[security] Nonrepudiation, and Trusting OpenID Providers

Thu Dec 10 03:46:27 UTC 2009

I agree with John.  You always have to trust someone/something.  Gmail could
use my bank's password recovery on my behalf.  I trust them not to.

What I think is more dangerous, however, is the lack of policy enforcement
or trust in that enforcement.  If I were a bank, I wouldn't want customers
logging in from a Gmail account because Gmail has much looser password
restrictions.  Since every web app has different requirements for security
and authentication and because it's so easy to just build the functionality
yourself, it makes more sense to just do it yourself.  Then you can enforce
whatever password policies you want.

In other words, I personally trust Google not to reset my passwords.  My
bank, however, why should they trust Google?  Or Joe-Shmo-Provider?

- Brandon

On Wed, Dec 9, 2009 at 9:44 PM, John Bradley <ve7jtb at ve7jtb.com> wrote:

> PS,
>
> If you use your email account for account recovery your email provider can
> get access to all of your other accounts.    That is one of the largest
> security problems.
>
> John B.
>
> On 2009-12-09, at 8:53 PM, Shearer, Charles Dylan wrote:
>
> John & Andrew,
>
> That you both for your interesting responses and pointers.
>
> Andrew, you said:
>
> your question moves past the security issues associated with the protocol
> and addresses trust issues in the ecosystem. As John indicated, at the
> current assurance levels and resulting appropriate usage scenarios, OpenID
> deployment and the trust levels associated with OP's are probably
> sufficient.
>
>
> And John said:
>
> Given that openID is only secure enough as a protocol for ICAM LoA 1
> (pseudonymous protecting no PII) the most practical path is to provide more
> trustable OP/IdP.
>
>
> I think you both are saying that, for what OpenID is intended to do, it
> works.  In other words, OpenID can be used if we can assume that the
> supported identity providers are honest.  I cannot understand why we should
> have to make that assumption at all.  There are well-supported technologies
> like PKI that provide identification, authentication, and even
> nonrepudiation without forcing us to put a very large amount of trust in a
> 3rd party.
>
> Indeed, my point is that OpenID forces us to put a *lot* of trust in the
> identity provider: we trust them with the ability to access our RP accounts,
> and we trust them with the knowledge of the fact that we use these RP
> services.  Just like most attacks, the probability that a provider (or
> employee thereof) will behave dishonestly is low (but certainly not
> negligible) --- but I should not have to take that chance.
>
> You might counter that we already put a lot of trust in organizations (that
> is, potential RPs) --- for example, banks --- so why shouldn’t we also trust
> an identity provider?  That is true, but there is a difference.  When I
> trust bank A, my trust is limited to the cash I give it and the credit
> accounts I have open with it.  When I trust bank B, my trust covers only the
> accounts I have with that bank.  When I trust Amazon.com, I trust it only
> with the information of one credit card.  When I trust my email provider, I
> trust them only with the email associated with that account.  The risk is
> limited in each case.  Now, imagine that I use OpenID to log into all four
> of those services: I am now trusting my identity provider with all four
> aspects of my life.
>
> Charles
>
>
> On 12/9/09 8:18 AM, "Nash, Andrew" <annash at paypal.com> wrote:
>
> Charles,
>
> your question moves past the security issues associated with the protocol
> and addresses trust issues in the ecosystem. As John indicated, at the
> current assurance levels and resulting appropriate usage scenarios, OpenID
> deployment and the trust levels associated with OP's are probably
> sufficient.
>
> The operating principals, policies controlling activities and personnel,
> audit trails and verifiable operation by third parties are all aspects to
> resolving the question you address. IFF you are dealing with low assurance
> credentials and low value transactions, then as David identifies it probably
> matters less and their are many alternative problems you could face from
> your ISP. However, for higher assurance levels, higher value transactions,
> more sensitive information or whatever, then you need to establish a trusted
> Identity Services Provider - hopefully using transparent mechanisms rather
> than just marketing spin, but heh.
>
> However, you should be aware that in the area you seem to be addressing,
> there are corresponding requirements and issues about how a relying party
> will treat the information that is provided to it and a need to verify (at
> higher assurance levels) that correct usage and protection of that
> information takes place. This is an issue that I raised several times in the
> context of the Federal Govt that Mary Rundle has taken up in the open trust
> frameworks discussion. For large scale operation this relying party
> assurance equivalent will not likely consist of audits, but probably will
> need to rely on relying party agreements (no pun intended :) ).
>
> The Kantara Initiative have done some good starting work on Identity
> Assurance models (based on earlier work by NIST et al). It needs lot more
> work and does not address a range of deployment challenges, but is worth
> taking a look at.
> --Andrew
>
>
>
> ------------------------------
> *From:* openid-security-bounces at lists.openid.net [
> mailto:openid-security-bounces at lists.openid.net<openid-security-bounces at lists.openid.net>]
> *On Behalf Of *John Bradley
> *Sent:* Tuesday, December 08, 2009 3:34 AM
> *To:* Shearer, Charles Dylan
> *Cc:* OpenID Security Mailing List
> *Subject:* Re: [security] Nonrepudiation, and Trusting OpenID Providers
>
> Charles,
>
> It is true that almost all assertion based protocols require that a RP and
> user have some trust in the OP/IdP.   This is equally the case for SAML and
> managed Info-Cards.
>
> Some thing like PKI and personal info-cards allow the user to have complete
> control over the authenticator.
>
> There are two basic options:
> 1 increase the trustability of the OP/IdP
> 2 Use multiple IdP simultaneously and prey.
>
> I don't personally believe that option 2 is all that practical or gives
> much more security for the average user.
>
> Given that openID is only secure enough as a protocol for ICAM LoA 1
> (pseudonymous protecting no PII) the most practical path is to provide more
> trustable OP/IdP.
>
> That said, with some of the v.Next changes openID will become appropriate
> for higher LoA.
>
> I don't think Gov or Banks are going to be comfortable with multi Auth
> solutions.  They are going to insist on trusted OP/IdP.
>
> You can have a look at the ICAM site to see where the US Gov is going.
> http://www.idmanagement.gov/drilldown.cfm?action=openID_openGOV
>
> I can see binding more than one openID to a RP to allow for recovery,
>  however that needs to be balanced against doubling the attack surface.
>
> Regards
> John Bradley
>
> On 2009-12-07, at 9:47 PM, Shearer, Charles Dylan wrote:
>
>
> I have some concerns about OpenID, and I would like to  see what those
> involved think about them.
>
> It seems to me that,  regardless of how OpenID is deployed, it is always
> possible for an OpenID  provider itself to authenticate with a relying party
> as any user by forging a  request to authenticate using the user’s
> identifier.  This is because a  relying party cannot tell the difference
> between a user attempting to log in  using his or her identifier, and the
> user’s OpenID provider spoofing that user  to gain access to whatever
> services the relying party provides to that user.   This seems to require
> that both users and relying parties put a lot of  trust in OpenID providers:
> for example, if I used my OpenID identifier for  online banking and email,
> my OpenID provider could easily access my email and  bank account.
>
> Additionally, even if we assume that OpenID  providers will not log into
> users’ accounts, I still cannot see how OpenID  could provide nonrepudiation
> regarding messages sent to a relying party by an  authenticated user: for
> example, if I authenticate with my bank using my  OpenID identifier and then
> use the bank’s “bill pay” service to pay a bill,  there’s no way the bank
> can prove that I ordered that payment because it is  possible that someone
> working for my OpenID provider logged in as me and  ordered it.
>
> Does anyone disagree with my  analysis?
>
> Dylan
> _______________________________________________
> security mailing  list
> security at lists.openid.net
> http://lists.openid.net/mailman/listinfo/openid-security
>
>
>
>
>
> _______________________________________________
> security mailing list
> security at lists.openid.net
> http://lists.openid.net/mailman/listinfo/openid-security
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openid.net/pipermail/openid-security/attachments/20091209/00465f51/attachment-0001.htm>