Fraud Response
AnantVerify tests whether the decision is sensitive to what matters—and insensitive to what does not.
The situation
A fraud system identifies an unusual transaction and recommends freezing an account, supported by transaction velocity, device changes, geography and unusual behavior. But freezing a legitimate customer's account can itself have significant consequences.
The decision-assurance problem
A traditional evaluation might ask whether the fraud model correctly classified the transaction. AnantVerify asks additional questions: was the state current, which evidence supported the action, did another trusted signal contradict it, would an irrelevant change alter the recommendation, how close is the case to a decision boundary, and what happens when multiple risk signals occur together?
How AnantVerify evaluates it
DTT can systematically transform the transaction state:
- Change the transaction description while preserving the financial behavior — the decision should remain invariant.
- Change the beneficiary to a newly created account — the decision may be expected to become more restrictive.
- Change the device and beneficiary together — a different escalation requirement may apply.
- Remove a stale risk signal — the decision may need to change again.
Business value
The result is not simply another fraud score. It is an assurance determination supported by the state, evidence, governing controls and tested decision behavior.
AnantVerify tests whether the decision is sensitive to what matters—and insensitive to what does not.