O£¿uz KAYNAR
%A Yasin G£¿RMEZ
%A Yunus Emre I£¿IK
%J -
%D 2018
%X As computers and the internet become indispensable elements of everyday life, the number of internet sites and web based applications have increased rapidly. The sharing of information, ideas and money through internet sites and applications has made information security an important and actual issue. Softwares such as security walls and virus programs are used for computer and system security have not been enough. For this reason, it is aimed to solve the possible threats by detecting abnormal behaviors with the intrusion detection systems that are proposed as an alternative to existing software. In this study, new data sets are obtained by applying chi square, information gain, gain ratio, gini coefficient, oneR, reliefF, genetic, forward and backward feature selection algorithms on KDDCup99 dataset generated for intrusion detection systems. In order to compare the new data sets obtained with the original size data set, different models are developed by using k-nearest-neighbor, support vector machines and extreme learning machines. In this study, all three methods for test data having best results are compared by using mentioned feature selection methods according to metrics such as success rate, precision, false alarm error, f-measure. As the result of experiments, it is seen that feature selection methods increases the success rate and enables models to function faster for all three classification methods. Besides, it shows that extreme learning machines are able to be integrated to intrusion detection systems conveniently and be used as an alternative method because of reasons such as having high success rate, being faster than other classification methods and having simple training algorithm
%K Sald£¿r£¿ Tespiti
%K Makine £¿£¿renmesi
%K £¿znitelik Se£¿imi
%K A£¿£¿r£¿ £¿£¿renme Makineleri
%U http://dergipark.org.tr/gazibtd/issue/36840/368583