BASALT · JOURNAL
How to remove confidential information from a PDF
“Confidential information” is a decision category, not a text pattern. Software can find account numbers and repeated names; it cannot determine whether a pricing formula, negotiating position, source identity, or product roadmap is confidential in context.
Convert policy into a review brief
Write down the audience, purpose, governing agreement or policy, information categories approved for removal, and whether visible reason labels are required. Include examples and edge cases. A vague instruction such as “remove anything sensitive” guarantees inconsistent results.
Create a value list for deterministic searches: project names, people, customers, codenames, domains, account numbers, and distinctive phrases. Add spelling variants and abbreviations. Then assign a reviewer to read for concepts a value search cannot find.
Review pages and hidden structures
Inspect text, scans, diagrams, charts, screenshots, headers, footers, speaker notes converted to PDF, and appendices. Apply destructive redaction: remove text objects, edit image pixels, and clear OCR beneath marked images.
Sanitize metadata and XMP, annotations, form values, attachments, hidden layers, link actions, JavaScript, thumbnails, and earlier revisions. PDF sanitization versus redaction explains why both operations are necessary.
Verify the release, not the project file
Export a new candidate and use an independent parser to extract text and enumerate hidden objects. Search every value, inspect images, and compare unmarked content with the source. If the candidate is later merged, compressed, signed, or watermarked, verify the transformed file again.
Record the released file's hash and the review instruction. That creates a useful answer to “what exactly did we send?” without storing confidential values in the log.
Frequently asked questions
Can automatic detection find all confidential information?
No. It can propose known patterns and terms, but confidentiality often depends on meaning and audience. Human review remains essential.
Is deleting a paragraph in a PDF editor enough?
Only if the editor removes the underlying objects and rewrites the file without recoverable history. Verify the exported bytes independently.
Do I need both redaction and sanitization?
Usually. Redaction removes approved page content; sanitization removes non-page carriers such as metadata, comments, attachments, and scripts.
Doing it in Basalt
Basalt combines searchable, reason-coded redaction with whole-document sanitization and an independent verifier. It refuses to save when it cannot prove removal, and the document engine cannot access the network. Download Basalt.
Redaction that proves itself
Basalt destroys the content you mark, then re-opens the file it wrote and proves the content is gone before it saves anything. Your documents never leave your Mac.