Development of a Method for Fraud Severity Measurement Based On Usage Profiling
The nature of fraud has changed from cloning fraud to subscription fraud, which makes specialized detection methodologies inadequate. Instead, the focus is on the detection methodologies that based on the subscriber’s calling activity or calling pattern, which can be roughly divided into two main...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
2006
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/5200/1/FSKTM_2006_14.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The nature of fraud has changed from cloning fraud to subscription fraud, which makes
specialized detection methodologies inadequate. Instead, the focus is on the detection
methodologies that based on the subscriber’s calling activity or calling pattern, which
can be roughly divided into two main categories: absolute analysis and differential
analysis.
Absolute analysis is capable at detecting the extremes of fraudulent activity. However,
absolute analysis cannot trap all types of fraud especially usage behavior fraud related.
An alternative approach to this problem is to perform a differential analysis against
subscriber’s behavioral patterns. Certain behavioral patterns may be considered
anomalous or abnormal for certain subscriber and potentially indicative of fraud but
would be considered acceptable for another. In order to overcome the uncertainty in behavioral patterns, in this research, we propose to conduct the usage profiling at
individual subscriber level. Usage profiling is a process of generating calling statistic
based on predefined categories, which involve some form of aggregation from
subscriber’s calling activity or CDR.
Usage profiling process will generate two forms of usage profile : usage profile history
(UPH) and current usage profile (CUP). In fraud detection system, comparison of these
two types of usage profile will generate a measure known as fraud severity
measurement. Implementation of the Hellinger distance for measuring a fraud severity,
lack of detection accuracy as this method does not properly define the measurement
scale as the Hellinger distance method will generate variation of values for fraud severity
measurement. Therefore, it is very difficult to define the actual severity level of detected
fraud.
In this research, we propose a new method for measuring fraud severity. The advantages
of the method are detection accuracy and detection speed. With the new method, the
severity measurement scale is properly defined and the detection speed is faster than the
Hellinger distance |
|---|