Code and continuous integrations are essential to fraud prevention and business intelligence
We believe a solution to all your fraud-related problems lies in code and its constant development. All our Machine Learning models are based on comprehensive data gathered by our profiler, provided by our clients and relevant third-party data.
Of course, it’s not a finished process (as fraudsters never sleep) and we obsessively seek out improvements because we believe that fraud prevention can only be successful with:
How does our anti-fraud solution work?
The Nethone’s system works towards profiling the users on the merchant site, detecting fraudulent behaviour and providing business intelligence insights. We collect all data out there starting from merchant’s databases through user’s interactions within their website.
The said data needs to be sent to our proprietary Machine Learning algorithms, tailor-made by our data scientists. It is delivered through the Inquiry API, Events API, Transactions API and Profiling Solution.
Panel - the way we visualize all of these data
Our panel also has a state of the art case manager for your fraud team. They can use it to automate their day-to-day activities to gain more time and information to tackle the most challenging cases that fall into manual reviews.
Aggregated traffic characteristics in a specified time frame
Let’s say you want to determine the traffic on your website in the context of fraud and payments.
The panel is the start of a comprehensive view of user profiles, visualization of data, and tracking of ML decisions based on signals and connections.
Inquiries with the context
User’s activity within the platform
In the Inquiries view, you can see all the user’s activity within one session gathered in a consecutive order on a timeline which we refer to as inquiries. e.g. creating an account or adding product to the cart. On top of that, in this view we provide a summary of the signals and algorithms which were triggered by each inquiry. Signals, in short, are notifications about abnormal situations which may be indicators of a higher probability of fraud. For example, you can see that clipboard was used on sensitive fields or incognito mode was detected. There are more than 100 of them and the list is growing.
Additional data worth noting
List of user’s actions that trigger Nethone’s recommendation
Different symbols represent different inquiries within the same activity session of the user. For example, you can check what card was used and find out that it was different for both inquiries.
No “black box” solution, only explainable AI
Together with the recommendation ‑ accept ‑ refuse ‑ review, we deliver a human‑readable explanation of the ML results in one place. We present features that have the most significant impact on recommendation (and contribute both, to the probability of fraud and contrary).
Interactive display of user’s activities and geolocalisation
The map view gives you in a quick glance all the geographic information you need about a given transaction attempt. Where did they buy from? What route did they buy? Where is the card from?
Analyze the web of relations in your merchant ecosystem
We are particularly proud of what we call the connection graph which shows the links between different profilings. All the nodes presented on the graph are related to each other by some feature. As you can notice these two profilings are connected by the same email address, user, IP address and cookie. This will put single profilings in context for you. You can check if the other profiling was detected as fraudulent or not and see how complex the whole network is.
Preference selection for a specific client
You can customise your view by selecting which features you want to track and arrange them according to your personal preference. You may choose from quite a few widgets, including i.e. signals summary, card data or related profilings. We add new ones all the time!