Deep Neural Network Architecture for Anomaly Based Intrusion Detection System

This paper presents a deep learning neural network for classifying the network traffic data. In this regard convolutional neural network with rectified linear activation function has been employed. The class score is computed at the final fully connected layer using the dropout mechanism. A k fold cross validation has been carried out to validate the model and the value of k is set to 10. NSL-KDD cup 1999 dataset has been used in the experiment to classify between normal state and 4 different attacks. The accuracy of the suggested model for NSL-KDD cup 199 along with another state of the art techniques is presented to show the effectiveness of the suggested model.