AI Machine Learning for
Fraud Detection Solution

"SEON's machine learning finds similarities based on your input and proposes rules to spot abuses you wouldn't notice otherwise."

G2 4.2 rating

Pre-built, Customizable, and Suggested Rules & Risk Scores

Fraudsters have behavioral patterns. Lean on our machine learning engine to create complex rules to beat them, without ever giving up control.

WhiteboxBlackboxLabels

Bolster your Fraud Detection

With precise and dynamic rules based on a variety of data and methods

Bolster your Fraud DetectionBolster your Fraud Detection

See How SEON Stops Fraud

Stop All Fraud-Related Threats

reviewer profile
reviewer company logo
"As soon as our staff used SEON on a regular basis, they could see its value. It immediately started saving them time as they didn’t have to trawl through the usual checkpoints. So much so, that the time spent by Air France analysts on manual reviews is down by an impressive 70%. Instead, all the information they needed was on one screen with a simple scoring system enabling them to see if a transaction was real or not."
reviewer profile
Eric Facquet
Deputy Manager of Fraud Prevention
reviewer profile
reviewer company logo
"The most useful thing for us was to be able to see if the email and phone numbers weren’t registered anywhere. We also rely on the device fingerprinting metadata in our scorecard to filter out junk users. We immediately saved about 6% of costs on our automated KYC checks. After only 2 months, we started seeing impressive results. Up to 3 times fewer cases of defaulting customers in new regions such as Romania and Europe. 4 months later, the fraud rates dropped by over 65%."
reviewer profile
Sergey Bogdanov
Chief Risk Officer
reviewer profile
reviewer company logo
"We were looking at another provider, but they came back to us with a 4-6 months integration window. With SEON, it was literally a phone call, sandbox tests on Monday and by the end of the week, it was done. One other thing I’d like to comment on is the pricing model. It’s really 2022, like for scaling together. The SEON model, it’s now one we actively look for with our other partners. Clearly an organization that approaches work and its customers in a different way."
reviewer profile
Coert Snyman
Senior Analyst
reviewer profile
reviewer company logo
"I tested some extra rules, and over a few months I managed to push SEON’s auto approval from an initial accuracy of 95% unto 99.5%. Taking the time to update your SEON rules is an investment that gives great results. And the best thing is that we can do so without deploying any software. From an engineering perspective, it is extremely time and resource saving, allowing us to react much faster."
reviewer profile
Phillipp Keller
Senior Product Manager
reviewer profile
reviewer company logo
"I couldn’t believe how fast the results came,” says Gergely. “Our chargeback rates dropped by 91% in one month only.” Moreover, the number of manual queries and fraud catch rates continued to improve after the initial 30 days as the system adjusted itself – part of the benefits of deploying a machine-learning algorithm to fine-tune accuracy."
reviewer profile
Gergely Kálmán
CEO
reviewer profile
reviewer company logo
"SEON allows us to stay one step ahead as by keeping advantage players at bay we are able to continue offering competitive bonuses with low wagering requirements to our players,” Daniel says. “And the fact that SEON team are so open to feedback and implement the feedback quickly gives us as their client a significant advantage which we couldn’t find with other competitors in their field."
reviewer profile
Daniel Saliba
Head of Compliance
reviewer profile
reviewer company logo
"It wasn’t just the technology that’s great though, the education side of things helped us understand the types of tactics and patterns used by fraudsters and put rules into place to prevent them."
reviewer profile
Stacey Pickering
Head of Operations
reviewer profile
reviewer company logo
"We deployed SEON both on our website and web app to gather device data as well as IP and email information. It was unbelievably fast, both in terms of integration and results. In 4 days only, our fraud rates were pretty much down to 0."
reviewer profile
Lewis Bye
Senior Projects and Operations Specialist
reviewer profile
reviewer company logo
"We overlay SEON’s digital and social lookup with device fingerprinting to get a good indication of whether the phone or tablet belongs to the actual client. We can instantly filter out obvious fraudsters with no digital or social presence, and we get better intelligence to automate our decision making."
reviewer profile
Juris Rieksts-Riekstinš
Head of Risk

AI & Machine
Learning

SEON includes powerful machine learning designed to help you identify risk patterns automatically. Based on flagged signups, logins, and transactions, it can recommend new rules to help spot patterns of user behavior. You can then calculate the accuracy of these rules and make decisions automatically according to confidence ratings.

Our whitebox rule suggestion algorithm is built on C5.0 and works best after processing at least 1,000 transactions. You can then receive suggestions for complex or heuristic rules to thwart even the most sophisticated fraudsters. Meanwhile, a separate blackbox ML module at SEON will examine each user’s activity and flag suspicious customers separately to the main engine, for added peace of mind.

The whitebox module provides human-readable reasoning to aid your analysts. The blackbox module uses CatBoost to calculate risk via a plug-and-play model that requires little interaction. As a SEON customer, you can choose to make use of both modules, just one or neither. The goal is always the same: to improve the accuracy of your fraud detection over time – based on your real and unique business landscape.

FAQ

Is SEON’s machine learning algorithm blackbox or whitebox?

SEON offers both whitebox and blackbox solutions. You can start straight out of the box with the blackbox machine learning tool, which will tell you when a user is demonstrating suspicious behavior. This is an extra check that does not interfere with your fully readable ruleset-based fraud scoring. Moreover, you can also make use of our separate whitebox solution that delivers human-readable and testable rules that will work specifically for you.

How long does it take for SEON’s whitebox machine learning to suggest rules?

Our whitebox machine learning engine delivers the best results after it’s been fed at least 1000 declined transactions (or user actions), ideally with labels

What algorithm is SEON’s machine learning model built on?

SEON’s whitebox machine learning algorithm, which suggests new rules, is built on C5.0. The blackbox module, which provides blackbox risk scoring, uses CatBoost.

Where can I find more information on using SEON Machine Learning Model?

You can learn more about how SEON’s SEON Machine Learning Model in our Knowledge Base. Here is an article too about fraud detection in machine learning too.

Read More About
AI & Machine Learning

Start fighting fraud now

Ready to see your fraud rates drop by the hour?
Get in touch for your free trial or to see how SEON can help your business grow safely today.

Machine creating SEON products on a conveyor belt.