Hi Colm,
When I looking at the krb5 source code, I found the function
cms_signeddata_verify in pkinit_crypto_openssl.c with the following comments:
" if (((si_sk = CMS_get0_SignerInfos(cms)) == NULL) ||
((si = sk_CMS_SignerInfo_value(si_sk, 0)) == NULL)) {
/* Not actually signed; anonymous case */
if (!is_signed)
goto cleanup;
"
When the client parsing PA-PK-AS-REP message, it will call
cms_signeddata_verify function. So my point from here.
But what you said let me doubt myself, I will take some time to dig into this
issue.
Thanks
Jiajia
-----Original Message-----
From: Colm O hEigeartaigh [mailto:[email protected]]
Sent: Wednesday, July 27, 2016 8:59 PM
To: [email protected]
Subject: Re: Anonymous PKINIT signatures
Hi Jiajia,
It's the client that's anonymous here, and not the KDC. This page leads me to
believe that the KDC does in fact sign the response to the client:
http://web.mit.edu/kerberos/krb5-devel/doc/admin/pkinit.html
" For anonymous PKINIT, a KDC certificate is required, but client certificates
are not.".
"The result of this operation will be in two files, kdckey.pem and kdc.pem.
Both files must be placed in the KDC’s filesystem. kdckey.pem, which contains
the KDC’s private key, must be carefully protected."
Colm.
On Tue, Jul 26, 2016 at 3:08 AM, Li, Jiajia <[email protected]> wrote:
> Hi Colm,
> >> However, the client doesn't use the certificate to verify a
> >> signature,
> and thus proving that the KDC knows the private key associated with
> the cert. Is this correct?
> You are right. I think anonymous case, not actually signed.
> Thanks,
> Jiajia
>
>
> From: Colm O hEigeartaigh [mailto:[email protected]]
> Sent: Friday, July 22, 2016 11:22 PM
> To: Li, Jiajia <[email protected]>
> Cc: [email protected]
> Subject: Re: Anonymous PKINIT signatures
>
> Hi Jiajia,
> So if I understand you correctly, what you are saying is that it is
> sufficient to verify that the Subject (alternative name) of the
> Certificate matches that of the "known principal" of the KDC? In other
> words, the KDC is not doing any asymmetric signature, it is just
> "presenting" the certificate to the client. The client verifies that
> the certificate is trusted, and then verifies that the KDC principal matches
> the certificate.
> However, the client doesn't use the certificate to verify a signature,
> and thus proving that the KDC knows the private key associated with the cert.
> Is this correct?
> It's a bit unusual from a security POV but I think it's ok. We're
> verifying trust in the certificate path and we're putting a hard
> constraint on the Subject of the certificate. A malicious KDC/MITM
> could forge a certificate, but then trust validation would fail, or
> else get a certificate for another KDC, but then the constraint would
> fail. So I think it's ok.
>
> Colm.
>
> On Fri, Jul 22, 2016 at 3:40 AM, Li, Jiajia <[email protected]<mailto:
> [email protected]>> wrote:
> Hi Colm,
> >> >However, I can't see where it is signing the response with the
> >> >private
> key associated with the KDC. This is a requirement for anonymous
> PKINIT
>
> Yes, you are right. The "Identity" should be used in anonymous PKINIT.
> But now in client PkinitPreauth, start from line 393, we skip to use
> the certificateSet which is returned by server, so now the code can't
> verify the kdc sans, edu and so on. Such as the function
> cryptoRetrieveX509Sans#PkinitCrypto is marked as TODO.
>
>
> Thanks
> Jiajia
>
>
> -----Original Message-----
> From: Colm O hEigeartaigh [mailto:[email protected]<mailto:
> [email protected]>]
> Sent: Thursday, July 21, 2016 7:27 PM
> To: [email protected]<mailto:[email protected]>
> Subject: Anonymous PKINIT signatures
>
> Hi all,
>
> I'm continuing to look at anonymous PKINIT as implemented in Kerby.
> I'm a bit puzzled by a few things relating to signatures and would
> welcome some feedback.
>
> Looking at the server PkinitPreauth, it appears that Diffie-Hellman is
> used to establish a shared secret key with the client. However, I
> can't see where it is signing the response with the private key
> associated with the KDC. This is a requirement for anonymous PKINIT, unless I
> am mistaken?
>
> Similarly, on the client side, it's not enough just to verify trust in
> the Certificate that's presented, it also needs to be using the
> Certificate to verify some signed data, to make sure that the KDC
> knows the private key associated with the Certificate...
>
> I've updated the code so that the server at least includes the "Identity"
> Certificate in the response to the client.
>
> Thanks,
>
> Colm.
>
>
> --
> Colm O hEigeartaigh
>
> Talend Community Coder
> http://coders.talend.com
>
>
>
> --
> Colm O hEigeartaigh
>
> Talend Community Coder
> http://coders.talend.com
>
--
Colm O hEigeartaigh
Talend Community Coder
http://coders.talend.com
