Realistic data for your teams. Real data stays behind the boundary.
The Oryvin obfuscation engine lets people and AI assistants work against de-identified data that behaves like the real thing, without ever handling the real thing.
A protected boundary between real data and everyone else
The engine connects to a source, applies the protection scheme defined for every field, and serves only the protected result.
Real data source
Scheduling platforms, document databases, uploaded files and other registered sources
Obfuscation engine
Applies the protection scheme field by field and keeps relationships intact
Protected output
De-identified, consistent data for developers, testers and assistants
Protected, and still useful
Relationships preserved
The same real person becomes the same fictional person across every record and every session, so joins, histories and edge cases still behave like production.
A scheme for every field
Each field is classified once: remove it, replace it with a consistent surrogate, or pass it through. Anything not listed is dropped.
Time-limited by design
Working copies are encrypted, held for the job and wiped when it completes. Surrogates can be set to expire on a schedule.
Named access
Every project has its own identity. Access to the engine is gated by a terms-of-use agreement that must be accepted for each version.
Operators in control
An operator plane lets administrators inspect and stop individual jobs without interrupting everyone else.
Many sources, one method
Connectors cover scheduling platforms, cloud databases and uploaded files, all protected the same way.
A portal for protecting a file in minutes.
Upload a data file in the Cane portal, mark how each field should be treated, and download the de-identified copy. Create a reusable source from the same upload when the data will be needed again.
- Every field starts as removed until you say otherwise
- Uploads are encrypted with their own key
- Administrators approve who can use which sources
The boundary in practice
Nobody outside the boundary, human or AI, has to be trusted with real values. Not the developer. Not the assistant. Not the support tool.
That is what lets teams test against realistic data, and lets AI help with the work, without turning every project into a protected-information project.
An AI agent that works inside the boundary.
Cane is a hardened coding and analysis agent for the private lane. It runs without outbound telemetry, in a sandbox, with explicit per-tool approval, against local models. Tasks that touch sensitive data stay where the data is.
- No hidden network calls: only what the task itself needs
- Per-tool approval policy instead of a single global switch
- Results returned as patches and reports a person can review
- Pairs with a second, independent reviewer for ordinary work
Give your teams realistic data without the exposure.
We will show the engine protecting a sample of your own data shape, end to end.
Request a briefing