Re: any discussion thread for page encodings?

That is what I meant – your (1); though it does not have to be an asymmetric key (cert) alone, could be symmetric key as well.
Trusted UA is just utopia, unachievable practically.

On 8/2/12 3:46 PM, "Richard L. Barnes" <rbarnes@bbn.com<mailto:rbarnes@bbn.com>> wrote:

No, you can't get that guarantee (where you==server), without either (1) device certificates or (2) a trusted UA.  Otherwise, the browser can lie about where it's doing the crypto, then do things wherever it wants, post keys to Twitter, etc.


On Aug 2, 2012, at 2:39 PM, Seetharama Rao Durbha wrote:

I think there are two use cases here.
We cannot support JS integrity (earlier Mountie's email) for reasons Ryan mentioned.
But, implicitly through the ability to interface with smart-cards (and other stores through provider interface hopefully), applications CAN get the guarantee that 'things are processed by a given device'.
-Seetharam
On 8/2/12 3:10 PM, "Richard L. Barnes" <rbarnes@bbn.com<mailto:rbarnes@bbn.com>> wrote:
So, in principle, there do exist techniques for guaranteeing that things are processed by a given device.  It's just that:
1. The relevant standards come from the Trusted Computing Group (not politically very popular)
2. They require per-device certificates (not very practical for browsers)
<http://en.wikipedia.org/wiki/Direct_anonymous_attestation>
I'm pretty sure that these could be adapted to Javascript, though it hasn't been done yet AFAIK.  It may be that the problems Ryan notes would still stand in the way.
My impression is that this group doesn't really to go down that path, especially not in the first version.  If there is a desire to do this later, it can be added onto the draft API.
On Aug 2, 2012, at 12:25 PM, Ryan Sleevi wrote:
On Wed, Aug 1, 2012 at 7:43 PM, Mountie Lee <mountie.lee@gmail.com<mailto:mountie.lee@gmail.com>> wrote:
Hi.
I agree there are no perfect protection for compromised user environment.
but I think if we can do, we have to do to the level we know at least.
I think.
Signed JS has same purpose for keeping JS integrity.
service provider want to make sure the JS codes in user agent are same to original server sent.
if server sent
crypto.sign("I agree to pay 100");
and user switched the string
crypto.sign("I agree to pay 10");
this can be problem in real field.
var integrity = crypto.getIntegrityValue();
can be one of example for my integrity checking mechanism.
also Signed JS can be (maybe) one of alternatives.
this is very basic concern.
best regards
Hi Mountie,
I'm afraid your solution is not really compatible with the JavaScript execution model nor of the general web security best practices.
A server has no way to guarantee that a client has not altered script. Nor can a user agent guarantee that either. There are lots of points where a user may alter code (they may use Web Inspector-style flows, they may open the browser process and rewrite its memory, they may inject executable code, they may inject a bootkit/rootkit, etc). This is a problem that neither can nor should be attempted by this WG.
Your example use case can be perfectly addressed through existing cryptographic transforms, and has been used by a variety of protocols in the past, so I don't think there is anything special about "protecting script" or integrity values needed.
If you're interested in schemes which alter the fundamental web security model and try to protect against these sorts of things, which equally exist as an unsolved problem in "native" code, perhaps the WebAppSec or SysApps WG may be more suitable.
Cheers,
Ryan

On Mon, Jul 30, 2012 at 2:01 PM, Ryan Sleevi <sleevi@google.com<mailto:sleevi@google.com>> wrote:
On Sun, Jul 29, 2012 at 9:46 PM, Mountie Lee <mountie.lee@gmail.com<mailto:mountie.lee@gmail.com>> wrote:
> Hi.
> thanks for your mail.
>
> for "prevent and control"
> the "same origin" policy is not enough I think.
>
> some attackers are focusing to compromise PC environment.
> many attempts are made on PC level.
>
> with "same origin" policy, well designed web application can protect XSS or
> Cross Site Scripting attack.
>
> I think we need one of followings.
> Integrity checking mechanism
> or
> protecting DOM object
>
> the description "webcrypto based methods should not permit DOM changes"
> in Transaction Security use cases
> in http://www.w3.org/wiki/KoreaWebCryptoUseCase
> is described based on same idea of me.
>
> regards
> mountie.
Thank you for the continued feedback.
Unfortunately, if I'm understanding correctly, such controls as you
propose have been agreed as out of scope according to our charter, and
would require rechartering this working group to consider them in
scope. This is listed under "Out of scope" is the statement
"access-control mechanisms beyond the enforcement of the same-origin
policy" at http://www.w3.org/2011/11/webcryptography-charter.html
However, I'm not sure I understand your comments regarding threats to
the PC level. A user agent cannot protect against a compromised
operating environment - whether malware, rootkit, or bootkit, these
attacks can all circumvent the browser security model (as well as the
OS and smart card security models, so these are not new problems).
Thus such attacks are out of scope for this WG, beyond possibly noting
this under security considerations.

It's not clear to me what exactly are you envisioning as suitable
controls. "Protecting DOM object" and "Integrity Checking mechanism"
are not clear, either in attacks or mitigation. Perhaps you can
provide an example of an "Attack", an example of a control, and a
description of how it would mitigate that attack. I just want to make
sure I've properly understood your question and it's out-of-scope
ness.
Note that, as I mentioned, the low-level API provides a way for
applications to provide custom integrity protection (eg: on top of
that afforded by SSL/TLS, which is the recommended means), by
protecting any sort of responses that will lead to DOM manipulation.
Also, please note that both the current low-level API and the previous
high-level API never directly interacted with the DOM, beyond the
incorporation of the DOM language for describing asynchronous event
handling.
If you're trying to prevent 'hostile' JS, then that is not a problem /
use case for this WG, and is indeed something better addressed at an
application layer (eg: via CSP).
Regards,
Ryan
>
> On Sun, Jul 29, 2012 at 6:07 PM, Ryan Sleevi <sleevi@google.com<mailto:sleevi@google.com>> wrote:
>>
>> On Sun, Jul 29, 2012 at 1:29 AM, Mountie Lee <mountie.lee@gmail.com<mailto:mountie.lee@gmail.com>>
>> wrote:
>> > Hi. Ryan.
>> > yes. I read in charter and use cases for "prevent and control"
>> > but no details.
>> >
>> > I remember someone mentioned "signed JS". but that is not acceptable for
>> > wider use.
>> >
>> > can we consider "arguments.callee" (but this is depreciated in ECMA-262)
>> > plus hash?
>> >
>> > If webcrypto api provide JS/DOM integrity checking feature in low level,
>> > web developers can use it for many purpose including "prevent and
>> > control"
>> >
>> > regards
>> > mountie.
>>
>> Hi Mountie,
>>
>> The intent of "prevent and control" is limited in the scope of same
>> origin. There are no attempts being made, and it's largely out of
>> scope for this WG, to try to change or redefine the web security
>> model. This includes the use of arguments.callee.
>>
>> All script executing in the context of the same origin is therefore
>> able to interact with this API. This is also highlighted in the
>> editor's draft under Security Concerns. Applications should thus make
>> use of available functionality, such as CSP, to reduce the risk of
>> cross-site-scripting and other script related attacks. Any further
>> refinements are better suited for the charter of the W3C's Web
>> Application Security Working Group.
>>
>> After our recent face-to-face, the question of signed JS, which is a
>> proposal that had limited qualification and was relayed second-hand,
>> was raised as a point of concern. Those who originally proposed it are
>> being asked to further clarify and refine what their use case was, to
>> better understand where it fits within the scope of this groups
>> charter.
>>
>> Note that with a low-level API, an application could certainly
>> introduce it's own signature verification methods around script
>> content that is passed to either JSON.parse or to eval, which afford
>> application-defined security without requiring this WG to define any
>> additional special APIs.
>>
>> Does that answer your question?
>>
>>
>> >
>> >
>> > On Sun, Jul 29, 2012 at 4:39 PM, Ryan Sleevi <sleevi@google.com<mailto:sleevi@google.com>> wrote:
>> >>
>> >> On Sat, Jul 28, 2012 at 8:55 PM, Mountie Lee <mountie.lee@gmail.com<mailto:mountie.lee@gmail.com>>
>> >> wrote:
>> >> > Hi. Ryan.
>> >> >
>> >> > thanks for your answer.
>> >> >
>> >> > as my understanding, the initial requirement has include "sign with
>> >> > user
>> >> > confirmation".
>> >> > that means the sign data must be same to the message shown to user
>> >> > screen.
>> >> > just with "sequence of bytes" are not enough for previous
>> >> > requirements.
>> >> >
>> >> > if it is handled in high level API,
>> >> > my additional question is
>> >> >
>> >> > is their any mechanism for preventing secret key material changes or
>> >> > other
>> >> > sensitive cryptographic values and settings changes?
>> >> >
>> >> > regards
>> >> > mountie.
>> >>
>> >> Hi Mountie,
>> >>
>> >> Yes, any scheme which relies on user agents presenting data in some
>> >> interpreted form is out of scope for the low-level API, and /may/ be
>> >> in scope for the high-level API. Because such schemes are
>> >> understandably complex (eg: PDF signatures, XMLsig,
>> >> normalization/canonicalization, etc), these are currently not the
>> >> focus of the working groups efforts, but may be in scope at a later
>> >> time. The current focus is on the low-level API, which does not focus
>> >> on such application-specific issues at this time, as even a low-level
>> >> API is a very ambitious and broad endeavor for user agents to
>> >> normalize.
>> >>
>> >> I'm not sure I understand your second question, but you may find it
>> >> answered by the Working Group charter at
>> >> http://www.w3.org/2011/11/webcryptography-charter.html , which states
>> >> that the API shall prevent or control access to secret key material
>> >> and other sensitive cryptographic values and settings.
>> >>
>> >> Regards,
>> >> Ryan
>> >>
>> >> >
>> >> >
>> >> > On Sun, Jul 29, 2012 at 10:41 AM, Ryan Sleevi <sleevi@google.com<mailto:sleevi@google.com>>
>> >> > wrote:
>> >> >>
>> >> >> Hi Mountie,
>> >> >>
>> >> >> There were early discussions about page encodings and normalization,
>> >> >> when the API included both a high-level and a low-level component.
>> >> >> Following our recent Working Group face to face, the Working Group
>> >> >> has
>> >> >> resolved to focus on the low-level API and ensure that it's in a
>> >> >> deliverable state, and then revisit the ideas for a high-level API.
>> >> >>
>> >> >> By virtue of being a low-level API, issues such as encoding are left
>> >> >> to the application author. This can be seen also in the removal of
>> >> >> DOMString (which is UTF-16) from being a valid input to the
>> >> >> CryptoOperation.processData method, since that would have opened
>> >> >> questions of "Is it a sequence of bytes or is it a string?"
>> >> >>
>> >> >> Right now, the API's interaction with data streams is accomplished
>> >> >> via
>> >> >> ArrayBuffer/ArrayBufferViews, which are treated as opaque sequences
>> >> >> of
>> >> >> bytes. Any further normalization or canonicalization should thus be
>> >> >> done at an application layer.
>> >> >>
>> >> >> Does this answer your question?
>> >> >>
>> >> >> On Sat, Jul 28, 2012 at 5:25 PM, mountie.lee@gmail.com<mailto:mountie.lee@gmail.com>
>> >> >> <mountie.lee@gmail.com<mailto:mountie.lee@gmail.com>> wrote:
>> >> >> >
>> >> >> >> Hi.
>> >> >> >> I'm Mountie Lee from Korea.
>> >> >> >>
>> >> >> >> when I read http://www.w3.org/2012/webcrypto/WebCryptoAPI/ and
>> >> >> >> searched
>> >> >> >> archive at http://lists.w3.org/Archives/Public/public-webcrypto/
>> >> >> >>
>> >> >> >> I can not find discussion about page encodings.
>> >> >> >>
>> >> >> >> because WebCryptoAPI is Javascript dependent and developers have
>> >> >> >> to
>> >> >> >> consider page encodings always.
>> >> >> >>
>> >> >> >> many asian sites use local encodings (EUC-KR, Shift-JIS, GB2312
>> >> >> >> ...)
>> >> >> >> and by the page encodings,
>> >> >> >> the javascript operation result is different.
>> >> >> >>
>> >> >> >> please anyone inform me the discussio thread for page encodings.
>> >> >> >>
>> >> >> >> regards
>> >> >> >> mountie.
>> >> >> >>
>> >> >> >> --
>> >> >> >> Mountie Lee
>> >> >> >>
>> >> >> >> Tel : +82 2 2140 2700
>> >> >> >> E-Mail : mountie@paygate.net<mailto:mountie@paygate.net>
>> >> >> >> Twitter : mountielee
>> >> >> >
>> >> >
>> >> >
>> >> >
>> >> >
>> >> > --
>> >> > Mountie Lee
>> >> >
>> >> > Tel : +82 2 2140 2700
>> >> > E-Mail : mountie@paygate.net<mailto:mountie@paygate.net>
>> >> > Twitter : mountielee
>> >> >
>> >> > =======================================
>> >> > PayGate Inc.
>> >> > THE STANDARD FOR ONLINE PAYMENT
>> >> > for Korea, Japan, China, and the World
>> >> >
>> >
>> >
>> >
>> >
>> > --
>> > Mountie Lee
>> >
>> > Tel : +82 2 2140 2700
>> > E-Mail : mountie@paygate.net<mailto:mountie@paygate.net>
>> > Twitter : mountielee
>> >
>> > =======================================
>> > PayGate Inc.
>> > THE STANDARD FOR ONLINE PAYMENT
>> > for Korea, Japan, China, and the World
>> >
>
>
>
>
> --
> Mountie Lee
>
> Tel : +82 2 2140 2700
> E-Mail : mountie@paygate.net<mailto:mountie@paygate.net>
> Twitter : mountielee
>
> =======================================
> PayGate Inc.
> THE STANDARD FOR ONLINE PAYMENT
> for Korea, Japan, China, and the World
>
--
Mountie Lee
Tel : +82 2 2140 2700
E-Mail : mountie@paygate.net<mailto:mountie@paygate.net>
Twitter : mountielee
=======================================
PayGate Inc.
THE STANDARD FOR ONLINE PAYMENT
for Korea, Japan, China, and the World