Detecting abnormal behaviors by using intelligent system

With the rising use of Internet technologies around world, number network intruders and attackers has skyrocketed. For this reason, introducing systems for detecting attacks within security measures stops hackers from gaining access to data. In order detect various forms assaults, development intrusion detection is quite crucial. The election features elimination unrelated information can improve classifier's accuracy execution because traffic dataset contains many useful unhelpful features. Along these lines, in work, three hybrid strategies selecting feature were implemented, which include constant by standard deviation, Quasi variance threshold, gain. These approaches used rank features, after best higher ranking was selected classification detection. tested on classifiers long shortـterm memory, convolutional neural network, CNN-LSTM. From acquired results, a high level detected, achieved cross-breeding selection different Furthermore, classifier highest rate, exceeded value 99.5%.