ABSTRACT
Total losses due to telephone fraud are estimated to be as high as billion dollars a year. While a significant amount of work has been done to develop secure interfaces, few of these interfaces are actually deployed. The less sophisticated methods of security that are in place are often rendered ineffective by technical means or by human error. This work assumes that there will always be some unauthorized users and that it is best to supplement a secure interface with the capability of detecting the unauthorized user. This paper describes ongoing research that attempts to use neural networks to detect unauthorized calling-card activity. The results are based upon analysis of actual calling-card activity obtained from the public switched networks
