On the Use of Unsupervised Techniques for Fraud Detection in VoIP Networks

摘要: In traditional telecommunication networks, fraud is already a threat depriving telecom operators of huge amounts money every year. With the migration from circuit-switched networks to packet-switched it expected that this situation will worsen. chapter, we present an unsupervised learning technique for classifying VoIP subscribers according their potential involvement in activities. This builds signature each subscriber describe his or her typical behavior. Then used as basis comparison evolves over time. An implementation prototype was developed and assessed against real-life data delivered by provider. The results were proven reasonable comparing another method, namely Neural Network Self Organizing Map (NN-SOM).