Home » Applications » ML21.Fraud


We find all suspicious transactions in your data

Financial institutions are charged with the task of working against fraud, money laundering and other undesirable activities. Fraud is often characterized in data by subtle deviations in actions.
ML21 is particularly suitable for detecting suspicious transactions. This is largely because our machine learning algorithms have the ability to extract patterns from seemingly unrelated data and help companies better detect fraud. Machine learning becomes even more pervasive with larger data sets, allowing companies to make informed decisions in real time and get clear advice on the risk profile of a given transaction.

Trustworthy IA

USPs of our ML for fraud prevention:

1. Realtime analysis;
2. Clearbox. Our ML explains to you why a transaction is “suspicious.”
3. Rapid implementation;
4. Feedback loop for fast learning curve;

Integration without code

CodeNext21’s anti-fraud solution is a piece of cake to implement. We offer customized implementation options to fit each case. Choose from a variety of open-source implementation options without code, or if you insist, we provide a variety of SDKs so your developers can dig in!

GDPR compliant

When creating risk profiles of users, personal information may be relevant. As a result, you will have to deal with regulatory compliance. CodeNext21 help you process the information correctly so that you remain AVG compliant.

Fraud checks
Damages from fraud
Cost savings from fraud

In a telecom case, we integrated our fraud prevention solution for a large wholesale telecom provider with more than 35 million calls per month. CodeNext21’s real-time monitoring system detects fraud (on average) 17 hours faster. In the most extreme case, we outperformed the existing solution by 34 days. Overall, 80% more fraud was prevented, saving our client tens of thousands of dollars per month.


Business case : Fraud in telecom transactions
In an application of our technology in the telecom domain (SIP), we were able to outperform an existing anti-fraud solution by 80%. A savings of tens of thousands of dollars per month for our client!


ML21 provides clear explanations of each fraud assessment. Supported by graphics, you get unambiguous justification for the outcome: Why exactly is the risk profile of this transaction high or not? So you remain in control of the results.

Superior speed

Compared to the existing anti-fraud solution, CodeNext21 detected fraud (on average) 17 hours earlier. In the most extreme case, fraud was detected by CodeNext21 32 days before the existing system was activated.


Our platform ML21 is capable of processing transactions into a risk score in near real-time. Our scalable solution allows millions of transactions per second to be assessed with a risk score. You decide from what risk score, on a scale of 0 to 100, you wish to block a transaction.


Every user is different, which is why CodeNext21’s platform learns from each of your users and adjusts its algorithm accordingly. This is why we produce 100x fewer false alerts than existing anti-fraud solutions.

Less downtime

Many billshock and anti-fraud systems block entire users when fraud is detected, causing downtime even for legitimate calls. Because CodeNext21 inspects individual calls, we are able to stop fraud but leave legitimate traffic untouched. This reduces customer downtime by more than 11x.

Legimate calls blocked

Choose the right path for the future

Brainstorming about applying our technology to your data? Curious about how your organization can handle data better? Contact us.