[security] security hole in signature algorithm

[Dick Hardt]

Actually, it is not a MITM attack. The real user has been replaced by  
a malicious user.

Per the rest of the thread, my understanding of the spec was  
incorrect and there is no vulnerability.

-- Dick

On 20-Nov-06, at 11:45 AM, James A. Donald wrote:

>     --
> Dick Hardt wrote:
> > Scenario ------------- By manipulating the return_to
> > parameter, an attacked can impersonate another user at
> > an RP.
> >
> > 1) Attacker goes to an RP and initiates an
> > authentication pretending to be http://victim.op.com.
> > RP sends the following parameters in the request:
> >
> > openid.return_to=http://rp.com/result?p=x
> > openid.identity=http://victim.op.com
> >
> > 2) The attacker modifies the parameters as such:
>
> This is a man in the middle attack.  To launch a man in
> the middle attack, one generally poisons DNS or routing
> tables, which can be done, and often is done, but is not
> that easy or that common.  One can also launch a man in
> the middle attack by interfering with the physical line
> coming from a particular home, or a particular
> institution.
>
> It is reasonable to provide security that fails to
> protect against a man in the middle attack, for such
> attacks are not the main threat, but the threat needs to
> be noted, and should be defended against if one can find
> a low cost way of doing so.
>
>
>
>     --digsig
>          James A. Donald
>      6YeGpsZR+nOTh/cGwvITnSR3TdzclVpR0+pr3YYQdkG
>      Feg5WSMP+ZF9MK8mPWm1MPQga0yF7ZwzeRCB9rmH
>      4WstoiW2104UktEffclaA6DuFGBVF5+Pv9KN9JK/r
>
>