The Data Minimalists Were Right All Along
Verifying a fact about a person and possessing their record have been treated as the same problem for as long as government has run computerized data matching. If an agency wanted to confirm that an applicant’s income fell below a certain threshold, that a Social Security number belonged to a living person, or that someone wasn’t collecting the same benefit twice under two names, the working assumption was that someone had to hold the underlying data to check it. That assumption served as the foundation for nearly every fraud-prevention proposal I have worked on, including my own. But it’s not true anymore.
Two architectures now let an agency get an answer without taking custody of the record behind it. The first is federation, where each agency keeps its own data and simply responds to a standardized query from another agency, the way Estonia’s X-Road has worked since 2001. No agency’s data gets pooled into a shared repository. The tax authority holds tax data, the population registry holds residency data, and the query travels to the source and returns an answer, encrypted and logged the entire way. X-Road has run for more than two decades, weathered repeated attacks from Russian-linked actors, and has no known major security breach to its name. More than twenty countries have adopted versions of it since, with Finland federating directly with Estonia’s system in 2018 under joint stewardship through the Nordic Institute for Interoperability Solutions.
The second architecture uses privacy-enhancing technologies, or PETs, a category that includes secure multiparty computation, verifiable credentials, and zero-knowledge proofs. Federation controls where data lives, PETs control what any party learns during the exchange, including the agency running the check. A zero-knowledge proof lets someone demonstrate a fact is true, such as income falling under a threshold, without revealing the income figure itself. Secure multiparty computation lets two agencies jointly compute an answer, such as whether the same person is drawing a duplicate benefit in both of their systems, without either side seeing the other's underlying records. PETs are not a recent invention. The cryptographic groundwork behind PETs traces back to the 1970s, and the term itself was coined in 1995, but it has been largely dormant as an operational solution to government problems until very recently.
In the last five years, governments have, for the first time, begun funding PETs as infrastructure. The US and UK announced the PETs Prize Challenges at the first Summit for Democracy in December 2021. They ran the competition through early 2023 with NIST and NSF administering the American side, and showcased the winning approaches at the second Summit for Democracy that year. NIST’s Privacy Engineering Program is currently running a working PETs testbed, built in part through the National Cybersecurity Center of Excellence, that develops what NIST calls model problems. Model problems contain a defined use case, a synthetic benchmark dataset, a reference implementation of the relevant PET, and a way to measure how well it performs. NIST has already built and run PETs model problems for genomic data analysis across research silos and de-identification of Census-style demographic data. The European Union’s eIDAS 2.0 regulation requires every member state to offer citizens a digital identity wallet built on selective disclosure by the end of 2026. Selective disclosure means someone can prove they meet an age or income threshold without handing over the underlying document. The National Association of State Workforce Agencies’ (NASWA) Integrity Data Hub already lets state unemployment programs check a claim against other states’ fraud indicators without any state turning over its full caseload.
The recent expansion of PETs is good news for the fraud prevention arena. The fact is that the public simply doesn’t trust the government with its data. Public concern about how government handles personal data has been rising for years. Pew Research Center puts it at 71 percent of adults in its most recent survey, up from 64 percent in 2019, with the increase concentrated among people who used to worry about it least. DOGE almost certainly had a hand in that increase. In February 2025, reporting revealed that individuals affiliated with DOGE had gained access to sensitive databases at Treasury’s Bureau of the Fiscal Service, the Office of Personnel Management, the IRS, and the Social Security Administration. DOGE staff frequently framed the justification as necessary to detect fraud and improper payments. Courts disagreed with how that access had been obtained and by March 2025, federal judges had blocked DOGE from Treasury, OPM, Education, and SSA systems on Privacy Act grounds, with one ruling warning that sensitive information had likely already moved outside the agency responsible for protecting it. DOGE became the case study for why the public won’t extend government the benefit of the doubt on data access, regardless of the justification attached to it. That distrust is rational, and any reform that still asks Americans to trust a new or expanded government repository is fighting an uphill battle against a fact pattern the public just watched unfold in real time.
Fraud remains and serious and growing problem in government benefits administration. The culprits today are more likely to be adversarial nation state actors and organized criminal networks rather than individuals lying on their applications. But the mechanism I and others have defaulted to for addressing it— a bigger and more capable central database— is no longer fit for purpose or politically viable. A federated query can confirm an identity is real and not already in use elsewhere without anyone ever seeing the full record behind it, and a cryptographic proof can confirm a business has an actual owner without that owner’s information sitting in a database that attracts thieves. Cryptographic techniques that can validate facts using minimal disclosure hold enormous promise in addressing the fraud prevention-privacy preservation tension we find ourselves in today. It’s time for Congress and the White House to invest in researching the use of these technologies in the fraud and improper payments arena.
Photo by Sasun Bughdaryan on Unsplash. Article first posted on GovIntegrity.