W3C home > Mailing lists > Public > public-wsc-wg@w3.org > April 2007

RE: ACTION 181: Summary of EV certificate discussion, prototype recommendation

From: <michael.mccormick@wellsfargo.com>
Date: Wed, 25 Apr 2007 14:20:51 -0500
Message-ID: <8A794A6D6932D146B2949441ECFC9D6802B4D3C4@msgswbmnmsp17.wellsfargo.com>
To: <pbaker@verisign.com>, <public-wsc-wg@w3.org>
I'm pretty okay with issuer logos.  I withdraw the phrase "and
certificate issuers" from recommendation 2.
 
My concern is more with relying party logos since users will view them
as site identifiers (much like favicons).  A legal process to prevent
ambiguity, resolve infringement, etc, would help but issuers can't/won't
be the enforcement point.
 
Cheers, Mike

  _____  

From: Hallam-Baker, Phillip [mailto:pbaker@verisign.com] 
Sent: Wednesday, April 25, 2007 1:03 PM
To: McCormick, Mike; public-wsc-wg@w3.org
Subject: RE: ACTION 181: Summary of EV certificate discussion, prototype
recommendation


I think that first we have to separate the issue of subject and issuer
icons
 
As far as I am concerned any issuer that is not prepared to put their
own brand on their EV certs shouldn't be issuing them.
 
Issues such as ambiguity and possible confusion are of course vital
here. It does not follow that the EV process needs to be the sole means
of addressing these issues however. There is an existing legal
infrastructure to deal with those issues. What is critical here is
whether or not a certificate holder is or is not within the reach of
that legal infrastructure, whether there is accountability.
 
To paraphrase Judge Hands here: A criminal is unlikely to attack an
infrastructure if the expected reward is less than the cost of making
the attack plus the cost of getting caught times the probability of
being caught.
 
We limit the expected reward using efficient revocation. I would be more
than happy to accept a tight constraint on freshness of recocation data
(OCSP with TTL of 30 minutes or so). 
 
We engineer a situation where the cost of reducing the chance of being
caught to an acceptable level renders the attack uneconomic.
 
 
Ambiguity is in many ways less of an issue where brands are concerned
than in the context of a subject name. There are multiple companies that
trade under the name 'Prince' but their logos are rather less likely to
be ambiguous than their domain name. 
 


  _____  

	From: michael.mccormick@wellsfargo.com
[mailto:michael.mccormick@wellsfargo.com] 
	Sent: Wednesday, April 25, 2007 1:47 PM
	To: Hallam-Baker, Phillip; public-wsc-wg@w3.org
	Subject: RE: ACTION 181: Summary of EV certificate discussion,
prototype recommendation
	
	
	Hi Phil,
	 
	I'm glad we agree about favicons.  I readily concede X.509 logos
are more secure than favicons.  However, unless there is a central
authority and process in place to ensure no two sites use visually
similar or indistinguishable images in their EV certificates, these
logos should not be relied upon as identity cues.
	 
	My reason for bringing this up is I've become aware many EV
issuers and RPs are planning to use these logos for branding and
identity assurance purposes.
	 
	Mike

  _____  

	From: Hallam-Baker, Phillip [mailto:pbaker@verisign.com] 
	Sent: Wednesday, April 25, 2007 12:35 PM
	To: McCormick, Mike; public-wsc-wg@w3.org
	Subject: RE: ACTION 181: Summary of EV certificate discussion,
prototype recommendation
	
	
	Well naturally I totaly disagree with the second recommendation.
	 
	If the issuer has taken appropriate steps to authenticate
certificate subject information and there are approriate accountability
controls it is appropriate to display it. 
	 
	And the certificate issuer can surely be trusted to authenticate
their own brand information. 
	 
	A favicon is objectionable because it is not authenticated and
the use is not accountable. Favicons are not intended to be an
authentication cue, are not secured as such but are easily confused as
being so. An authenticated logotype is explicitly intended for that
purpose and the issuing controls governing their use are designed to
ensure that the level of security is appropriate.
	 
	 
	But since the recommendation does not mention logotype
information at all and it did not come up in the conversation I don't
see where the justification for a prohibition would be.
	 
	
	

  _____  

		From: michael.mccormick@wellsfargo.com
[mailto:michael.mccormick@wellsfargo.com] 
		Sent: Wednesday, April 25, 2007 1:20 PM
		To: public-wsc-wg@w3.org
		Cc: Hallam-Baker, Phillip
		Subject: RE: ACTION 181: Summary of EV certificate
discussion, prototype recommendation
		
		
		This is excellent.  I'd like to propose a couple minor
additions to Phil's recommendation:
		 
		1. Web clients MUST verify revocation status of the EV
certificate and associated CA certificates as a precondition for
displaying a positive security signal.  Web clients SHOULD use OCSP
protocol to perform this check if it is available from the certificate
issuer.
		 
		2. Web sites and certificate issuers SHOULD NOT use
X.509 logographic extensions to visually brand EV certificates.  Web
clients SHOULD NOT display logos attached to EV certificates.
		 
		Rationale:
		 
		1. It is critical to the credibility & trustworthiness
of EV SSL that compromised certificates can be revoked, and that
revocation prevents web clients from signaling positively.
		 
		2. See my earlier email about favicons.  My reasons for
mistrusting certificate logos are essentially the same: They are
mistakenly viewed as identity cues or security context signals by most
users, which makes them easily exploitable.  Also they blur the
distinction between chrome and content.
		 
		Thanks, Mike
		 
		P.S. In interest of full disclosure, Wells Fargo is a
member of the CAB Forum and will likely be a commercial issuer of EV SSL
certificates.

  _____  

		From: public-wsc-wg-request@w3.org
[mailto:public-wsc-wg-request@w3.org] On Behalf Of Hallam-Baker, Phillip
		Sent: Wednesday, April 25, 2007 9:52 AM
		To: public-wsc-wg@w3.org
		Subject: ACTION 181: Summary of EV certificate
discussion, prototype recommendation
		
		
		Extended Validation (EV) Certificates are X.509v3
certificates that are issued in accordance with minimum authentication
criteria defined by the CA-Browser Forum, as demonstrated by a Web Trust
audit of the issuer. 

		When certain browsers (currently IE7 with the
anti-phishing filter enabled or Firefox with a plug-in) visit a site
secured with SSL and an EV certificate an enhanced user experience is
presented. Currently in IE7 this user experience consists of the address
bar turning green and the certificate subject and issuer being displayed
alternately next to the address bar. Advertising and User education
programs for EV certificates frequently use phrases such as 'Green for
Go'.

		The current CAB-Forum criteria are specified at:
http://www.cabforum.org/EV_Certificate_Guidelines.pdf

		The core principle of EV certificates is to assure the
relying party that there is accountability. The EV issuing criteria
require a certificate issuer to verify that the organization specified
in the certificate subject exists as a legal entity, that the physical
contact address is valid, that the party making the application is
authorized to do so by the specified subject, that the applicant has
control of the domain name specified in the certificate and that the
applicant has possession of the private key corresponding to the
specified public key.

		If the issuance process should fail the certificate
issuer is to be held accountable, hence the display of the issuer name
in the EV security experience.

		An EV certificate does not guarantee that a vendor
selling a $3000 plasma TV will deliver a product that works or even that
the TV will be delivered at all. But an EV certificate does guarantee
that either the vendor can be held accountable for the default or the
certificate issuer held accountable for the failure of their
authentication program.

		In addition to requiring minimum authentication
standards and a WebTrust audit, the CABForum guidelines require CAs to
implement certain technical measures. Certificates must conform to a
specific profile of the X.509v3 and PKIX standards. Wildcard
certificates are not permitted, the maximum validity interval is one
year, a means of efficient revocation must be supported and certain
minimum key sizes must be employed. These measures are generally
consistent with community consensus amongst network security
specialists.

		Although the EV guidelines do not mandate support for
publication of status information by OCSP, accountability of the
certificate issuer creates a significant incentive to do so. 

		While EV certificates are designed to be compatible with
the existing deployed base of browsers the use of EV makes use of
certain protocol extensions which are already highly desirable to be
even more desirable. In particular support for the certificate host
identification extension and OCSP token stapling. 

		Although the EV security experience in IE7 is designed
to resist certain types of attack it is impossible to provide absolute
guarantees of security on a platform that is not considered trustworthy.
For example the IE7 does not allow users to add EV signing certificates
to the root store. IE7 only presents the EV security experience if a
root certificate that is countersigned by a Certificate Trust List
signed by an offline root. An attacker can only circumvent this
requirement by compromising the IE7 executable.

		Prototype recommendation:

		Web clients SHOULD present a distinctive user experience
in response to presentation of an EV certificate. Such a user experience
SHOULD be guarded against emulation by browser content. The certificate
subject and issuer SHOULD be displayed in the primary user experience.

		In order to facilitate accessibility such a user
experience SHOULD NOT use color alone, although if color is used the
color green SHOULD NOT be used EXCEPT to indicate a positive condition
and the color red SHOULD NOT be used except to indicate a negative
condition.

		In order to reduce the consuption of IPv4 addresses Web
Clients SHOULD support the TLS host name identification extension on all
SSL/TLS transactions regardless of whether an EV certificate is
presented or not.

		For efficiency, Web Clients SHOULD support the OCSP
token stapling extension of TLS on all SSL/TLS transactions regardless
of whether an EV certificate is presented or not.
Received on Wednesday, 25 April 2007 19:20:59 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 5 February 2008 03:52:47 GMT