Developing new deep-learning model to enhance network intrusion classification

Network traffic has recently known tremendous growth, and it is set to explode over the next few years. Alongside the increase in traffic, network attacks have become more complex, advanced, and efficient. Therefore, intrusion detection systems (IDS), among other countermeasures, must be adapted accordingly to the development of new threats, which implies the design of new detection methods with better accuracy and adaptability characteristics. Furthermore, methods training and validation can be conducted only on the grounds of adequate datasets. Therefore, using updated datasets and efficient classifiers are key factors. In this paper, we introduce a new Deep Neural Network (DNN) based IDS model for network traffic classification. Experimental analysis is carried out using both the CICIDS2017 dataset, which contains many new and up-to-date attacks alongside the well-known NSL-KDD dataset. The results are analyzed based on different performance metrics. The proposed model proves an accuracy of 99.43% and 99.63% using CICIDS2017 and NSL-KDD datasets, respectively. Furthermore, the performance of the proposed DNN model has been compared with the most recent schemes and higher accuracy is achieved.