FiVerity

The removal test leaves the rules based systems the company was founded to beat, and the Federal Reserve's own finding that those miss up to 85 percent of synthetic identity fraud is the measure of what would be lost. Machine learning performs the detection, pattern matching enables privacy preserving comparison across institutions, and risk scoring and fraud classification run automatically. A further intelligence layer shipped in March 2026. Identifying a fabricated identity that has no prior fraud history is inference by definition.

The division is stated as the operating principle, with the company describing its approach as combining machine learning with the expertise of each institution's own fraud analysts rather than replacing them. Outputs are proactive alerts, risk scores and classifications that analysts act on, and where the platform touches regulatory filing it pre-fills reports rather than submitting them, leaving the institution to review and sign what carries its name.

Joint investigations are conducted by institutions with the platform connecting them. What is not described is any confidence threshold or what an analyst is expected to verify before acting on a network alert.

The performance claim is framed against the named alternative rather than in isolation, stating detection of more than half the synthetic identity fraud traditional models miss, with the Federal Reserve's finding that rules based systems miss up to 85 percent supplying the baseline. That is the right construction, since it tells a buyer what improvement to expect over what they run today. Risk scoring is described as transparent and classification as automated.

What is absent is the other half of the measurement: no false positive rate, no precision figure and no validation result, which matter more than usual because network alerts affect institutions that never saw the original evidence.

Backing is unusually well matched to the product: 14.5 million dollars across four rounds from investors including a bank on the cap table and a firm specialising in the intersection of financial technology and traditional banks, whose partner, a former New York Federal Reserve regulator, sits on the board. A national cyber forensics organisation acts as research partner, and the company states government regulators have recognised its approach.

One credit union is named publicly and adoption is otherwise described as regional and national banking communities. The performance claim is specific, catching more than half of previously undetected fraud. Headcount is 30 and the company has operated since 2017.

This is the most complete answer to the consortium question in this index. Other vendors here disclose that cross institution learning occurs, which is honest as far as it goes and leaves open what actually crosses the boundary. FiVerity specifies the mechanism: pattern matching in place of personal information, double blind encryption on shared fraud identities, and anonymised exchange, so an institution contributes intelligence without contributing its customer records. The Federal Reserve's position that no single organisation can stop this fraud alone is the argument for pooling, and this is the design that makes pooling defensible rather than merely useful.

Privacy is the architecture rather than a policy statement, which is what earns this grade. Institutions exchange pattern matches rather than consumer personal information, the identities of detected fraudsters are shared under double blind encryption, and the company describes the exchange as anonymised throughout, so participants gain intelligence about fraud without gaining access to each other's customers.

That is a design decision that constrains what the platform itself can hold, and it is the correct answer to the central objection to any fraud consortium, namely that pooling data means exposing customers who did nothing wrong.

No attestation, certification, trust centre or enumerated framework was located. Cryptographic design is described in some detail and is credited elsewhere in this assessment, but a described mechanism is not an assessed control set, and institutions joining a network that carries intelligence between competitors would be expected to ask for one.

This company names the statutory provision that makes its entire business model lawful, which almost nothing else in this index does. Joint investigations run under the information sharing safe harbour created by anti terrorism legislation, the provision that permits financial institutions to exchange information about suspected money laundering and terrorist financing without breaching customer confidentiality.

Suspicious activity report pre-filling addresses a second defined regulatory artefact with statutory content requirements. Beyond that, the Federal Reserve's published position on collaborative defence is cited as the basis for the approach, government regulators are stated to have recognised the method, and a former Federal Reserve regulator sits on the board.

The privacy design limits exposure and does not remove the central risk, which is misidentification propagating across a network. Synthetic identity detection works by finding records that look fabricated, and the signals it relies on, thin history, inconsistent data and sparse corroboration, describe real people as well as invented ones: recent immigrants, young adults with no credit history, people rebuilding after disruption, and anyone whose records are genuinely incomplete.

A flag raised at one institution reaches every participant, which is what makes the network valuable and also what makes an error expensive. No false positive rate, testing across population groups or review process was located.

No guarantee, indemnity or correction process was located. Participating institutions benefit from alerts and from shared investigation. The individual has no described position at all, and the network structure makes theirs the weakest in this index: someone wrongly identified as a synthetic identity is flagged across many institutions at once, will not be told, cannot know which institution originated the pattern, and has no stated route to have a correction propagate back through the network that carried the original flag.

No model provider, hosting arrangement or subprocessor list was located. One upstream relationship is named, a national cyber forensics organisation acting as research partner, which indicates where threat intelligence expertise comes from. The material dependency is the network itself, since detection quality scales with how many institutions contribute, and no participant count is published, so a prospective member cannot judge what they would be joining.

The deployment claim is the strongest practical feature for the segment served, with the solution described as embedded and requiring no integration, alongside a secure programmable interface for those who want one, and the platform is positioned to connect the fraud systems, data and teams an institution already runs rather than replacing them. For community banks and credit unions with no engineering capacity that removes the usual barrier entirely. No named core banking, fraud or case management system appears, and no developer documentation was located.

No hosting provider, region selection, residency commitment or private deployment option was located. Exposure is domestic and the encryption architecture means the most sensitive material never leaves the contributing institution in identifiable form, which narrows the question considerably without answering where the network layer itself operates.

No pricing, packaging or basis of charge was located. Adoption cost is addressed directly, with the solution described as embedded and requiring no integration, which for smaller institutions with no engineering capacity is often the binding constraint rather than licence cost. Network products also raise a pricing question nothing answers: whether contributors of intelligence pay the same as consumers of it.

Buyers span banks, credit unions, payment institutions and online lenders, with explicit focus on community banks, regional banks and newer lenders, and the network extends beyond buyers to regulators and law enforcement agencies who both contribute and receive. Trade coverage notes smaller institutions in particular turning to the platform, which matters because they are the ones least able to see fraud patterns from their own data alone. Scope is domestic and confined to fraud and financial crime rather than the wider risk stack.