An NMF-Based Traffic Classification Approach towards Anomaly Detection for Massive Sensors
Akira Nagata, Kotera Kohei, Katsuichi Nakamura, Yoshiaki Hori · 2014
For a computer network in the era of big data, we discuss a behavioral anomaly detection system which makes it possible to analyze and immediately detect anomaly traffic behavior. Many sensor devices connect to the network and tend to generate their application traffic at quite a low communication rate. In order to observe necessary traffic information for traffic analysis in a short time, the monitoring system integrates traffic statistics of flows sent from devices which are considered to generate traffic for the same application. It detects anomaly traffic behavior on the basis of application analysis using NMF(Non-Negative Matrix Factorization). This paper describes a basic design of our prototype development.