SITR/SHAKEN implementation in effect today (June 30 2021)

Michael Thomas mike at
Fri Jul 9 23:33:15 UTC 2021

On 7/9/21 3:32 PM, K. Scott Helms wrote:
> On Fri, Jul 9, 2021 at 4:47 PM Michael Thomas <mike at 
> <mailto:mike at>> wrote:
>     On 7/9/21 1:36 PM, K. Scott Helms wrote:
>     > Nothing will change immediately.  Having said that, I do expect
>     that
>     > we will see much more effective enforcement. The investigations
>     will
>     > come from the ITG (Industry Traceback Group) with enforcement
>     > coming from FCC or FTC depending on the actual offense.  The
>     problem
>     > has been that it's been far too easy for robocalling companies
>     to hop
>     > from one telecom provider to another.  Now there are requirements
>     > around "know your customer" that telecom operators have to
>     follow and
>     > the ITG will have a much better chance of figuring out who the bad
>     > actor is than they have in the past.
>     The thing is that that shouldn't have been held up by rolling out
>     STIR.
>     With email, there was nothing akin to the FCC so it was really the
>     only
>     name-and-shame stick we had. This could have been done years ago.
> It wouldn't work the same and I say that as someone who ran email for 
> ISPs for decades and just got done with a STIR/SHAKEN implementation.  
> There's far more money in robocalls than there ever has been in spam.  
> The other thing is that the technologies are fundamentally different.  
> Don't get me wrong, there are parallels, but comparing email to the 
> various flavors of telephony (POTS, SIP, MGCP, H.323, etc) isn't all 
> that useful because they're so different. When I get an email as a 
> provider I can always figure out the path it took to get to me.  The 
> same is not at all true for a call, though I can trace it to a degree 
> via the CDRs from carriers I work with.  Much of the call path will be 
> opaque and often in the case of robocallers it's intentionally so.  
> Number porting is another (big) difference.  We could always choose to 
> forward email for a customer who left our service, but imagine if 
> email literally let that person take their address with them 
> regardless of who was providing the hosting for the email.
Once it hits a SIP gateway it's pretty much the same. The thing that I 
think that made the made the biggest -- and this coming from one of the 
inventors of DKIM -- was shutting down open relays. With email there was 
no back pressure on ESP's to close them so DKIM was at least a way to 
name and shame, or at least that was one of the goals at least in my 
mind. It's hard to say whether there was actually cause and effect, but 
the reality now is that open relays are pretty much gone -- at least 
with ESP's.

I was one of the early Cassandras telling people that 
P-Asserted-Identity was going to lead to exactly what has happened for 
which I got told I was wrong, and then threw up my hands and went and 
worked on email. This could have been dealt with 15 years ago and it 
could have trivially piggybacked off of the DKIM work -- heck I even 
hacked a SIP stack to prove the point. And it should have learned the 
lessons with email which it apparently has not because the 9th of July 
don't seem any different than the 30th of June.

Also: since there are PSTN gateways which fundamentally can't be secured 
spammers will take advantage of the holes as they become economically 
viable. The entire scheme of attesting for e.164 addresses was a 
complete waste of time. The problem is with the SIP gateway that onramps 
the bogus calls not whether somebody should be able to assert a given 
address real time. Those onramps could have taken those measures a 
decade ago but didn't just like the open email sewers before requiring 
submission auth. I think that Peterson has some other wacky shit to make 
PSTN gateway traversal better, turning a Rube Goldberg contraption into 
something even worse. The entire thing is madness with its complexity 
but I would expect nothing less from the SIP WG.

>     > Longer term I worry that this will lead to more attacks on PBXs,
>     > eSBCs, and VOIP handsets to be able to call either from that
>     endpoint
>     > itself or be able to use the SIP credentials. The market for
>     robocalls
>     > will certainly not disappear.
>     >
>     A meta question that really needs to be asked these days is why we
>     even
>     need telco telephony anymore. A lot of problems go away if you are
>     not
>     in thrall to 100 year old technology and its accreted kruft.
> Robocalls really aren't a product of the legacy PSTN. Today almost 
> none of them originate from anywhere but VOIP. Now, you can certainly 
> say that if SS7 had robust authentication mechanisms that we could 
> then trust caller ID (more) but there's no sign of us abandoning the 
> PSTN anytime soon.  Having said that, there's any number of protocols 
> we rely on today that have the exact same gap.  BGP is arguably even 
> worse than SS7.

I'm not speaking about PSTN qua PSTN/SS7, etc. I'm speaking about being 
shackled to an architecture that isn't relevant today. The 
bellheadedness of STIR is frightening: they actually scrape SIP From: 
addresses and have a regex to determine if it's a telephone number. 
People don't use telephone numbers for anything but phones, and all of 
the rest is email addresses. There is fundamentally nothing stopping SIP 
From: mike at be how reach me. STIR leaves that use case in limbo: 
I've never been able to get a straight answer about it, which means in 
reality it's not supported.  I'd actually like that quite a bit more 
especially since it would make it difficult right off the bat to spoof 
my domain vs. spoofing area codes. Doubly so for spear phishing. We 
already know how to do federation: email. There is just no justification 
for keeping the legacy PSTN kruft, and even worse wagging the dog.

Mike, Cassandra to the end

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the NANOG mailing list