Currently, most fraud detection solutions for financial crime and tax fraud are focused on risk assessing entities. That is to say, they evaluate the risk of each individual or businesses separately. While this helps prioritize risk-based investigations by alerting investigators to the likelihood of fraud, it does not necessarily maximize deterrence. Therefore it does not always lead to optimal audits.
The current process for fraud detection almost always requires the use of hybrid analytics for scoring businesses or individuals. This typically involves a combination of methods such as business rules, predictive modeling, anomaly detection, social network analysis and other tactics.
Each entity is scored taking into account a wide range of attributes. These attributes include information about financial transactions (volume, frequency and parties involved), socioeconomic characteristics (such as income and profession) as well as information about direct and indirect associations with known fraudsters. Once a score for each entity has been established, alerts are generated for entities with a high risk score. The alert then forms the basis for opening a case which requires the involvement of an investigator. However, the process does not always facilitate optimal audits.
Work towards more optimal audits
The existing technologies have helped tax authorities, welfare departments, financial institutions and other users of fraud detection solutions to know better than ever before which entities are involved in fraudulent activity.
Although this is a major step forward in fraud detection, it introduces an apparent challenge. Pursuing every detected fraudster is a costly activity. While tax authorities and financial compliance agencies have increased resources to match the demand for investigations, a more sustainable solution is required.
Developing analytics to optimise audits and investigations in a way that will facilitate behavioral change and deter fraud before is committed is the most effective way to tackle the problem in the long run.
But how can this be done? By focusing on the network of fraudulent activity, not just the individual entities. Prioritising and spreading investigations across different high risk networks of individuals or businesses rather than focusing on entities with high risk will optimise audits and maximise deterrence.
This approach facilitates strategic detection by identifying the root cause in each network. Plus, it ensures, through network communications, that potential fraudsters become aware of the fact that they are constantly subject to risk assessments, and if they commit fraud they will be detected.
Although this has been a well-known principle for many decades, the technology and data required to identify networks has emerged only recently. Social network analysis can identify associations of entities involved in underlying fraud for many industries. The same technology can also be used for optimising investigations in a way that will maximise deterrence.
One way this can be achieved is by moving away from entity level alert generation to network level alerting. Generating alerts at the network level helps inspectors become more strategic in their investigations. Instead of focusing their efforts on uncovering the fraudulent activity of a particular individual they can uncover the fraudulent activity of the entire network in one investigation.
This change will maximise the yield, or revenue recovered, from the investigation. More importantly, it will prevent and deter future fraudulent activity given the visibility of the investigation in the entire network. Finally, the fraud network is likely to break up completely after a single investigation.
Learn more about using network analysis for fraud detection in the whitepaper, A Layered Approach to Fraud Detection and Prevention.