Offline Fraud Call Detection By Using Artificial Neural Network

Abstract

Telecommunication Fraud can be defined as an illegal use of telecom infrastructure likemobile communications with an intention for not paying services, misuse of voice calls (or data, SMS, MMS), cheating in subscriptions and using illegally services in the networks of telecom providers. Telecommunication fraud has continuously been causing significant financial loss to telecommunication customers in the world for several years.Traditional approaches to detect telecommunication frauds usually rely on constructing a blacklist of fraud telephone numbers. However, attackers can simply evade such detection by changing their numbers, which is very easy to achieve through VoIP (Voice over IP). To solve this problemfeed-forward neural network is usedas a software capable of detecting fraud call for offline data, calldetailed recordswere collected, prepared to be coded and analyzed through extracting features, then rules has been built to check whether the call was normal or fraud. constituting Feed-forward neural network system was done, data has beendivided into two datasets which were Training data and Testing data, learning and validation for neural network were done, then the mean square error has been measured.This technique was designed for telecommunication fraud call detection, depending on analyzing contents of a call Instead of relying on call type and caller number then constructing a blacklist of fraud numbers.It was found that the mean square error has reduced by increasing the percentage of testing data, as well as increasing training data leads to minimizing the validation and the mean square error would be minimized, the obtained results had led to increased system accuracy to detect fraud call.