Ebook Adaptive Fraud Detection
In the United States, cellular fraud costs the telecommunications industry hundreds of millions of dollars per year (Walters and Wilkinson 1994; Steward 1997). One kind of cellular fraud called cloning is particularly expensive and epidemic in major cities throughout the United States. Cloning fraud causes great inconvenience to customers and great expense to cellular service providers. Existing methods for detecting cloning fraud are ad hoc and their evaluation is virtually nonexistent. We have embarked on a program of systematic analysis of cellular call data for the purpose of designing and evaluating methods for detecting fraudulent behavior.
Cloning fraud is one instance of superimposition fraud, in which fraudulent usage is superimposed upon (added to) the legitimate usage of an account. Other examples are credit card fraud, calling card fraud and some forms of computer intrusion. Superimposition fraud typically occurs when a non-legitimate user gains illicit access to the account or service of a legitimate user. Superimposition fraud is detectable if the legitimate users have fairly regular behavior that is generally distinguishable from the fraudulent behavior.
This paper presents a framework, and a corresponding system, for automatically generating detectors for superimposition fraud. We have applied the system in the domain of cellular cloning fraud. Under the framework, massive amounts of cellular call data are analyzed in order to determine general patterns of fraud. These patterns are then used to generate a set of monitors, each of which watches customers’ behavior with respect to one discovered pattern. A monitor pro les each customer’s typical behavior and, in use, measures the extent to which current behavior is abnormal with respect to the monitor’s particular pattern. Each monitor’s output is provided to a neural network, which weights the values and issues an alarm when the combined evidence for fraud is strong enough.
This article is organized as follows. We rst describe the problem of cellular cloning fraud and some existing strategies for detecting it. We then describe the framework in detail using examples from the implemented system. We present experimental results comparing the system against other known methods for detecting fraud. Finally, we discuss the evaluation and describe issues in the future of automatic fraud detection.
Download
PDF Ebook Adaptive Fraud Detection
Posted in :
