How to redact a Word document permanently

Why a black highlight is not redaction

Word stores text and its formatting separately. A black highlight or a black font colour changes the formatting only: the words are still in the file, and anyone can select them, copy them, search them or remove the highlight. Saving the document as a PDF keeps them as well.

Why a black box drawn over text hides nothing

Where a Word document keeps information

A .docx file is a package of several parts. Personal data can sit in any of them:

• the body, with its tables, text boxes and footnotes or endnotes;
• the headers and footers of every section;
• comments, and the names of the people who wrote them;
• tracked changes: deleted text stays in the file until the changes are accepted;
• fields, whose instructions can hold a name or a path;
• pictures, which can show a signature, a stamp or a scanned identity document, and their descriptions (alternative text);
• the document properties: author, last modified by, company, manager, title;
• embedded objects, such as a spreadsheet pasted into the page, and the thumbnail of the first page.

Redacting by hand in Word

Work on a copy. Accept or reject every tracked change, delete the comments, then replace each value to remove with neutral text using Find and Replace. Remember the variants: a name with and without its first name, an identifier with and without spaces.

Pictures need another tool: a black rectangle drawn over a picture in Word is a separate shape that can be moved away. Finally, Word's Inspect Document (File › Info › Check for Issues) removes comments, revisions and document properties.

This works for a short document, but every step is manual and nothing tells you that a footnote or a header was missed.

Redacting a Word document with Obscelis

Import the .docx file. Obscelis reads the body, then the headers, footers, footnotes, endnotes and comments, the deleted text of tracked changes, field instructions and picture descriptions, and reads the text in pictures with on-device OCR. It proposes names, addresses, e-mail addresses, phone numbers, IBANs and identifiers; you accept or reject each one, and draw zones over what a picture shows.

The copy is a Word document again. Each masked character becomes █, so the layout holds and the length of a value stays visible; masked areas of pictures are painted black and every picture is written again; embedded objects are emptied, the thumbnail is removed, and the document properties and the names of comment and revision authors are cleared unless you choose to keep them.

Before the copy is published, Obscelis opens it again and searches every part for the masked values. If one remains, for instance in a part it does not review such as SmartArt text, the copy is refused rather than delivered with it. Everything happens on your computer.

Download Obscelis

Check the copy before sending it

1. Open the copy and search for each masked value, including part of it.
2. Show the comments and tracked changes: there should be none left that you did not mean to keep.
3. Open File › Info and read the properties: author, company, title.
4. Send the copy, never the original, and name it so that nobody confuses the two.

Questions