COWL: A Confinement System for the Web | Hacker News Reader

COWL: A Confinement System for the Web

cowl.ws

62 pointsfla8 comments

This has some heavy hitters behind it, both in authorship and their organizations.

I wondered how it compares to Caja, and this is what their "Towards .." paper states:

"Caja, ADSafe, and FBJS offer sandboxing by defining safe subsets of JavaScript; these subsets tend not to support JavaScript’s full functionality, and as retrofits atop existing browser interfaces, they are also vulnerable to various attacks"

Caja is something you can do server-side to filter JavaScript. COWL is a modification to the browser security architecture. Hence, unlike Caja, a web site cannot unilaterally deploy COWL. On the other hand, if COWL or something like it makes it through the standardization process, it would have a far more significant impact on web security than Caja.

Caja is (more?) often deployed on the client side to filter Javascript.

JavaScript sandboxing is a sort of orthogonal concern to what Cowl is after. Cowl needs sandboxing to work (either by built-in browser support, or by Caja et al), but even with a sandbox, you still need some sort of policy for when to allow data in and out of the scripts, and that's one of the main things Cowl provides.

I like the page, and the example. The key is "privacy with functionality". You can write some JS with certain code paths able to access outside resources, but once you enter constrained code paths, you get limited basically through a dynamically applied SOP.

Their example of a password strength checker -- say you download updated bitmaps (bloom filters) from a central location which alert for risky passwords being used. Now you can trust any code path which actually sees the password will be locked down through taint detection.

I wonder if there's a basic mechanism, or if they have to identify every possibly avenue for tainting? There are a lot of static locations you could stash bits, so are you just denied access to all of that? Or somehow is it all mirrored?

I want to make sure I really understand what they are saying can be trusted, what privacy guarantees are provided? What exactly is sandboxed?

Hello, one of the paper authors here.

One of the jobs that our implementation has to do is close down all channels by which JavaScript could leak data: so this includes postMessage, the DOM, XHR, cookies, local storage... fortunately, these overt channels are well specified DOM APIs so from the implementor's perspective it's not too difficult to lock them all down. (And we just deny access, there's no "shadow" structure s involved.) There's nothing too fine grained going on: if you read private data, everything it isn't allowed to flow to is locked down.

I wish COWL went as far as this proposal for domain sandboxing:

http://mortoray.com/2014/09/30/a-secure-and-private-browser-...

I know Google would never agree to do something like that - unless Mozilla, Apple (the new "privacy" company) and Microsoft (another one pretending to defend user privacy) would put pressure on Google by adopting such a system, together.

> Resources can still be included from a variety of domains, but the data they use is exclusive to this sandbox. A script from google.com on youtube.com would be completely isolated from the same script running at blogger.com.

1. How would that work exactly? There's nothing stopping the script from sending data back to a server for later association with another site.

2. How would a site (or the user) authorize specific 3rd-parties to operate within a given site? This problem is seen by anyone running NoScript today, observing how sites break/work as a subset of 3rd-party scripts is incrementally enabled.

3. If servers are sharing data in real time with commercial 3rd-parties, would they be required to disclose this to users? Publishers don't disclose this today. NoScript allows users to identify 3rd-parties and block them. If this is done on the server, is there a loss in transparency?

4. In general, systems which implement strong isolation are immediately met with user requests to relax this isolation in specific contexts. E.g. Apple added inter-app communication in iOS8. How can "contextual whitelists" be maintained for cross-domain risk management? Should these be determined by the browser vendor, user, or server? Do we need a multi-stakeholder model like CSS, with user preferences negotiated at runtime with publisher preferences? What happens when proprietary DRM is running in the client?

Reply on news.ycombinator.com