ANN Mechanism for Network Traffic Anomaly Detection in the Concept Drifting Environment

For timely and effectively detecting the anomalous network traffic, this study proposes an ANN-based machine-learning algorithm that is able to detect abnormal network traffic in the concept-drifting environment. For exploring the timeliness and effectiveness, this study sets up an experiment with a set of real network traffic data provided by Research Center for Information Technology Innovation, Academia Sinica, Taiwan. The proposed mechanism is implemented via TensorFlow and multiple Graphic Processing Units (GPUs). The experiment results show the proposed mechanism can timely and effectively detect the anomalous network traffic via using an enough amount of GPUs.