The Applied Neurocomputing Centre at Didcot, Oxfordshire-based AEA Technology has been working to develop a system that detects organised credit card fraud. The work was funded by Barclays Bank Plc. The idea is to determine the patterns in which people are using credit cards as a regular means of income and to discover how much of the fraud is organised as opposed to opportunist, in an effort to combat fraudulent transactions in the future. Organised crime detection lends itself to a neural network-based approach. At the moment the only way of tracking such crime is very haphazard – basically by tip-off to police. Barclaycard Ltd supplied a six-month snapshot of fraud data from 2,000 credit card accounts, some 40,000 transactions, all of which had some continental European transactions on them. One database was set up containing information such as the date of the theft of the card; the date it was reported stolen; how it was stolen; and general information such as the expiry date. A second database included information on all the transactions that took place after the cards were parted from their rightful owners. This included the type and location of shop that goods were purchased in; the amount spent; and the period over which the transactions took place. A control set of fraud data that was already known about was also included by Barclaycard in the research data to see if the AEA system was capable of finding it. The researchers fed this transaction data into a Kohonen neural network to create clusters of data and find similarities and differences in the fraudulent transactions. The researchers came up with a breakdown according to transaction category, country of purchase and transaction value and then developed transaction and spend curves. Training was done on an RS/6000 for 20 minutes, but later this was found to be overkill and one minute of training on a standard 80486-based machine was sufficient. The network was highly sucessful in finding groups of organised crime, concluded Iain Strachan of AEA. The one Barclays asked us to identify was easy. Other patterns of fraud that Barclays didn’t know about were also found. The control set of data was organised lorry driver fraud, which incidentally was investigated by the UK investigative television programme, The Cook Report. Drivers were using stolen cards to pay for petrol and toll road costs over to France and Spain. It can cost something like รบ200 to take a lorry across France using toll roads. 15% of the card fraud in the study took place at service stations. Another pattern found was the use of the same card in disparate towns in a short space of time. In these cases it would not be possible for one person to travel the distances in that amount of time, showing that stolen cards were being duplicated. Unsurprisingly Barclaycard is keen to keep much of the information the system came up with under wraps because of the competitive advantage it could offer if used commercially. Negotiations are currently underway between AEA and Barclays to see if the research can be taken one step further and used by the credit card company as a business application. There are important problems to be overcome if this is to happen because even if a pattern of fraud is detected, it is not easy to stop. Not all retail outlets check cards online, for example. Another factor is the priority the crime is given by the police, because it is generally considered that only the banks lose out through such crime. In practical terms it would also be difficult to get police on the Continent to take part in combatting such crime.
