[Pki-users] Best way to bypass challenge password verification

[Fraser Tweedale]

On Thu, Dec 04, 2014 at 10:36:10AM -0500, Peter Beal wrote:
> Hi Fraser,
> 
> Thanks for the quick response.
> 
> I saw the ticket get opened.  Thanks for jumping on that as well.
> 
> What I see at our EST based client as it is building the CSR is the
> following value being used for the challenge password
> 
> (gdb) p tls_uid
> $35 = 0x7ffff001a4d0 "H9wOU8vX9HX5s/hH"
> (gdb)
> 
> We take the TLS UID and run it through a base64 encoding and the above
> string is what is placed in the challenge password field. This appears to be
> a printable string for us as it's placed into the CSR.  We verify the CSR at
> the EST server and then send it on to Dogtag unmodified.  I haven't go into
> the RA or Dogtag to verify this, but I suspect this string is still set to
> this value when it arrives at Dogtag.
> 
> Thanks again for you help,
> Pete
> 

The DER data in the Dogtag log output shows that the
challengePassword attribute value has UTF8String encoding (tag
0x0C), not PrintableString (tag 0x13).  Dogtag MUST be able to parse
UTF8String challengePassword attributes, but according to the RFCs
the value SHOULD have PrintableString encoding, so you may want to
look into it.

I have a patch for the decoding issue in the works; should get it to
the pki-devel list for review this morning.  What OS and Dogtag
version are you running?  Hopefully we can get that patch in the
next Dogtag release for your distro.

Regards,

Fraser

> On 12/3/14 10:44 PM, Fraser Tweedale wrote:
> >On Wed, Dec 03, 2014 at 02:16:47PM -0500, Peter Beal wrote:
> >>Hello,
> >>
> >>Our group is developing its own RA to be part of a PKI solution that
> >>includes Dogtag.  Our solution uses EST (RFC 7030) between the the clients
> >>and the RA, with the EST protocol terminating at the RA. The RA then takes
> >>the CSRs and sends them on to Dogtag using REST commands.  This project has
> >>been going very well, however we did just hit one issue that we were hoping
> >>others might be able to provide some guidance on.
> >>
> >>The EST protocol defines a feature called Proof Of Possession (PoP) where
> >>the clients insert the TLS unique ID of the TLS session between it and the
> >>EST server, in our case our RA. This TLS UID is sent in the challenge
> >>password attribute field of the CSR so that it can be signed by the client.
> >>The EST server is responsible for verifying this TLS UID, and once this
> >>verification is performed the value in the challenge password field no
> >>longer has any meaning. Because the CSR cannot be resigned at this point,
> >>the challenge password cannot be taken out of the CSR.  This CSR is passed
> >>as is along to Dogtag and we're currently finding that Dogtag is checking
> >>the CSR and does not like the challenge password attribute:
> >>
> >>[03/Dec/2014:13:37:59][http-bio-8444-exec-3]: Start parsePKCS10():
> >>MIIBdDCB3gIBADAUMRIwEAYDVQQDDAkxMjcuMC4wLjEwgZ8wDQYJKoZIhvcNAQEB
> >>BQADgY0AMIGJAoGBAMPNKtwZO82WfR5u/hS+SsVghdS9jD5BQS6Z5ymXrKcr9R4t
> >>DtSeEQ+AtCZ5qVNXcangb00vJgVqgS/7NH4MzSqscgNzZdpx9+mPklvOUuqTuYCv
> >>MFIlMwP/2DJ6TBrmF86vJ1I0GmZAyTSzHg4V4YWaN7r0V7x0RyvFqoBnZU51AgMB
> >>AAGgITAfBgkqhkiG9w0BCQcxEgwQTUR4clVZNmd6TnZqdWRmNzANBgkqhkiG9w0B
> >>AQsFAAOBgQAAKLbWGndYFfa+8IhopufYOEKIOAmcT+Nhr27vFt5I4ymoUwSlKX9L
> >>K+KpLho5Q2GsRoItNXJ6VxRcGe1CPZBW2ef7yPdZaKhFmnxXsYVQaqPY5BGI8kAY
> >>MMMr75WQcpn+XUpu+FNB4F2j8YY314u2rsplCMbOdR4tcrgc8WqucA==
> >>
> >>[03/Dec/2014:13:37:59][http-bio-8444-exec-3]: EnrollProfile: parsePKCS10:
> >>signature verification enabled
> >>[03/Dec/2014:13:37:59][http-bio-8444-exec-3]: EnrollProfile: parsePKCS10:
> >>use internal token
> >>[03/Dec/2014:13:37:59][http-bio-8444-exec-3]: EnrollProfile: parsePKCS10
> >>setting thread token
> >>[03/Dec/2014:13:37:59][http-bio-8444-exec-3]: EnrollProfile: parsePKCS10
> >>java.io.IOException: DerValue.getPrintableString, not a string 12
> >>[03/Dec/2014:13:37:59][http-bio-8444-exec-3]: EnrollProfile: parsePKCS10
> >>restoring thread token
> >>
> >>So we're wondering what the best approach might be to handle this. Is there
> >>a way to configure Dogtag so that it will ignore the challenge password?
> >>
> >>Thanks very much,
> >>Pete Beal
> >>
> >Hi Pete,
> >
> >Dogtag tries to decode the request object as a whole (and is
> >failing; see below).  There is no way to bypass this.
> >
> >PKCS #9 (RFC 2985) §5.4.1 "Challenge password" states that "PKCS #9-
> >attribute processing systems MUST be able to recognize and process
> >all string types in DirectoryString values."  Therefore, this is a
> >bug in Dogtag, and I will file a ticket.  (Incidentally, OpenSSL
> >also fails to deal with with this part of the request object, but
> >the problem is isolated to the attribute.)
> >
> >The problem should not present when using PrintableString instead of
> >UTF8String.  Since according to RFC 7030 (EST) the value of the
> >challengePassword attribute is base64 encoded, it is printable, and
> >furthermore RFC 2985 also states, "ChallengePassword attribute
> >values generated in accordance with this version of this document
> >SHOULD use the PrintableString encoding whenever possible."  If you
> >have control over the clients (I understand you may not) then using
> >PrintableString instead of UTF8String is recommended and should get
> >you over the line with Dogtag until the decoding issue is fixed on
> >our end.
> >
> >Cheers,
> >
> >Fraser
>