[VoiceOps] SHAKEN testing services

Nathan Anderson nathana at fsr.com
Tue Jun 28 18:59:50 EDT 2022 

I could swear that I saw examples of decoded PASSporT JWTs where "orig" and "dest" fields had "+1" prefixed to the NANP number, but darned if I can find them now...

Reading through RFCs 8224 and 8225 again, though example numbers referenced outside of a fully-formed example JSON are all + prefixed, the actual numbers within the JSON are not (e.g., see: RFC 8225, 5.2.1.4).

I also just spotted this language in RFC 8224 section 8.3:

  *   Implementations MUST drop any "+"s, internal dashes, parentheses, or other non-numeric characters, except for the "#" or "*" keys used in some special service numbers (typically, these will appear only in the To header field value).  This MUST result in an ASCII string limited to "#", "*", and digits without whitespace or visual separators.
  *
  *   Next, an implementation must assess if the number string is a valid, globally routable number with a leading country code. If not, implementations SHOULD convert the number into E.164 format, adding a country code if necessary; this may involve transforming the number from a dial string (see [RFC3966]), removing any national or international dialing prefixes or performing similar procedures.

So I guess the implication is that since the number is assumed to be global at the outset, the + is unnecessary (as well as actually against spec)?

Talking to Carlos about this, apparently some STI-VS implementations do allow for the +.  At the moment, it so happens that Twilio is back-end-ing the Sansay test # (though this is subject to change in the future), and perhaps they just don't.  I did find it odd that the various test services in question would consider a malformed (by spec) PASSporT to simply be unverified rather than report that the verification failed, but also found some Twilio documentation on their own implementation and what the possible causes for certain results might be.  Taken from https://www.twilio.com/docs/voice/trusted-calling-with-shakenstir#changes-for-twilio-customers --
'No-TN-Validation' Possible causes:

  *   A malformed E.164 phone number
  *   [...]
  *   SHAKEN PASSporT orig field doesn't match with actual callerid in the SIP messages (P-Asserted-Identity, Remote-Party-Identity, or From header).
  *   SHAKEN PASSporT dest field doesn't match with the actual destination of the call in the SIP Request-URI.
  *   [...]

Now, I would argue that many of the causes that they list in their documentation that can lead to this result should actually be considered failed validations instead; however, I guess a case could be made that if a PASSporT is signed with a cert that has a proper chain-of-trust back to a legit STI-CA & shows no signs of tampering, but otherwise does not contain the proper 'orig' or 'dest' numbers, that it is not strictly speaking an invalid PASSporT, just one that merely doesn't actually speak to the call in question that it's attached to.

I suppose it's also possible that this STI-VS implementation may allow for + prefix in the phone numbers within the PASSporT contents, but engages in strict matching of 'orig' and 'dest' with the TNs in the SIP headers; that is to say, if you have the + in the PASSporT but NOT in From: and/or P-A-I:, but the TNs are otherwise the same, then it still technically & literally "doesn't match".  Further testing would be needed to determine if maybe this was actually why our PASSporTs were returning this result.

-- Nathan

________________________________
From: VoiceOps <voiceops-bounces at voiceops.org> on behalf of Jared Geiger <jared at compuwizz.net>
Sent: Tuesday, June 28, 2022 3:05 PM
Cc: Voiceops.org <voiceops at voiceops.org>
Subject: Re: [VoiceOps] SHAKEN testing services

I've also been unsure how the ANI should be encoded in the Identity header. Before S/S, most termination vendors wanted the CallerID to be in 10 digit form without a leading +1. Only a couple vendors could support e164 format with a leading + or +1.

On Tue, Jun 28, 2022 at 2:14 PM Nathan Anderson <nathana at fsr.com<mailto:nathana at fsr.com>> wrote:
Following up here to let you all know that after consulting with Carlos, the only obvious difference that could be seen between our PASSporTs that were not being validated by both the Sansay & IDT tools and the ones being generated by our upstreams that were being validated, was that we were formatting the origination and destination numbers in globalized E.164 format within the PASSporT JWT, complete with the + prefix, while the others included the U.S. country code of 1 but no + prefix.

I removed the +, and instantly both the Sansay and IDC tools could decode and verify our PASSporTs.

Much thanks to Carlos and Sansay for the assist in hunting this down.

-- Nathan
________________________________
From: VoiceOps <voiceops-bounces at voiceops.org<mailto:voiceops-bounces at voiceops.org>> on behalf of Nathan Anderson <nathana at fsr.com<mailto:nathana at fsr.com>>
Sent: Tuesday, June 28, 2022 1:24 PM
To: Dovid Bender <dovid at telecurve.com<mailto:dovid at telecurve.com>>
Cc: Voiceops.org <voiceops at voiceops.org<mailto:voiceops at voiceops.org>>
Subject: Re: [VoiceOps] SHAKEN testing services

Dovid,

Our carriers swear up and down they are passing the Identity header we are sending to them; I opened tickets with them to verify.  And that if we do not send an Identity header, both of them will attest the call themselves.

At this point, I also have no reason to doubt them, because we are getting identical results across two different carriers, and upon further testing, what I've found is that if we *don't* send our own Identity header, our calls pass validation with the PASSporT that our carriers generate & send.  It's when we send our own that we get the "No-TN-Validation" result back from the Sansay test number.  Which now makes me suspect that there must be something wrong with what we're doing (even though that Voxbeam number that I referenced in my prior response has no problem reading back our attestation level to us).  I just don't know *what* is wrong, because none of these testing services actually give back detailed info about the mode of failure in the case of a problem.

I've reached out to Carlos/Sansay to see if they can pull logs on their side in order to tell me what they're seeing when we call their number.

You also implied that when you call the Sansay test number, you actually get back an attestation level from it.  That also has not been our experience: the Sansay number tells us whether the TN validation passed, failed, or is simply unvalidated...it doesn't read back the attestation level if it passes.  So I'm not sure why your experience is different there, unless you mistakenly wrote "Sansays number" when you meant something else (like "Voxbeam number" maybe).

-- Nathan

________________________________
From: Dovid Bender <dovid at telecurve.com<mailto:dovid at telecurve.com>>
Sent: Tuesday, June 28, 2022 3:18 AM
To: Nathan Anderson <nathana at fsr.com<mailto:nathana at fsr.com>>
Cc: Carlos Perez <cperez at sansay.com<mailto:cperez at sansay.com>>; Voiceops.org <voiceops at voiceops.org<mailto:voiceops at voiceops.org>>
Subject: Re: [VoiceOps] SHAKEN testing services

Nathan,

Could it be your carriers that are dropping the headers? We have tested with a self signed cert and expected to have issues across the board. I don't want to name shame so overall using a few carriers to terminate calls to call Sansays number this is what we found
- For some carriers regardless of what we sent attestation always seemed to have come back as B.
- For others where we send a self signed cert the attestation always comes back as C.
- For the carriers where if we send a self signed cert it comes in as C,  if we send nothing the call comes in as B.
- Looking at our apache logs we mainly see Verizon wireless getting our certs (could be others are caching?).  Surprisingly #2 is Telnyx followed by random requests from AWS, Google and MS IP's.

On Mon, Jun 27, 2022 at 7:47 PM Nathan Anderson <nathana at fsr.com<mailto:nathana at fsr.com>> wrote:
Carlos,

Nifty; thanks.  Unfortunately, I tried calling via 2 different carriers who both pass through Identity headers if supplied, but got No-TN-Validation result both times, implying that the PASSporT is getting stripped out somewhere along the way.  Your number looks to be supplied by Level3, so a bit surprised at this result since I know for a fact both of the services I tried have direct relationships with L3 & so I would think that they'd look-up LRN & then try to send the call direct to them rather than punt it to their LCRs, but *shrug*​.

I don't have direct access to L3/Lumen term...any chance I can send the INVITE directly to an SBC on your side to just bypass the problematic path(s)?  (I presume I'd have to supply you with an IP on our side for you to whitelist...)

If I get a failed validation result from the service, is it possible for somebody at Sansay look up on your side what specific part of the validation failed for a given call?

At the moment, here is where I stand:

Found a Voxbeam number (302-257-7304) that will read back attestation level.  This will read back exactly the attestation level that I supply in the PASSporT that I attach to the call, but it "works" regardless of whether I'm using self-signed testing cert or an actual, valid SHAKEN cert.  So this service doesn't actually seem to be validating anything.  At least I know that Identity header is making it all the way there & that it's encoded properly...

IDT's service, on the other hand, will show that validation passed and the correct attestation level if I do not transmit my own Identity header, and instead have upstream carrier attest it themselves.  However, if I send our own Identity header using *either* our self-signed cert *or* our Sansay SHAKEN cert, I get "N/A - problematic" as the result, with no other explanation.

If the Sansay testing service ends up having no problem validating our PASSporTs, that will be reassuring from one perspective, but concerning from another (why can IDT not validate any of our calls that we sign ourselves using our production cert, but Sansay can?).  So I'm kinda rooting for the Sansay test service to fail when validating our calls...

-- Nathan

________________________________
From: Carlos Perez <cperez at sansay.com<mailto:cperez at sansay.com>>
Sent: Monday, June 27, 2022 3:37 PM
To: Nathan Anderson <nathana at fsr.com<mailto:nathana at fsr.com>>
Cc: Voiceops.org <voiceops at voiceops.org<mailto:voiceops at voiceops.org>>
Subject: Re: [VoiceOps] SHAKEN testing services

Hi Nathan,

We've put together something like you've requested (also open to the public) recently.
In the future we will also provide the SPID in the report, right now you get the status of the validation (Passed, Failed, No Validation) and the attestation level. Feel free to dial +18586098097.

Regards,

Carlos Perez
Sansay, Inc.
+1 858-754-2216 Direct
+1 858-754-2211 Support
+1-858-754-2200 Main

https://calendly.com/cperez-sansay/30min


On Mon, Jun 27, 2022 at 3:34 PM Nathan Anderson <nathana at fsr.com<mailto:nathana at fsr.com>> wrote:
Is there a list somewhere of third-party SHAKEN testing services available to the public?  Although our own STI-VS can validate our own PASSporTs signed by our newly-minted private key, I want to verify what the world at-large is seeing before I flip the switch on for all outgoing calls (& potentially cause a bunch of problems if it turns out that something's wrong).  So I'm envisioning something along the lines of a phone # one can call, and then get a report back of whether the terminating provider was able to validate the contents of the Identity header or not, and if not what the exact problem is.  (Obviously the path between us and them would need to be IP end-to-end unless out-of-band SHAKEN is in place for the PSTN "glue" circuit.)

I've found a couple, but the results for the ones I've tried (e.g., IDT BYOC) are woefully simplistic and don't really explain what went wrong if something does go wrong.

Thanks,

-- Nathan
_______________________________________________
VoiceOps mailing list
VoiceOps at voiceops.org<mailto:VoiceOps at voiceops.org>
https://puck.nether.net/mailman/listinfo/voiceops
_______________________________________________
VoiceOps mailing list
VoiceOps at voiceops.org<mailto:VoiceOps at voiceops.org>
https://puck.nether.net/mailman/listinfo/voiceops
_______________________________________________
VoiceOps mailing list
VoiceOps at voiceops.org<mailto:VoiceOps at voiceops.org>
https://puck.nether.net/mailman/listinfo/voiceops
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://puck.nether.net/pipermail/voiceops/attachments/20220628/f2affb1c/attachment-0001.htm>

More information about the VoiceOps mailing list