Want to ensure financial documents cant be parsed by automated systems

  • OsrsNeedsF2P@lemmy.ml
    link
    fedilink
    arrow-up
    28
    ·
    21 days ago

    PDF scanning is done by both OCR and PDF analysis so no. If you, a human can read it, a bot can read it too.

    Your best bet is the classic inserting BS in a 1-hex-off-white font

  • Natanael@slrpnk.net
    link
    fedilink
    arrow-up
    7
    ·
    21 days ago

    There’s DRM solutions but they’re by definition not perfect, if it can be read then a photo can be taken

  • cannedtuna@lemmy.world
    link
    fedilink
    arrow-up
    3
    ·
    21 days ago

    OCR cannot scan documents that have been certified or digitally signed.

    Note that once you certify a document it can no longer be edited, combined with another PDF, or have pages inserted or extracted.

    Once a PDF has been digitally signed it is locked and you can no longer add pages, delete pages, or read it via OCR.

    • GBU_28@lemm.ee
      link
      fedilink
      English
      arrow-up
      8
      ·
      21 days ago

      What? If the document is accessible, and human readable, it’s parsable by OCR

      • cannedtuna@lemmy.world
        link
        fedilink
        arrow-up
        1
        ·
        20 days ago

        I don’t know what to tell you dude. A certified or digitally can’t/wont be read by OCR. A digitally signed document legally certifies that the document has not been modified. PDF editors such as Bluebeam or Adobe will not or cannot process a certified or digitally signed document.

        I’m not sure if that limitation is due to the process by which the document is certified or if it is a feature of software conforming for legality reasons. I’m not going to research this for OP, I’m just providing a simple and best accurate answer.

        Maybe current AI has better abilities to process document text? I’m not sure, maybe. But you’d think this would be a shared concern with groups wanting to protect documents for the same reason and therefore encryption would match.

        If it’s just the legality of it stopping a company from providing the feature, you would think most companies would want to keep out of legal hot water and would then disallow OCR processing. In this case sure there could be software that doesn’t conform, but for most application purposes I don’t think you’d have to worry too much.

        • BCsven@lemmy.ca
          link
          fedilink
          arrow-up
          3
          ·
          edit-2
          20 days ago

          Lots of software can manipulate PDF. Open PDF in libredraw change pages,print as PDF or export as PDF. A system that skims content is purposely going to bypass any signed restriction.

          Edit: Here’s how to bypass restriction in Paperless OCR.

          The parameter PAPERLESS_OCR_USER_ARGS: ‘{“invalidate_digital_signatures”: true}’ in the context of Paperless-ngx and OCRmyPDF allows OCR processing of PDF documents that have been digitally signed by intentionally invalidating those signatures. In its standard configuration, OCRmyPDF does not process documents with digital signatures so as not to compromise their integrity. Setting this parameter to true allows OCR on such documents

        • GBU_28@lemm.ee
          link
          fedilink
          English
          arrow-up
          3
          ·
          20 days ago

          Many alternative OCR tools now simply screenshot the page. This is a cracked issue.

    • MystikIncarnate@lemmy.ca
      link
      fedilink
      English
      arrow-up
      4
      ·
      20 days ago

      This works, right up until you introduce PDF compatible software that doesn’t give a shit about your rules, of which there’s plenty.

      You can also print/scan, or even print to PDF to get around such limitations. The original document cannot be altered since that would invalidate the digital signature on the file, but you can create a perfect digital copy, omitting the signature, and modify it however you want.

      If online systems that are skimming documents for their contents don’t give a shit about what the signature is, and simply take a copy and OCR it to train an AI or amalgamate the information for data harvesting or other purposes.

      I get what you’re saying and in concept, it should be fine, the problem is that it’s a software lock/restriction on a file type that isn’t inherently closed source, unknown, nor was the PDF format built to be secure from the ground up. So we’re applying security to a system that wasn’t built for it.

  • MystikIncarnate@lemmy.ca
    link
    fedilink
    English
    arrow-up
    2
    ·
    20 days ago

    I would OCR it myself, but edit the meta data in the file so that the text in the OCR metadata is lorem ipsum.

    So any bots that assume that the OCR text is what’s on the image in the PDF (and why wouldn’t they), it will only read useless junk. Only someone reading the text from the image would “see” it, and only a bot programmed to OCR a file that already has OCR metadata would realize that there’s any inconsistency.

    I’m not entirely sure how to accomplish that, but I’d figure it out if I was worried about the data being compromised.

    Personally, I would simply keep the file in an encrypted container, then I wouldn’t worry about what can scan the file since it would be entirely unreadable ciphertext without the correct security key or passphrase.