You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@sling.apache.org by "Roy T. Fielding" <fi...@gbiv.com> on 2010/02/02 03:59:11 UTC
Re: TRACE method
On Jan 28, 2010, at 7:49 AM, Ian Boston wrote:
> One of our QA people is reporting that Sling servlets support the TRACE method, which can be used for XXS attacks.
No, it can't, or it least it doesn't make any difference whether
TRACE is supported or not because the security leak is allowing
javascript to send cookies and credentials with TRACE. The original
report that TRACE was at fault is ridiculous, and just keeps getting
repeated because "security" reports never die.
The only reason to disable TRACE is to satisfy MIS managers who don't
actually understand security but want to disable it anyway.
....Roy
Re: TRACE method
Posted by Ian Boston <ie...@tfd.co.uk>.
On 2 Feb 2010, at 09:11, Felix Meschberger wrote:
>>
>>
>> btw, SlingSafeMethodsServlet.doTrace looks like it might be vulnerable to Response splitting, it echos headers back to the response stream without making them safe.
>
> Agreed with Bertrand, lets create an issue and fix this in the
> SlingSafeMethodsServlet.doTrace method (probably just omitting any known
> security relevant headers like Set-Cookie and Authorization).
>
> Regards
> Felix
It might just need
headerName = headerName.replace('\n',' ').replace('\r',' ');
headerValue = headerVaue.replace('\n',' ').replace('\r',' ');
or something like that since the split can be applied to any header.
Ian
Re: TRACE method
Posted by Felix Meschberger <fm...@gmail.com>.
Hi,
On 02.02.2010 09:47, Ian Boston wrote:
>
> On 2 Feb 2010, at 02:59, Roy T. Fielding wrote:
>
>> On Jan 28, 2010, at 7:49 AM, Ian Boston wrote:
>>
>>> One of our QA people is reporting that Sling servlets support the TRACE method, which can be used for XXS attacks.
>>
>> No, it can't, or it least it doesn't make any difference whether
>> TRACE is supported or not because the security leak is allowing
>> javascript to send cookies and credentials with TRACE. The original
>> report that TRACE was at fault is ridiculous, and just keeps getting
>> repeated because "security" reports never die.
>>
>> The only reason to disable TRACE is to satisfy MIS managers who don't
>> actually understand security but want to disable it anyway.
>
> The report was from a QA team running through a battery of known "vulnerabilities"
>
> I am sure your right, Should I revert the patch, which, IIUC copies the behaviour in Apache Httpd and allows configuration?
I would keep it (as Roy said "to satisfy MIS managers who don't ...")
>
> btw, SlingSafeMethodsServlet.doTrace looks like it might be vulnerable to Response splitting, it echos headers back to the response stream without making them safe.
Agreed with Bertrand, lets create an issue and fix this in the
SlingSafeMethodsServlet.doTrace method (probably just omitting any known
security relevant headers like Set-Cookie and Authorization).
Regards
Felix
Re: TRACE method
Posted by Ian Boston <ie...@tfd.co.uk>.
On 2 Feb 2010, at 08:51, Bertrand Delacretaz wrote:
> Hi,
>
> On Tue, Feb 2, 2010 at 9:47 AM, Ian Boston <ie...@tfd.co.uk> wrote:
>> ...The report was from a QA team running through a battery of known "vulnerabilities"
>>
>> I am sure your right, Should I revert the patch, which, IIUC copies the behaviour in Apache Httpd and allows configuration?...
>
> I haven't followed the details of this conversation but I'm fine with
> TRACE being disabled by default.
>
>>
>> ...btw, SlingSafeMethodsServlet.doTrace looks like it might be vulnerable to Response splitting, it echos headers
>> back to the response stream without making them safe....
>
> This would warrant a fix and an integration test...but if we agree to
> disable TRACE by default that wouldn't be urgent ;-)
https://issues.apache.org/jira/browse/SLING-1344
Question:
URI encoding for the name and the value or just the value ?
if Just the value, remove \n & \r from the name just in case someone manages to get that through ?
Ian
>
> -Bertrand
Re: TRACE method
Posted by Bertrand Delacretaz <bd...@apache.org>.
Hi,
On Tue, Feb 2, 2010 at 9:47 AM, Ian Boston <ie...@tfd.co.uk> wrote:
> ...The report was from a QA team running through a battery of known "vulnerabilities"
>
> I am sure your right, Should I revert the patch, which, IIUC copies the behaviour in Apache Httpd and allows configuration?...
I haven't followed the details of this conversation but I'm fine with
TRACE being disabled by default.
>
> ...btw, SlingSafeMethodsServlet.doTrace looks like it might be vulnerable to Response splitting, it echos headers
> back to the response stream without making them safe....
This would warrant a fix and an integration test...but if we agree to
disable TRACE by default that wouldn't be urgent ;-)
-Bertrand
Re: TRACE method
Posted by Ian Boston <ie...@tfd.co.uk>.
On 2 Feb 2010, at 02:59, Roy T. Fielding wrote:
> On Jan 28, 2010, at 7:49 AM, Ian Boston wrote:
>
>> One of our QA people is reporting that Sling servlets support the TRACE method, which can be used for XXS attacks.
>
> No, it can't, or it least it doesn't make any difference whether
> TRACE is supported or not because the security leak is allowing
> javascript to send cookies and credentials with TRACE. The original
> report that TRACE was at fault is ridiculous, and just keeps getting
> repeated because "security" reports never die.
>
> The only reason to disable TRACE is to satisfy MIS managers who don't
> actually understand security but want to disable it anyway.
The report was from a QA team running through a battery of known "vulnerabilities"
I am sure your right, Should I revert the patch, which, IIUC copies the behaviour in Apache Httpd and allows configuration?
btw, SlingSafeMethodsServlet.doTrace looks like it might be vulnerable to Response splitting, it echos headers back to the response stream without making them safe.
Ian
>
> ....Roy
>