Fraud Detection on International Direct Dial Calls

Abstract

Telecommunication operators’ losses a considerable amount of their revenue due to fraudulent activities. These even causes poor service quality which leads to a bad customer experience. Technological advancements favor fraudsters with new tools and techniques. Subscriber Identity Module box (SIM box) bypass fraud is such a type which immerged with Voice Over IP (VoIP) technology. These are used to bypass the legitimate interconnect gateway route for International Direct Dials (IDD) calls considering the tariff is much higher than locally terminated, thus benefitting from the difference. These activities can be detected by various approaches, such as, profiling, identifying calling patterns or test call generation, etc. But it must be in an expeditious way in order to minimize the loss. In this research, Call Detail Records (CDR) are used in an hourly manner to differentiate fraudulent from legitimate subscribers with better performance. Artificial Neural Network (ANN) was used as the classification technique in building the prediction model. Training data set was prepared over one-month data. Results shows that the system performed with 99.59% accuracy.