Intrusion Traffic Detection and Characterization using Deep Image Learning
Gurdip Kaur, Arash Habibi Lashkari, Abir Rahali
Abstract
The security community has witnessed an unprecedented upsurge in cyber attacks in recent years. These attacks have proved to be successful in achieving their catastrophic objectives. Intrusion detection and prevention systems remain the principal point of defense against these devastating attacks. However, most of the anomaly datasets in the past are neither up-to-date nor reliable. Researchers used various machine learning techniques to classify anomaly-based attacks due to their capability to keep pace with the evolution of such attacks and gave encouraging predictions. Nevertheless, deep neural networks turned out to be revolutionary in detecting and characterizing such intrusions. In this paper, first of all, we propose an imagebased deep neural model to classify various attacks by using two comprehensive datasets called CICIDS2017 and CSE-CICIDS2018. Secondly, we provide a list of best network flow features to identify these attacks. We deploy a convolutional neural network model to classify and characterize different attacks with promising evaluation results.