Nikolai Sokolov

My whole beat is the difference between what a document appears to hide and what it no longer contains at all, and that gap is where disclosures happen. Drawing a black rectangle over text does not remove the text. The characters remain in the content stream, and so do they when the file is opened in an editor, copied from, or processed by anything that extracts text. Redaction means rewriting the content stream to omit the characters entirely, and the tool doing it has to be trusted to have done so. Metadata is the quieter version of the same problem. A document carries a producer, a creation tool, and sometimes an author, and any of those can identify a person or a system you would rather not name. Passwords split into two kinds. An owner password controls whether the file can be opened. A user password controls permissions such as printing and copying, and it is enforced by the application rather than by cryptography, which is why removing it takes seconds. That distinction matters for compliance work, because a permission password has never protected anything. Encryption strength depends on the revision used. An early revision uses a scheme with a known weakness, and a later one does not, so two files both described as password protected can differ by orders of magnitude in how long they hold. Every page here says plainly that these controls are deterrents, not boundaries. Signed documents add a further wrinkle, because altering the content invalidates the signature and the whole point of signing was to detect that alteration. Redaction and signing are in tension, and the order they happen in decides which one you end up with.