Compute on encrypted data.
Wodan builds the infrastructure for data to remain mathematically encrypted throughout. The model never sees plaintext. Neither does the infrastructure it runs on. Neither do we.
Only the key holder does. No key, no data.
What is FHE?
Run external AI on your data while it stays encrypted, so the provider never sees a single cleartext moment.
Under DORA Article 28, the inference step stops being a third-party risk event, because there is no exposure for a third party to create.
The provider becomes a processor that mathematically cannot reach your data.
What is third-layer encryption?
Data has long been protected in two places: at rest, when it is stored, and in transit, when it moves.
The third place was always the open one, because data had to be decrypted the moment it was used. Wodan closes that gap.
Encryption-in-use lets data be processed while it stays encrypted, so it is protected at rest, in transit, and in use.
What does FHE make possible?
Working with sensitive data has meant choosing between two losses.
Anonymise it and you strip out the detail the model needs. Leave the detail in, and the data sits exposed the moment it is processed.
Wodan removes the need to choose: data stays encrypted the entire time, including during processing.
Why is it next-generation?
Twenty years ago, the padlock, the SSL certificate, in your browser meant nothing to most people.
Today every banking app is expected to show one.
Encryption-in-use sits at that same point: unfamiliar today, expected tomorrow, and in time simply what “secure” will come to mean.
How does the Wodan technology work?
1
Requester (application) sends a clear text request to Wodan’s client. Client runs in the data owner’s infrastructure (on-prem or cloud).
2
Client encrypts the data before sending it to the server.
3
Private keys are stored in the client.
4
Send encrypted data to the server along with “evaluation keys” to allow encrypted computation.
5
Server returns encrypted answers to the client.
6
Client decrypts the encrypted answers and provides it back to the requester in clear text.
Integration Model
Drops into existing AI/ML pipeline
No re-engineering. Point your application to the client API.
Deployment Model
On-prem · Cloud · Hybrid · Edge
Docker containers, all major cloud platforms.