Applying Recurrent Neural Network to Intrusion Detection with Hessian Free Optimization

摘要

With developing a network communication technology, cyber attacks which threaten users safety are increasing. Consequently, many studies are being carried out to protect the user security. One of them is an intrusion detection system (IDS). In this paper, we apply recurrent neural network with hessian-free optimization which is one of the deep learning algorithm for intrusion detection. We use DARPA dataset in order to train and test the intrusion detection model. It was used for the 1999 KDD Cup contest dataset. It composed of 41 features and 22 different attacks. We chose salient features for training the model and analyzed a result of experiment with various metrics. We found that our result is superior to the existing studies through comparing the performance.