Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Tue, 30 Jul 2013 00:18:21 -0600
From: Kurt Seifried <>
CC: "Christey, Steven M." <>,
        Evan Teitelman <>,
        "" <>
Subject: Re: CVE Request - Coin Widget serves code over plain

Hash: SHA1

On 07/28/2013 06:07 PM, Christey, Steven M. wrote:
> Kurt Seifried said:
>> The problem is not in the code, the problem is in how the code
>> is served/distributed. CVE is traditionally for software and not
>> for services. So under a simplistic reading of that strict
>> definition I would say this doesn't deserve a CVE.
> tl;dr I've looked into this issue some more and in general, I
> agree.
> I suspect there are a couple issues here; note that these are some
> of my impressions and not anything "official" from CVE:
> 1) The downloading and execution of code from an "http://" URL is
> subject to various attacks, including DNS spoofing and MITM.  To my
> way of thinking, the main issue is that the code is not downloaded
> in a way that preserves its integrity AND ensures that it was
> downloaded from a trusted source.  For CVE, a core question for
> inclusion is, "does this download of code WITHOUT an integrity
> check (CWE-494) happen automatically, or is there a documented
> manual step in which the administrator is expected to verify the
> integrity?"
> 2) With Coin Widget in particular, the widget is available in
> github, and people can download the code and install it on their
> own servers (see the "Download the Source Code" link from the main
> page at  Thus Coin Widget can be offered
> as a customer-controlled "product" (not just a service) and, as a
> product, could qualify for a CVE, but read on...
> 3) However, from, it appears that the Coin
> Widget installation documentation tells the installer to modify
> widget/coin.js to point to an admin-controlled source.  This
> suggests that it's an admin-controlled configuration, which may
> exclude it from CVE.
> 4) The "Wizard" that generates Coin Widget code for people is out
> of scope - this is inherently "site-specific" in that there would
> be no customer actions to fix a vulnerability; the Coin Widget
> admins could modify their code to avoid use of http:// URLs
> entirely, without any action on behalf of customers.
> 5) One could argue that this issue is due to a fundamental problem
> in HTTP, and as such, HTTP should be "blamed" for not having
> integrity checks; but, to assign a CVE to every commonly-used
> protocol that doesn't use encryption is not necessarily
> appropriate, either.
> All in all, for now, it seems to me that this particular Coin
> Widget issue is out of CVE's scope because of the
> software-as-a-service and configuration considerations, but the
> general issue of "reading and executing scripting code from http://
> links without verification" may qualify.

So like CVE-2009-3555 we have a single CVE for the issue and any
future instances can be submitted to cve-assign@ for addition? Problem
is this CVE would become monstrously huge and unwieldy.

>> However the world is changing, for example a program that
>> included an auto-updater component that was advertised as being
>> "Secure" but went over HTTP would probably qualify for a CVE.
> It probably would, but since this might theoretically affect any
> software that's 5 years old or more and downloads anything over
> unencrypted channels without integrity checks, the raw number of
> CVEs that could be assigned is rather daunting.

Yup. Maybe only apply it to common software/software that advertises
"Secure" updates? Seems like a cop out (I vote we do it right or not
at all, but first define what "right" is =).

>> Steve I'm bouncing this to you, I'm inclined to NOT assign a CVE
>> since it opens up a huge can of worms (every single bit of
>> JavaScript served from HTTP and not available via HTTPS ever),
>> but I can also see how it should maybe get a CVE.
> My gut reaction is that you might be treating this as a more
> complex issue than it really is.  Simple delivery of code over HTTP
> might affect Coin Widget and many other packages, regardless of
> whether *some* code is delivered over HTTPS.

One thing I've been thinking about is all the software on the planet
that I have no real clue as to how secure/well developed it (to say
nothing of how well a packager like
Fedora/Debian/Ubuntu/Dreamhost[1]/etc cares for it). Some signals
exist, like do they have a security@ contact that actually replies? Do
they have a bug tracker? Do they offer updates? Did they set static
SSH host keys? In general we have no way of knowing, perhaps
OSVDB/someone wants to start collecting data on known good and known
bad software projects in addition to vulnerability info (e.g. has
"bugtracker, URL is at, bonus point for supporting security bugs that
are private)"? But then the legal issues "you said something mean
about us so we're taking you to court in the UK for libel" come up and
I basically give up since I have no answer for that one =).


>> The good news is that future versions of Firefox are implementing
>> a security policy that when loading a page from HTTPS they will
>> not load page components from HTTP, which would fix this issue.
>> Hopefully all the browsers do this.
> While there are no formal rules, CVE generally considers "typical
> behavior of market-leading browsers" as acceptable considerations
> for determining a vulnerability; e.g., many XSS attack variants
> only apply to 1 or 2 browsers.
> - Steve

- -- 
Kurt Seifried Red Hat Security Response Team (SRT)
PGP: 0x5E267993 A90B F995 7350 148F 66BF 7554 160D 4553 5E26 7993
Version: GnuPG v1.4.13 (GNU/Linux)


Powered by blists - more mailing lists

Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.