W3C home > Mailing lists > Public > public-webappsec@w3.org > September 2015

Re: SRI: edge case when loading the same stylesheet twice in a document

From: Jonathan Kingston <jonathan@jooped.com>
Date: Fri, 18 Sep 2015 00:52:23 +0000
Message-ID: <CAKrjaaULhJhFifZ9aNTj6UPBF0Ny6Qwb6X0iGWfHNruNaFcmcg@mail.gmail.com>
To: Brian Smith <brian@briansmith.org>, Francois Marier <francois@mozilla.com>
Cc: "public-webappsec@w3.org" <public-webappsec@w3.org>
The cache is also never populated on a failed SRI check right?

On Fri, Sep 18, 2015 at 1:06 AM Brian Smith <brian@briansmith.org> wrote:

> On Wed, Sep 16, 2015 at 4:52 PM, Francois Marier <francois@mozilla.com>
> wrote:
>
>> <html>
>> <head>
>> <link rel="stylesheet" href="style.css"
>>       integrity="sha256-correcthash">
>> <link rel="stylesheet" href="style.css"
>>       integrity="sha256-incorrecthash">
>> </head>
>> </html>
>>
>> In Firefox, because of internal optimizations in the CSS loader, we
>> don't check integrity if we have already successfully loaded a
>> stylesheet (with same URI + CORS mode + referrer) within a document.
>>
>> It's not related to the HTTP cache, but rather to an in-memory data
>> structure that gets reused if an author loads the same stylesheet more
>> than once in a document. If you reload the page or load the stylesheet
>> from a different document, its integrity gets checked again.
>>
>
> Presumably it wouldn't be terribly difficult to add the digest to that
> data structure so that this is a non-issue.
>
> However, consider the threat model. The primary threat is that the host of
> the stylesheet IS NOT trustworthy, but the host of the web page IS
> trustworthy. Thus, the attacker can modify the stylesheet but not the web
> page. Therefore under SRI's primary threat model there is no way for the
> attacker to induce this condition, so there's no security issue.
>
> (Note that your particular example doesn't seem very realistic because the
> untrusted host is the same as the trusted host. In general, it is a good
> idea to always use separate hosts in SRI examples, so that the examples
> make sense when considering SRI's threat model.)
>
> Cheers,
> Brian
>
Received on Friday, 18 September 2015 00:53:04 UTC

This archive was generated by hypermail 2.3.1 : Monday, 23 October 2017 14:54:15 UTC