A Step Forward to Revolutionize Intrusion Detection System Using Deep Convolutional Neural Network

Cyber-threats have become a devastating issue in the era of technically advanced world where data and information are strenuously used for predictive analytics. To prevail over this situation, the role and responsibilities of cyber-security predominantly increase. Network intrusion detection system (NIDS) has become an indispensable part of cyber-security which analyzes all the incoming and outgoing packets of a network and detects any malicious content. This paper proposes an advance intrusion detection system (IDS) based on optimized convolutional neural network (CNN-IDS) which is an enhancement over basic CNN that chooses the best suited model during the time of training, and later on, the selected model is used for testing purpose. The benchmark dataset NSL-KDD has been used for experimental purpose. The reduction of various unnecessary features from NSL-KDD dataset has been performed using information gain technique, and finally, the model has been evaluated through several machine and deep learning frameworks. Based on the experimental results, it is concluded that the proposed CNN-IDS model outperforms among all the classifiers.