PDFium sends HTTP(S) requests without prompting
Reported by
c...@ecraig.com,
Feb 10 2017
|
||||||
Issue descriptionUserAgent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_11_6) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/55.0.2883.95 Safari/537.36 Steps to reproduce the problem: 1. Go to https://brainwashing.secur3.us/test.pdf in Chrome 2. Observe that this document triggers a web request without asking permission. Alternatively, follow instructions to generate a canary PDF document from http://canarytokens.org/generate and load the file with Chrome What is the expected behavior? Chrome/pdfium should be prompting before making an external web request to prevent inadvertent privacy exposure. Adobe's PDF products for example has a permissions request. OS X Preview tool simply ignores the external resource referenced in the document. What went wrong? By opening this PDF document in Chrome, the use is redirected without prompt to an arbitrary HTTP server. This is undesirable as it is a privacy exposure revealing the IP address and user-agent that opened the PDF. Did this work before? No Chrome version: 55.0.2883.95 Channel: n/a OS Version: OS X 10.11.6 Flash Version: Shockwave Flash 24.0 r0
,
Feb 10 2017
The same risk inherently holds true for HTML documents, and it's not clear that users have any different expectations of privacy when opening a PDF file vs. a HTML file.
,
Feb 10 2017
For plugins not to be able to make HTTP requests has never been a Chromium security goal, and I don't think it is for the privacy team, either. PDFium can do much less than Flash, for example. PDFium is just another web contents renderer. Anyway, I'll leave it for Privacy team to decide. Maybe I'm wrong.
,
Feb 10 2017
Thank you for your feedback and for relaying my concerns onto the privacy team. In general, I had not thought of this as being web content rendering in the way that for example Flash is. In general, I think that Google has very high-standards for privacy and therefore I was very unpleasantly surprised to find that it is actually doing less (in this situation) to safeguard my privacy than Adobe, Apple, or Microsoft.
,
Feb 13 2017
To me it makes a big difference in which context the user sees a PDF: - If I open foobar.com/test.pdf, I would not mind if test.pdf fetchs resources because foobar.com has the same power to leak the fact that I open the PDF. - If I open a local/downloaded file, I would not expect that to have the capability to communicate to the outside world. I'd be supportive to a prompt in the latter case and open to a prompt in both cases. I'd be curious how often it would be seen by users.
,
Feb 13 2017
Yes, the open file from local file:// case is certainly more concerning, but also consider that the file on foobar.com can trigger DNS and make requests to other sites. This is important if foobar.com may be a "trusted" CDN or any site allowing user uploads. The Firefox and Apple Preview behavior seems more appropriate to me. (PDF does not trigger external resource requests in my tests.)
,
Feb 13 2017
Additionally, I would like to point out that there is the potential for unexpected or abusive behavior if the PDF document makes a /URI reference back to itself as shown with https://brainwashing.secur3.us/foo.pdf In my browser, this document makes the tab reload in a loop and if I try to open a new tab, the file is also loaded in that tab and keeps reloading itself. Similarly, if I open a new window, the document is loaded in the new window rather than opening the "new tab page" as my settings indicate should happen.
,
Feb 13 2018
This issue has been Available for over a year. If it's no longer important or seems unlikely to be fixed, please consider closing it out. If it is important, please re-triage the issue. Sorry for the inconvenience if the bug really should have been left as Available. If you change it back, also remove the "Hotlist-Recharge-Cold" label. For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
,
Jun 4 2018
|
||||||
►
Sign in to add a comment |
||||||
Comment 1 by c...@ecraig.com
, Feb 10 2017