Unfraud

Unfraud

Deep Learning

Founded: 2014
HQ: Rome / Italy

We are a sophisticated cutting-edge solution for online fraud prevention and we believe that prevention is the most effective way to secure the growth of online business.  We use Artificial Intelligence to recognise fraudster’s activity in real time, protecting the business from any fraud-related cost such as chargeback or bad reputation. There is no need of change anything in our customers’ businesses, allowing for an easy, flexible and fast deployment.

Process Steps

  • Risk Engine
  • Data Provider
  • Manual Review Cockpit
  • Chargeback Handling
  • Callcenter Fraud
  • KYC/AML/ID Verification

Technology Coverage

  • Machine Learning
  • Device Fingerprinting
  • Behavioral Analytics
  • Chargeback Guarantee

Reference Customer

Telecom Italia

Terna

Volagratis

lastminute.com

Rumbo

Integration

Magento

Shopify

Fraud Solution Profile

We are mainly active in the transaction security field, and this represents our core business.  We provide a complete cybersecurity platform, that enables our client to have a clear picture of how the business is going and how frauds are related to it.  Our platform comprehends different modules, each one for a specific use: among the most popular there is the Deep/Dark web recognition tool.  We also offer a Dashboard from which it is easy to control different statistics about the business, see how UnFraud is running, the main actions taken, which transaction has been blocked and give us feedbacks.

Thanks to our flexible AI engine, we are able to cover issues less related to former frauds.  An example of the application of our AI, is for scam and lie detection.  These issues are mainly typical for marketplaces (recognize fake product reviews for example) or classifieds websites (e.g. Craigslist).

 

For another customer, we employed our AI based system to analyse the video stream and distinguish between real physical intrusions and false positives. This allowed our customer to free up workforce for other more important tasks, as our systems automatize almost the whole process of surveillance monitoring.  This is why we say that our system and our AI engine is really flexible.  Our AI expertise allowed us to create and offer general purpose AI systems, that are immediately ready for any kind of data.  A really not common feature, we would say a really rare one, is that we entirely rely on deep learning.  Generally all systems based on deep learning are heavily tailored on a single specific problem, while our experience and skills allowed us to go further and overstep this problem.

 

We are able to accept numeric data, categorical, Boolean, and even text data.  These are automatically converted in a deep learning friendly format, and thus we can nonchalantly go from fraud detection on financial transaction (mostly numerical data) to scam/lie detection (mostly natural language data) to behavioural anomaly detection (mostly behaviour and device fingerprinting – i.e. non-human readable – data) immediately and within the same exact system without any tuning.

 

This is possible since our experience in AI begins in bioinformatics and computational biology, building up systems for tasks like:

  • discernment between normal and cancerous cells
  • discernment between populations of cells with different mutant proteins
  • discernment between tissue at different developing stages in embryos

 

Those data ranged from strictly numerical to computer vision approaches; this set the stage for our future growth, and gave us the experience and know-how to treat different kind of data within the same area of interest. Our growth path has been huge since the beginning.  We built an AI know-how starting from simple machine learning algorithms to neural networks and eventually to the cutting edge approach in academic literature, deep learning.  With this know-how it was straightforward to go towards deep learning based AI systems that were able to tune themselves on any new problem they are presented with.

 

Share this page: