Skip to content

vfps

vfps

A very fast and resource-efficient pseudonymization service, with gRPC, REST and FHIR APIs, an admin UI, and highly available deployments.

Get started View on GitHub

~12,000 req/spseudonym creation on a single instance
< 10 msP99 request latency
< 2 sfrom cold start to first pseudonym
~1,000 req/swith just 1 CPU and 128 MiB memory

How these numbers were measured

  • Namespaces, one level or many


    Each namespace has its own pseudonym format and length. Build multi-level pseudonymization by chaining child namespaces off a parent, and hand out several pseudonyms per original value where a study needs them.

    Namespaces

  • Oblivious, verifiable pseudonyms


    Derive deterministic pseudonyms with an RFC 9497 VOPRF. The key holder never sees the value it pseudonymizes.

    VOPRF pseudonyms

  • gRPC, REST and FHIR


    Every operation is available over gRPC and JSON-transcoded REST, described by an OpenAPI specification. FHIR clients can use the MII pseudonymization operations.

    MII FHIR pseudonymization operations

  • Admin UI and CSV jobs


    Browse and create namespaces, then pseudonymize, de-pseudonymize, import or export larger-than-memory CSV files as background jobs that stream straight to and from S3-compatible storage.

    Admin UI

  • Namespace-scoped access control


    OIDC sign-in, separate read, write and reverse-lookup grants per namespace. Vfps-issued personal access tokens and service accounts with fine-grained access for machine users.

    Access control

  • Built for production


    Stateless replicas on PostgreSQL, a hardened Helm chart, and OpenTelemetry metrics and traces.

    Production deployment