IMPROVING DETECTION FOR INTRUSION USING DEEP LSTM WITH HYBRID FEATURE SELECTION METHOD

Due to the importance of intrusion detection system, which is considered supportive enhancing network security. Therefore, we seek increase efficiency systems through use deep learning mechanisms. However, algorithms still suffer from problems in process classification and determining presence type attack, causes a decrease rate, an number false alarms, reduces system performance. This due large redundant features that are not relevant dataset. To find solve for this problem, here propose hybrid algorithm based on feature selection technique, helps reaching goal optimally by choosing best most important features. And it works integrating three ways minimize deleting static features, do have much value information gain done before training stage model LSTM as preprocessing CSE-CIC-IDS data set, improving performance By minimizing time processing increasing rate accuracy ratio. The results experiment showed high 99.84%.