Continuous Online User Authentication in Banking
Customer: Large Bank in the CEE Region
Goal: Decrease the number of fraudulent online banking transactions
and reduce the cost of transaction authorization
Challenge
By the first half of 2023, the country's annual incidence of credit transfer fraud exceeded 22 million USD. In 2019, this amount was below 3 million USD; by the latter part of 2020, it only rose to over 4.5 million USD. Additionally, the cost of sending OTPs via SMS has been steadily increasing.
Solution
Graboxy uses movement biometrics to continuously authenticate users. It gathers cursor movement data from the online banking platform through the browser to create identity profiles. If the real-time cursor movement analysis shows a divergence from the user's biometric profile, Graboxy flags the suspicious user sessions. Flagged users can be locked out or re-verified.
How it works
Step 1
The user logs in to the banking account and carries out the usual actions by moving the cursor.
Step 2
Graboxy analyzes the cursor dynamics of the user in real time.
Step 3
By the time the user initiates a transaction, Graboxy authenticates the user.
Authorized user
Silent 2FA completed
Unauthorized user
Request for traditional 2FA
Results
Cost saving on
OTPs via SMS:
-85%
Fraud detection accuracy*:
97%
* Where sufficient training data was available.Authenticated user sessions daily:
100K+
The National Cyber Awards
The Financial Services 2024
Multi-Award-winning cybersecurity platform
built on cutting-edge movement analysis AI by Cursor Insight
Winner of SigWiComp by the German Research Center for AI
Remote Monitoring Security Solution of the Year
Winner of the Mouse Dynamics Challenge Competition of Balabit
BOLD Awards 2024
Boldest Cybersecurity
Our Blog
Gait Dynamics & Body Dimension Report Accepted as Evidence by an EU Court
This week, two leading outlets covered a breakthrough for Cursor Insight and for the broader digital forensics community. Forensic Focus published a new interview with our CEO, Tamás Zelczer, on what makes AI-driven forensic technology truly usable in practice. And Biometric Update reported that our computer vision-based gait dynamics + body-structure analysis methodology has been […]
How Accurate Is Walking Recognition? Uncovering the Precision of Our Technology
What we learned from over 90 bodies in motion, thousands of gait cycles, and near-perfect recognition rates At Cursor Insight, we’re passionate about understanding how people move—because movement, it turns out, is one of the most personal things about us. Your gait, the way you walk, is as unique as your fingerprint. And just like […]
Feature screening for ultra-high dimensional, multi-class data
Predictive modelling has become a widely used technique for businesses to gain insights from their data. It is often the case that tens (or even hundreds) of thousands of features are present in the data in question, which can be problematic when training a machine learning model. Feature screening is the process of selecting a small subset of […]
