NSECURED MemberGo / Financial services

Let fraud models analyze transactions within a defined purpose.

Fraud analysts need transaction features to identify anomalies. Financial institutions can register the fields a risk workload needs and distinguish fraud analysis from marketing through service identities and purpose controls.

An analyst at transaction monitoring screens

Financial services

Which data does this workload need?

Fields to protect

Customer transactions, sensitive financial fields and customer identifiers. Transaction features and identifying details can be governed separately.

Fields eligible for this purpose

An approved fraud workload may receive custom fields such as transaction amount band, frequency and merchant category.

Synthetic data example

Two access decisions in context

Access allowed

Purpose
fraud_detection
Fields
txn_frequency amount_band
Result
200 · access_id: demo_7c20

In a transaction-monitoring pilot, the approved fraud workload can read frequency and amount band to score a synthetic transaction.

Request denied

Purpose
marketing ×
Fields
txn_frequency amount_band
Result
403 data_access_denied

A marketing workload requesting the same transaction fields for product recommendations is denied when only fraud analysis has been authorized.

The fields and decisions here are illustrative. Actual access depends on tenant settings, the basis for processing and the approved purpose.

Workflow

How the workflow fits together

  1. 01

    Choose the fraud features

    The risk team lists the model inputs and identifies the source, purpose and owner of each field.

  2. 02

    Separate service identities

    Set up separate test identities for fraud and marketing so their purposes and permissions can be evaluated independently.

  3. 03

    Review access and records

    Run both purposes against one synthetic transaction. Verify which features were released and check that audit records omit their values.

Proof of concept

Test two purposes against one synthetic transaction

Ask risk and security teams to choose three transaction features. Submit fraud and marketing requests, then compare the decisions, stop behavior and internal trace records.