Share a document without leaking metadata
A document carries more than its pages: the author's name, the software and the machine, revision history, comments, sometimes tracked changes that were never accepted. Three tools in order find that, remove it, and give the recipient a way to confirm they got the file you sent.
The steps
- 1
Inspect and clean the file
The inspector lists what the file says about its author and history, and writes a cleaned copy with the properties, comments and hidden content removed. Read the findings before cleaning; sometimes the surprise is that the file is not the version you thought.
Open Document Privacy Inspector - 2
Scan the visible text for personal data
Copy the document's text and paste it. The scanner finds emails, phone numbers, card and account numbers, addresses of servers and keys, each verified by checksum where one exists, and can produce a redacted copy of the text to check against the document.
Open Sensitive Data Scanner - 3
Hash the cleaned file to send alongside
A SHA-256 of the final file, sent in the message body, lets the recipient confirm the file arrived unaltered. It is also a record of exactly which version you sent.
Open Hash Generator
Each tool offers the next one from its result. The file is handed on in memory, in this tab, and is not stored.
What the inspector removes
For Office files: the core properties (author, last modified by, company), the application properties, comments, tracked changes that were still pending, custom XML parts and embedded thumbnails. For PDFs: the info dictionary, XMP metadata and embedded files. It rewrites the container rather than editing it in place, so the cleaned copy contains only what was written back, and it re-inspects the result so you can see what remains.
Why the text scan is separate
Metadata is about the file; the scan is about what the pages say. An offer letter with the author removed still contains a salary and an address. The scanner finds the shapes it can find, by pattern and checksum, and is explicit about what it cannot: a person's name or a street address have no pattern, and a clean scan does not mean an anonymous document. It is a second look, not a certificate.
- Contracts and offers going outside the company.
- Reports built from a template that has someone else's name in its properties.
- Anything produced by a shared machine, whose properties may name the wrong person entirely.
Sending the checksum
Put the hex digest in the covering message, not inside the file. The recipient runs the same hash on what they received and compares. The point is less to defeat tampering, which a checksum sent over the same channel cannot do, than to settle which version was sent when a question comes up later.
Questions
- Does cleaning change the document's content?
- No. Text, formatting and images are untouched. Comments and pending tracked changes are removed, because they are hidden content; accept or reject changes first if they should become part of the document.
- Can I do this for many files at once?
- The Batch Workspace inspects PDFs in bulk and reports the findings as a CSV, and hashes any number of files to a checksum list. Cleaning is per file, because each cleaned copy should be read.