PDF Prompt Injection Toolkit – Hidden LLM Payloads in PDFs(github.com)
github.com
PDF Prompt Injection Toolkit – Hidden LLM Payloads in PDFs
https://github.com/zhihuiyuze/PDF-Prompt-Injection-Toolkit
3 コメント
I think the easiest solution to prevent such injections is just to render PDFs by page into images, and feed the images into whatever downstream task. I believe this is enough to prevent the attacks techniques listed in the repo :)