Fraud consortium

A group of organizations that pool fraud information so each one can benefit from what the others have seen.

If one member confirms that an identity was used for fraud, the others can check against that knowledge before they accept the same identity.

A shared arrangement in which participating organizations contribute fraud-related data, such as flagged identifiers or confirmed cases, to a common resource that members can query. A member is an organization that contributes to and uses the shared resource.

A consortium typically works like this:

1. Members agree on rules for what data they contribute and how it can be used.
2. Each member sends records or flags about confirmed fraud to the shared resource.
3. When a member sees a new case, it queries the shared resource.
4. The query returns whether the identity or detail appears in what others have contributed.

Consortia come in several forms:

• Shared blocklists, which hold identifiers tied to confirmed fraud.
• Industry fraud databases, which collect confirmed cases from members in a sector.
• Data cooperatives, which pool a wider set of identity and behavior data.

They share a common design trait. In the classic consortium model, records move to a central place, which creates a pool of sensitive data that needs protection.

That design raises several concerns:

• Privacy and legal limits, since moving customer data between organizations can conflict with privacy rules.
• Concentration risk, since a central store is an attractive target.
• Data quality, since one member's mistake can affect others.
• Governance, since members need clear rules for correcting and removing entries.

Privacy-preserving networks take a different approach. Raw data stays inside each institution, and only a limited, de-identified signal returns.

A fraud consortium sits next to terms that are easy to mix up:

• Negative list / blacklist is a list of flagged details, which a consortium can share among members.
• Network-based fraud detection is the broader approach, and consortia are one way to build it.
• TrustGraph keeps raw data in place and returns a de-identified signal, so it does not work like a central data pool.