3 ms·
Allowing known-risky formats that keep failing over and over is negligent. But if you look at this from the opposite direction, you're essentially arguing that
by Silhouette 11y ago
Allowing known-risky formats that keep failing over and over is negligent.
But if you look at this from the opposite direction, you're essentially arguing that we should only use technologies that are known, or at least reasonably expected, to be extremely safe.
Given that in general humanity hasn't yet figured out how to create such technologies, and that numerous formats we use every day on the web to great overall benefit would not qualify, that seems a tall order.
- Dylan16807 11y ago> But if you look at this from the opposite direction, you're essentially arguing that we should only use technologies that are known, or at least reasonably expected, to be extremely safe. No I'm not. Go ahead and use a new technology. But don't use a proven-bad technology. If you tried a reasonable amount and don't know about security holes, that's one thing. If someone shows you the security holes, and you don't fix them, that is where you're a bad actor.
- Silhouette 11y agoGo ahead and use a new technology. But don't use a proven-bad technology. But the "proven bad" technology you're talking about here is just incorporating any third party content in your site. Obviously that is a security risk if the third party isn't perfect about policing what they host. On the other hand, billions of resources are served in that way every day, and the web is a much better place for it. Only a tiny fraction of those third party resources are hostile, and most of the ones that are will be closed down rapidly by the third party service themselves once discovered. So is this really in "proven bad", "known-risky" territory, or are we actually talking about "very rare" dangers and a lot of hyperbole here?
- Dylan16807 11y agoThe problem is incorporating third party content that is neither screened nor sanitized. The proportion doesn't matter as much as the fact that there is nothing stopping attacks. They only get cleaned up afterwards. You don't expose your users to attack without warning. I think very few websites allow one person to embed arbitrary scripts that will be shown to another person.
- Silhouette 11y agoThe problem is incorporating third party content that is neither screened nor sanitized. It is impossible to screen or sanitize third party content if the third party is hosting it and the user loads it when your page refers to it. The third party can change that content at any time, without your knowledge or consent. This is how almost all ad networks work. It is also how almost all CDNs, web font services, image hosting services, etc. work. I think very few websites allow one person to embed arbitrary scripts that will be shown to another person. Every single site on the web that hosts jQuery via a CDN does exactly that. This single example alone represents many millions of sites.
- Dylan16807 11y agoWhy are you conflating individually-trusted CDNs with the servers of some random guy? And allowing only images, like most embeds do, is a form of sanitation. I'll repeat myself. "I think very few websites allow one person to embed arbitrary scripts that will be shown to another person." This is not happening as a result of you using an image host. No scripts are involved there. This is not happening as a result of the site using a CDN. No user triggered that load of jQuery. It's fine to load jQuery from a specific server that you trust. It's also fine to load ads from the ad network's server, as long as they are policing uploads properly. The problem is they usually don't.
- Silhouette 11y agoIt's fine to load jQuery from a specific server that you trust. It's also fine to load ads from the ad network's server, as long as they are policing uploads properly. The problem is they usually don't. You keep saying they usually don't, but billions of harmless ads are served every day while only a tiny fraction of the served ads are malicious. I just don't see how it's reasonable to assume depending on a third party ad network for content is fundamentally risky yet depending on some other third party service is not. CDNs and other hosting services get hacked and serve malicious content sometimes too, but that is also very rare and also usually gets fixed very quickly if it does happen.
- MichaelBurge 11y agoYou can use a theorem prover like Coq to generate software that proves an image parser correct.
- Silhouette 11y agoExistence proof of a 100% safe image parser built using a formal theorem prover, please.
- Dylan16807 11y agoFor an image parser all you really need is to use a language without buffer overflows.
- Silhouette 11y agoThere are other potential attack vectors than buffer overflows, though that seems the most likely source of vulnerabilities if you're going to limit images to JPG or PNG. Of course now you're not only prohibiting third party resources except images, you're even prohibiting modern image formats like SVG, which is a little ironic since SVG-based ads might be smaller and/or look cleaner than equivalent bitmaps.