Engine · PDF tools

PDF redaction

Follow a selected phrase through preflight, removal and verification in a new PDF.

Workflow

  1. Choose the text to removeA RedactionTarget uses a one-based page and unrotated content-stream coordinates, not screenshot pixels. The target text also identifies what to scrub from document-wide related fields.
  2. Check every targetResolve each target against actual text operators. Reject the entire request if a target is missing or a font, page rotation, overlapping object or attachment cannot be handled safely. No partial target success is advertised.
  3. REFUSED · keep source intactRefusal is the correct result for an unsupported document. A visible black rectangle does not prove the underlying text or embedded content has been removed.
  4. Remove, do not coverRewrite supported content streams to remove the target strings and scrub those strings from related metadata, annotations, form fields and bookmarks across the document. Write a new copy.
  5. Extract text independentlyThe redactor verifies the output using independent extraction. Do not treat output-file creation alone as success. This check covers supported text removal, not every possible hidden representation of information.
  6. REFUSED · check did not passIf the independent check cannot confirm removal, do not offer the copy as successfully redacted. Report the named reason so the caller can select another supported processing path.
  7. REDACTED · verified new copyReturn the output path and the removed target strings. Preserve the original separately and describe the supported scope to the user.

Failure boundaries

  • Screenshot coordinates usedTarget may not match: Request refused
  • Image overlaps targetUnsupported request: No claimed image redaction
  • One target unmatchedWhole request refused: No partial-target success
  • Text still extractableVerification failure: Do not deliver as redacted
  • Verifier unavailableRefused: No success claim

engine.pdf.redaction