Method
Test set and document types
We will add the test set and the document types as soon as the measurement is released.
What counts as an error is fixed before the measurement. We distinguish four kinds:
| Error type | What it means |
|---|---|
| Whole value visible | A name, identifier or credential appears in full in the shadow copy. |
| Part visible | Only a part of a value is still in the shadow copy. |
| Wrongly replaced | Something that is not a secret was replaced, or replaced with a placeholder that changes the meaning. |
| Translation back faulty | Out and back does not produce exactly your text. |
Results by language
So far only German is measured. For the other languages there is no measurement yet.
| Language | Status | Result |
|---|---|---|
| German | measured | values to follow |
| English | not measured | no result yet |
| Dutch, Danish, Swedish | not measured | no result yet |
| Spanish, Polish | not measured | no result yet |
| French, Italian | not measured | no result yet |
Roadmap for more languages
This is an order, not a promise of dates.
- German (measured)
- English
- Dutch, Danish, Swedish
- Spanish, Polish
- French, Italian
Known limits and error types
- No detection is perfect. A small remainder can be left even in a shadow copy. When in doubt, nothing is sent.
- With scans, text recognition can make mistakes.
- A person who can only be recognised by their role is not caught reliably by every stage.
- A separate test series for code, configuration files and logs is still to come.
- IT-InterMirror does not handle audio or video yet.
Who measured?
Own measurements
We measured these values ourselves. There is no third-party review as long as no external review exists. The values are not a certificate.
Interested?
We are looking for partners. We test IT-InterMirror with made-up or cleared examples first. Only then does real text go through.
This page is also available as Markdown for AI reading systems. /en/measurement.md