Challenges

Identity fraud

This type of fraud most commonly occurs when a fraudster uses a stolen identity of an unaware victim and tries to use this profile to apply for money anonymously. A high percentage of fraudulent transactions and instability in fighting fraud result in growing numbers of unpaid loans. Thus digital lenders seek solutions allowing them to identify anomalies in the network connection or unusual behaviour.

Investment fraud

It is one of the most commonly used practices in the digital finance space. With social engineering techniques, fraudsters manipulate victims to get their credentials, access to their PCs (via remote desktop protocols), personal documents or even their money. Since fraudsters can hide behind the “real” identity of a person who even won’t notice the online fraudulent activity, financial institutions need profiling to recognize if they are dealing with a regular user or an excellent imitation.

Solution

Based on constant feedback and fraud reports from the client, Nethone constantly improved the performance of antifraud models on Smartney's traffic. The key metrics - precision and recall, were enhanced by over 10p.p. each in less than a year of cooperation. That resulted in better identification of fraudsters and releasing sales potential with a higher acceptance of high-quality customers, even though a new type of fraud appeared, which had not previously occurred with this client. To answer Smartney's needs, Nethone engineered open ports detection, and a dedicated Data Scientist extracted features based on a person's place of residence and postal codes geolocalization.

Nethone proposed a proprietary ML model, which has been extracting over 5,000 attributes about each Smarney customer–fingerprinting their device, using remote desktop protocols or checking their behaviour to spot any repetitive patterns and abnormalities that may imply fraud. The connection graphs in the Nethone panel helped Smartney visualize and understand how fraudsters operate–their habits and behaviours. That resulted in a 3-times higher fraud detection for applications marked by Nethone than for apps not verified by Nethone signals.

Models are constantly refined and modified by a dedicated Data Scientist based on the acquired knowledge and close cooperation with the client. And in case of any fraud attack, Nethone can quickly intervene by adapting logic to block fraudulent activity.

Results