"Thousands of mobile phones are stolen each month in South Africa alone; worldwide it is a serious problem," says Paul van der Merwe, Director, Knowledge Integration Dynamics (KID). "One of the most prolific fraud crimes in America today is the use of stolen cellular phone numbers. Much like credit card fraud, the lawful owner of the cellular phone number may not know they are being victimised until significant abuses appear on their monthly bill. This crime requires specialised investigative techniques and procedures."
Van der Merwe says new discoveries and developments in fraud detection systems in the UK, for instance, have shown that the types of call made, the numbers a person rings, the length of calls and the time of day a person makes them are characteristic behaviours that are specific to individuals. And it is these 'biometrics' which can be used to detect fraud.
These fraud detection systems use pattern recognition software built into intelligent agents called sentinels - which assemble behaviour profiles of subscribers on a network. "If the software detects unusual activity on an account, it will send a text message to the mobile phone. The users will then have to punch in a PIN to identify themselves, or if they have a 'pay-as-you-go' phone, top up the credit to validate their ID. If they fail to do so, the phone will be cut off.
"This type of software differs from existing fraud-detection systems because it analyses behaviour dynamically. By not having fixed rules, it can recognise that users might make more calls than normal on New Year's Eve, for example, and let these through."
Van der Merwe says as telephone networks grow in sophistication, so do the possibilities for fraudulent activity on those networks. However, a combination of data visualisation and data mining techniques give useful results in fraud investigations. In a particular case study, he says, BT in the UK found a fraud was based on generating large numbers of calls to bogus premium rate services. The bills for these calls were never paid but BT was required to pass on the revenue generated by them to the service providers.
From an initial set of suspect telephone numbers, premium rate services, and their call records, and using Alta Analytic's Netmap, BT was able to pick out likely fraudulent callers and identify the structure of the criminal gangs. "In order to focus on just the most active premium rate services, BT filtered out much of the data and got a clearer view: the suspected co-ordinating role of the mobile telephone users became more apparent. BT was able to identify different gangs at work, and their organisation, by looking at emergent groups in the filtered data."
Finding emergent groups can be considered as a form of clustering. "BT identified three different gangs, each with its own premium rate service number/s and co-ordinating mobile phone/s. Using this technique helped focus the efforts of the investigators on those who were most likely to be the organisers of the fraud. This resulted in a number of arrests and presentation of the analysis as evidence in court."
Van der Merwe, advises that Netmap, distributed locally by KID, provides detailed analysis and data filtering functionality by identifying and graphically displaying patterns, trends and relationships contained in corporate data. "Companies can easily spot and dynamically filter this information to identify trends and patterns in their business processes.
"Netmap supports train-of-thought analysis and assists users to follow their instincts through visual representations of the data to discover patterns without needing to know any database programming techniques or having to ask an IT person to redesign a report.
"The searching and cross-referencing capabilities of the software enable companies to considerably reduce the costs of fraud," van der Merwe adds.
For details contact Paul van der Merwe, director of Knowledge Integration Dynamics (KID) on tel: (011) 787 8802 or e-mail: email@example.com