How does Adyen's latest risk engine work?

Adyen’s latest risk engine is designed to detect, stop, and fight fraud to achieve maximum conversion at a minimum cost. To find this sweet spot, we use machine learning models trained on Adyen’s platform-wide data to sharpen and automate risk decision-making.

Machine Learning

With the Premium version of the risk engine, there are two machine learning models responsible for identifying and blocking suspicious transactions:

• Bot attack risk model: detects the likelihood that a payment is part of a card testing attack; available in both premium and basic risk engines.
• Fraud risk model: detects the likelihood that a payment is fraudulent; available in the premium risk engine.

The models screen transactions pre-authorization using the information provided in the payment request and historical payment data in the Adyen Ecosystem. Transactions are blocked based on the probability that they are fraudulent. 

Risk rules

In addition to machine learning models, risk engine risk profiles include risk rules. Risk rules can be assigned the following actions:

• Allow: When an allow rule is triggered, the risk engine does not block the transaction.
• Block: When a block rule is triggered, the risk engine blocks the transaction.
• Review (Premium): When a review rule is triggered, the transaction is sent to Case Management (Premium feature).
• Check for 3DS: when linked to a Dynamic 3DS rule, the transaction is sent to Dynamic 3DS.