RawPower: Deep Learning based Anomaly Detection from Raw Network Traffic Measurements

Machine learning models using deep architectures (i.e., deep learning) have gained path in recent years and have become state-of-the-art in many fields, including computer vision, speech recognition and natural language processing. However, when it comes to network measurement and analysis, classic machine learning approaches are commonly used, heavily relying on domain expert knowledge. In this work, we explore the power of deep learning models to perform anomaly detection on network traffic data, taking as input raw measurements coming directly from the stream of monitored bytes. Our initial results suggest that deep learning can enhance anomaly detection without requiring expert domain knowledge to handcraft input features.