- From: Stephen Chenney <schenney@chromium.org>
- Date: Fri, 27 Mar 2015 09:40:31 -0400
- To: Daniel Holbert <dholbert@mozilla.com>
- Cc: Doug Schepers <schepers@w3.org>, David Dailey <ddailey@zoominternet.net>, www-svg <www-svg@w3.org>, Chris Lilley <chris@w3.org>
- Message-ID: <CAObCcUrcSAkvAEVrpRtZWZ=MVDf7N=Gdov1KRR4pZ3ukMJV3pw@mail.gmail.com>
[Resending from correct address.] On Fri, Mar 27, 2015 at 12:35 AM, Daniel Holbert <dholbert@mozilla.com> wrote: > On 03/26/2015 08:48 PM, Daniel Holbert wrote: > > (One scenario I've thought of, which might be what you're getting at > > [please bring up other scenarios though]: some interactive user-provided > > SVG content, where if you click just the right spots, some horrifically > > obscene picture pops up. This is a problem, though it also seems like > > it could be a problem for e.g. animated GIFs or simple animated SVG or > > video -- some arbitrary frame of the animation/video could have the same > > obscene content. I'll grant that it may be more discoverable via > > automated tools in these non-interactive formats, though. (to the extent > > that your tools can actually tell that it's obscene.) > > One other trick in this category -- even with static SVG, a sneaky user > could sneak something objectionable past automated filters & human > content-screeners using CSS media queries in their SVG. They could use a > media query to control whether a piece of content is visible, based on > the viewport-size (which I think would translate to the image's size), > or something else that can vary between the content-screener's > environment & the actual-site-on-a-particular-user's-device. > > I think this is somewhat similar to burying an image at frame N of an > animated GIF (where N is huge), so it's not entirely a problem that's > new when you introduce SVG. But I'll grant that tools can probably scan > GIF-frames more trivially than they can probe the entire space of > media-query-influenced renderings. > Daniel, you've hit on exactly the kinds of reasons why I think event driven or navigation driven content might be harder to screen for objectionable content. I agree that screening is technically possible, and that existing systems might miss content. My goal was to add context to the discussion and try to guide expectations as to the impact of improving interactivity in SVG-as-image. In particular, turning on SVG on a social site is not just an issue of enabling upload and display, so one might expect much greater push back from site developers. Cheers, Stephen.
Received on Friday, 27 March 2015 13:40:59 UTC