Public trial instance. Anyone can open these pages and see the documents issued here. Use sample documents, never real ones. Resolver addresses are shown without their host part.

Offline marking · Tardos code

Issue marked copies

Every recipient gets a copy with its own fingerprint. If a copy leaks, TAPAK can point to the copy it came from and give the chance that the accusation is wrong.

1 Document
2 Plan
3 Review text
4 Issue

Document and recipients

The master document stays on this device. The registry stores only the document’s hash and each recipient’s code commitment.

Master document (PDF) Drag it here or choose a file. A PDF with a text layer, or scanned (at least 200 dpi), up to 40 MB.

Recipient list (CSV) Drag it here or choose a file. Required columns recipient_id and nama (name); jabatan (position) and satuan (unit) are optional.

Issuer (optional) Recorded in the registry as the issuing party.

Background pattern A layer of code under the page text. Choose by how the copies will be shared.

Screen (default) Invisible. On paper it depends on the printer: it was read on the printer tested, more weakly than the print profile. Print Strongest on paper: read almost without errors on the printer tested. The page looks light grey.

Print-profile background: pages look light grey, on screen and on paper, and the texture of the pattern may show on close inspection. Use this profile only when the copies will be printed and you accept that look.

Word shift scheme Word and line shifts. On paper they add evidence; it is the background pattern that names the copy there.

v1 Control and data word groups alternate, 0.05 mm shifts. v2 (default) Every word group carries a bit, 0.025 mm shifts: more positions in a sharp photo of a printed page.

Code and canary settings

Coalition size resisted (c) How many recipients might collude by comparing copies.

Maximum chance of a false accusation For the whole recipient list, fixed before any analysis.

Canary domain Leave empty to turn canary tokens off.

Reading the text layer and testing the visual gate. For a scanned PDF: reading the word boxes of the scanned pages with Tesseract (a few seconds per page).