Machine-Learning-based algorithms lend themselves as useful tools for analysing the possibilities of credit fraud. This is typically done by running the algorithms on various transactional data and analysing for hidden patterns.
How it differs from conventional techniques
Conventional statistical techniques for analysing frauds, such as regression analysis, assumes not only the existence of a particular pattern/relationship in the development data but also makes the assumption that such patterns/relationships would stay constant. In reality, such patterns/relationships have drifts. Drifts may have been induced artificially by a fraud perpetrator to avoid detection (E.g. Series of transactions involving small amounts of money, but adding up to a considerable amount) or may have crept in due to the nature of transactions (E.g. On account of seasonal fluctuations).
ML-based algorithms can be used to detect and flag suspect patterns. Such suspect transactions can be subjected to further investigations and courses of action. Extending the argument, the algorithms can be used to reduce false positives as well. For example, seasonal fluctuations in a series of transactions can be flagged separately and allowed to pass (if genuine).
Use Cases
ML techniques can be intelligently applied to a variety of use cases, particularly in the case of frauds:
- Fund Diversion: Machine Learning can be used to detect any unusual patterns pointing towards possible cases of diversion of fund
- Are funds being routed to individuals or groups of individuals on a regular basis?
- Are funds being transferred to third parties on pre-set dates (say, the 2nd fortnight of every month)?
- Are fixed amounts getting transferred? Alternatively, are amounts being broken down into smaller chunks and transferred, to avoid detection?
- A combination of the above situations
- Transactions with Blacklisted Parties: Transactions with parties who are internally or externally blacklisted (E.g. present in the AML watch list). ML-based algorithms can be used to monitor any transactions with such parties.
- End Use Monitoring: Are the funds being transferred totally unconnected to the lines of business of the borrower?
- Network Analysis: Is the beneficiary indirectly related to the borrower? (E.g. Beneficiary is one of the common directors)
With electronic transactions surging, it would be impossible for banks and other financial institutions to keep a tab on transactions manually. ML-based algorithms, working on banks’ data, can provide an effective way to keep fraudulent transactions under control.
Author
Kasthuri Rangan Bhaskar
VP, Financial Services Practice & Risk SME (Lead) at BCT Digital
Mr. Kasthuri is the Risk SME (Lead) at Bahwan CyberTek with profound experience in Market Risk & Credit Risk, and has over 15 years of experience in the BFSI sector. He has experience working with some of the large mainstream BFSI labels in the country.
Kasthuri Rangan Bhaskar
VP, Financial Services Practice & Risk SME (Lead) at BCT Digital
Mr. Kasthuri is the Risk SME (Lead) at Bahwan CyberTek with profound experience in Market Risk & Credit Risk, and has over 15 years of experience in the BFSI sector. He has experience working with some of the large mainstream BFSI labels in the country.
