An Empirical Study on Applying Anomaly Detection Technique to Detecting Software and Communication Failures in Mobile Data Communication Services

Abstract

A mobile operator offers many mobile data communication services to its users, such as e-mail, Web browsing, company proprietary services. Although quick detection of communication and software failures are important to improve users' satisfaction, such a quick detection is difficult because the services are served by many servers, network nodes and mobile terminals. Thus we developed the anomaly detection technique for the mobile operator's network to detect anomalies caused by communication failures such as server and network halts. Our technique is based on the observation that users reconnect to servers many times when a communication failure occurs. It is useful not only to detect such communication failures, but also those which would be caused by software failures of mobile terminals and servers. This means that a mobile operator would be able to detect software failures missed at the testing period. In this paper, we empirically study how our technique is used to detect software failures of mobile terminals.