Anomaly Detection Algorithm Test Bench for Mobile Network Management

Huge amounts of operation data are constantly collected from the performance monitoring and system logs from servers in communication networks. One of the key applications for system operators is to detect anomalies in the data. The anomalies may indicate various types of situations that need attention. They can be results of malfunctions of the hardware or software components. They may also indicate new unseen activity of the customers or possible unauthorized intrusion. There is a large selection of anomaly detection methods to choose from. In this paper we present an application developed for testing anomaly detection algorithms in network management environment. We introduce a test bench implementation and illustrate its properties with two local anomaly detection methods. One is based on SOM (Self Organizing Map) and clustering. Another one is an extension to a basic clustering method, utilizing two layered clustering. The test bench consists of compiled MATLAB code and is excellent for rapid prototyping of algorithms. It can be easily deployed in any network operator’s environment for testing and demonstration purposes without massive software installations. It helps the network operators in selection of the algorithm and its parameters for the final applications