Intrusion Detection and Classification Based on Deep Learning

Cyberattacks aiming to disrupt the confidentiality, integrity and availability of systems by penetrating the network infrastructure of organizations are becoming increasingly widespread. These attacks carried out by attackers cause anomalies in normally functioning networks. Detection of these intrusions have of great importance in the protection of networks. Basically, Network Intrusion Detection Systems are tools that prevent and detect malicious activities or policy violations against networks by monitoring network traffic. In the scope of this study, supervised learning classification-based RNN, LSTM and GRU algorithms for intrusion detection on networks are applied comparatively on the UNSW-NB15 dataset. The main objective of the study is to compare the success of deep learning algorithms and reach the most appropriate model for intrusion detection and classification. The accuracy values of the models are 98% and FPR values are 0.014, 0.011 and 0.011 for the RNN, LSTM and GRU models, respectively.