Assurance

[Eric Goodman <>]

The difference around FIDO (and SQRL, for that matter) is that there's 
basically a two-way validation happening. The private key generated by the 
FIDO/SQRL client is based on identity information from the server. At least 
in SQRL, the URL of the site being authenticated to also comes into play.

As I understand it, it's sort of like a SAML metadata validation, in that the 
acceptable endpoints of an authentication message for a specific 
service/target are pre-defined. It's unlike metadata in that there's no out 
of band exchange of the endpoints, (it's instead part of the account setup, 
defined when the new service-specific private key is created).

If a MiTM gets iTM, it will look like a new service/endpoint, and so the 
authentication assertion the MiTM gets from the user won't match up to the 
assertion the target service needs. (And the user should be alerted that 
he/she is creating a new account, not authenticating with an existing one).

That's at least how I understand the SQRL version of the technology, I think 
FIDO is relatively similar, but I'm doing some guessing here.

--- Eric

> -----Original Message-----
> From: 
> 
>  
> [mailto:]
>  On Behalf Of Tom
> Scavo
> Sent: Thursday, April 17, 2014 1:40 PM
> To: 
> 
> Subject: Re: [Assurance] can two-factor be hacked ?
> 
> On Thu, Apr 17, 2014 at 4:06 PM, David Walker 
> <>
> wrote:
> >
> > I was just reading the FIDO Alliance's U2F overview
> > (https://fidoalliance.org/specs/fido-u2f-overview-v1.0-rd-
> 20140209.pdf
> > ), and I think they have a solution to this issue.  They use public
> > key cryptography, though not PKI, for mutual authentication of the
> > service and the token, which (I think) should defeat this kind of
> > man-in-the-middle vulnerability.
> 
> Duo Push uses public key crypto as well:
> 
> https://www.duosecurity.com/blog/heartbleed-defense-in-depth-part-2
> 
> but that still doesn't prevent a MiTM (as discussed earlier in the
> thread). Doesn't U2F require a button push by the user? If so, then how
> does it avoid this problem?
> 
> Tom