Finance
Boosting fraud detection accuracy in banking
Document fraud is a growing challenge in the banking sector. We partnered with a Belgian bank to automate their fraud detection system using a combination of traditional and innovative approaches.
To protect confidentiality, we may alter specific details while preserving the accuracy of our core contribution.
Context & objectives
A Belgian bank faced a growing problem with forged documents and fraudulent mortgage applications, which led loans to be wrongfully approved for customers unable or unwilling to repay them.
The bank's fraud detection process relied on individual agencies manually checking each document. When an agent identified a suspicious document, they escalated it to a team of experts for review. Unfortunately, this slow and tedious process was prone to human error, and fraudulent documents could slip through undetected.
The bank asked Agilytic to develop a proof of concept for an automated fraud detection system with two main objectives:
Save employees time by eliminating manual document checks
Improve fraud detection accuracy
Approach
We adopted complementary approaches to develop our automated fraud detection system.
Optical character recognition
On the one hand, we used traditional approaches based on optical character recognition (OCR) to detect alterations in text documents. These compare value and format inconsistencies in documents, such as spaces between letters.
The main advantage of OCR is that does not require large amounts of training data so long as documents follow a specified template that the system can analyze.
Steganographic detection
On the other hand, we adopted an innovative approach involving steganographic fraud detection to detect manipulations on image documents.
This graphical inspection technique can detect fraud in even the trickiest cases, going beyond what our eyes can see and extending down to the pixel level of these image documents. We tested this algorithm on thousands of pay slips and correctly identified 70% of forged documents.
Combining these approaches significantly improved the system's ability to detect fraud accurately. OCR offers speed and simplicity, while steganography delivers high accuracy in cases where fraud might otherwise go unnoticed.
Results
We designed the system to help agencies quickly and easily check documents for fraud. It either detects fraud directly or flags suspicious elements for agents to review. Using the system's guidance, teams can evaluate documents much faster, streamlining the entire process.
A key point to remember when automating processes that previously involved highly manual tasks is that achieving 100% automation isn't always realistic. However, automating even a small portion of high-volume, error-prone work can deliver significant benefits to an organization.
Given the high cost of fraud detection for banks, automation is an incredibly worthwhile investment: beyond saving time and improving accuracy, automated fraud detection helps banks manage risks more effectively by identifying and addressing fraudulent activity before it escalates.
To safeguard confidentiality, we may modify certain details within our case studies.