Elevate Fraud Detection Precision for Digital Lenders

It’s always a precarious challenge to strike the balance in fraud detection for digital lenders — while being too lenient can leave the door open to fraudsters, being too stringent can alienate genuine users and hurt your business development.

‍

The key is, how to precisely spot on the fraudulent intent while keeping the rest of it runs smoothly.

‍

To address this problem, we have enhanced our models to fine-tune the accuracy of our risk assessments, covering our flagship product lines from Application Fraud Detection, KYC++, and Device Fingerprint.

‍

Application Fraud Detection: What’ New?

‍

Customer Stratification Strategy Enhancement

‍

Our updated account information module now supports dynamic customer segmentation. This enhancement allows digital lenders to implement more targeted risk management strategies tailored to each segment, thus enabling a more efficient decision-making.

‍

This update provides digital lenders with a few key benefits:

‍

1. Targeted Risk Control: By accurately identifying and segmenting customers, digital lenders can apply tailored risk control strategies. High-risk customers undergo more stringent verification, reducing the likelihood of fraud, while low-risk customers benefit from faster processing.
2. Optimized Resource Allocation: Dynamic segmentation enables smarter resource distribution, focusing intensive monitoring on high-risk segments while automating lower-risk areas. This approach enhances decision-making, leading to more effective risk management strategies and greater accuracy in fraud detection.

‍

Spatiotemporal Analytics Feature in Device Information

‍

By integrating the time zone district into device information, we enhance our spatiotemporal analytics capabilities. This update allows digital lenders to gain detailed insights into where and when fraudulent activities occur.

‍

Here’s how it looks like that improves precision of application fraud detection:

‍

1. Geographical and Temporal Insights: Identify regions with high fraud incidence and understand the timing of fraudulent activities, enabling targeted, region-specific, and time-based risk controls.
2. Comprehensive & Real-Time Risk Monitoring: Combine spatial and temporal data with other fraud indicators for a complete risk view, allowing for real-time anomaly detection and swift responses to suspicious activities.
3. Advanced Pattern Recognition: Uncover organized fraud schemes by analyzing unusual behaviors and movement patterns, using network analysis to detect complex fraud activities.

‍

A Leap of KYC Enhancement

‍

Overall Optical Character Recognition (OCR) Accuracy

‍

With the integration of Large Language Model (LLM), the system can smartly examine document even if they’re not in the standard layout, slightly blurred or partially obscured. Our document classification accuracy has also been boosted to 99.9%, reducing false positive to just 0.06%. This improvement is reflected on 116 document types in 30 countries.

Detect Forged Document and Enhanced Liveness Detection in Indonesia

1. The latest update includes forged document examination, enabling detection of fake documents and screen captures.
2. Liveness detection solution can now better detect users wearing headscarves and hats, offering seamless and inclusive experience. As a result, the false rejection rate has been reduced by over 80%.
3. During the liveness detection process, it is now viable to detect face under the condition where there’re multiple faces shown on screen. Face that takes up the largest portion of the screen will be detected and verified. This will ensure smoother user experience during the identity verification process.

Increase Document Verification Accuracy in Philippines

The new OCR model can now deliver Philippines Tax Cards at 98.2% accuracy and Philippines Health Cards at 97.31%.

‍

New Labels and Cheating Tools Fortified for Device Fingerprint Technology

Device fingerprint is a crucial tool to differentiate genuine users from potential threats, it can be done so by identifying unique device ID, analysing user behavior and screen through global historical data. We’ve added new risk labels to identify device that falls under high risk through our historical database. The system is also updated with new cheating tools discovered such as emulator, hook tool, group control, VPN, tampered device info, troll store and quick macro.

We're always on our way to perfecting fraud detection, focusing on enhancing precision to help you confidently identify fraudulent activities with fewer false alarms. Contact us today, and let's work together to elevate your fraud detection capabilities and secure your financial operations!

‍