← All topics

Focus area

Privacy-preserving identity

Resolving who someone is without capturing what they are.


Correlating offline and online presence through intermediate identifiers, so that useful inference can happen without the personally-identifying payload. The question is not whether you can identify someone — it is whether you should have to.

The received wisdom of the last twenty years of digital identity is that correlation requires disclosure. To stitch two records together you needed a common key, and the common keys that worked — email addresses, phone numbers, device graphs — were exactly the fields you shouldn't have been collecting in the first place.

The alternative has existed in academic cryptography since the 1990s. Private set intersection, zero-knowledge proofs, trusted execution environments, federated learning — the primitives are now fast enough to deploy in production systems. What's missing isn't the math. It's the engineering discipline to pick them up, and the commercial incentive to ship them instead of cleartext pipelines that are easier to build and cheaper to operate.

The 2014 patent mechanism — chaining across intermediate identifiers so neither endpoint needs to co-observe — is the architectural template. The 2026 refinement is that every intermediate can now be a cryptographic artifact instead of cleartext. Same graph. Smaller blast radius when something leaks.

Where it shows up

Three applications.

  • 01

    Healthcare record reconciliation

    Stitch the same patient across hospital, pharmacy, insurer, and wearable without any party seeing the others' raw identifiers.

  • 02

    Sybil resistance for benefits and voting

    Detect one person running many wallets or many absentee ballots, without unmasking legitimate participants.

  • 03

    Advertiser-publisher clean rooms

    Discover shared audiences without exporting either side's customer file — the original motivating context, now with better primitives.

Related focus areas