August 10, 2020
Cleaning up money laundering
Money laundering is big business but wholly illegal big business. It has an enormously negative impact on local, national, and international economies as well as providing the financial means to fund other criminal activities such as people trafficking and drugs. By definition, money laundering is activity carried out to obscure the source of money that has been obtained illegally.
Writing in the International Journal of Business Intelligence and Data Mining, researchers from the Sultanate of Oman and Saudi Arabia describe a new dynamic approach to identifying suspicious financial transactions that might be part of the chain in a money-laundering scheme.
Abdul Khalique Shaikh of the Department of Information Systems at Sultan Qaboos University in Oman and Amril Nazir of the Department of Computer Science at Taif University, in Al-Hawiya, Saudi Arabia, explain that among the many millions, if not billions, of financial transactions carried out every day, a worrying proportion will be associated with money laundering. Identifying such illegal transactions is difficult especially as the criminals carrying out such transactions are well aware of the tools used by banks and financiers to spot suspicious money movements and as such can usually obfuscate the activity very efficiently.
The team has devised a way to profile individual users and to flag up activity that is genuinely suspicious without the false positives that might otherwise interfere with genuine banking and other financial transactions members of the public might carry out entirely legitimately.
"The approach works based on the dynamic behavior of customer transactions that measures the customer's own transaction history, profile features and identifies suspicious transactions," the team writes. They have tested the approach against realistic data and validated the result with confirmed suspicious customers. The dynamic approach has an accuracy of well over 90 percent, which exceeds that seen with statistical models based on pre-defined rules, the team concludes.