Identifying Fraud Using Correlation
Using correlation to detect fraud is a relatively recent event. The first application of correlation in a forensic setting was at a fast-food company where correlation was used to identify restaurants with sales patterns that deviated from the norm. These odd sales patterns played an important part in identifying sales numbers that were influenced by fraud. The next correlation application was at an electric utility where correlation was used to identify customers with electric usage patterns that differed from the seasonal norm. The usage patterns together with other red flags were successful at detecting some large-scale electricity theft. A recent correlation application was at a consumer goods manufacturer where correlation was used to identify retailers with coupon redemption patterns that differed substantially from the norm. Again, the redemption patterns together with other red flags were successful at identifying highly suspect patterns in coupon submissions.
Correlation is usually used to detect fraud on a proactive basis. According to the IIA (2004), controls may be classified as preventive, detective, or corrective:
- Preventive controls These controls are there to prevent errors, omissions, or security incidents from occurring in the first place. Examples include controls that restrict the access to systems by authorized individuals. These access-related controls include intrusion prevention systems and firewalls, and ...