There's no decryption here: it's an HMAC. Think of it as a magic token that's 
irrevocably bound to the content. Since its creation involves a secret, nobody 
else can forge it. It's almost sorta kinda like a cleartext password, only not, 
since obviously a cleartext password would be useless.

Is it better than what you provide? Not sure, quite possibly not. What it would 
definitely provide is something that crypto-types would understand and trust 
more than "Really, honestly, we're sure you can't forge email headers". Nobody 
trusts email much (well, everybody does because have to, but we don't WANT to) 
and so crypto types aren't likely to be thrilled with "Sure, just send email in 
this format to this address and lo! it will execute the embedded command(s)".

OpenSSL will generate HMACs easily:
echo -n "the command" | openssl dgst -sha256 -hmac "the secret"
(pipe that to wherever as appropriate).

Admittedly this adds an extra step, although, as noted earlier, an Outlook 
macro could probably automate it. But end-users would need at least OpenSSL and 
that macro. 

One thing we don't know (not blaming ya, it hasn't been particularly relevant) 
is what the footprint of this facility is: is there software installed on the 
endpoints already? Or is the idea that Joe Sysprog is at a BBQ and needs to hup 
some STC or submit some job and wants to do it via a quick email from his 
phone? That latter capability would be way cool when needed--but only if it's 
provably secure. Otherwise any security person worth their salt will say "No 
way are we enabling that!" There's a lot of appeal to making such things 
lightweight, of course; I'm a big fan of "As few parts as is practical". Or as 
Einstein may or may not have said, "As simple as possible, but no simpler"!


ObAnecdote, only peripherally related: over 40 years ago, we had a service 
virtual machine on the University of Waterloo VM system called PRIV. You could 
send it an SMSG with a command and it would look up your userid and the command 
and see if you were allowed to execute that command. This gave us the ability 
to provide VERY granular PRIVileged command execution: "BOB can FORCE BOBSVM 
but can't FORCE any other IDs", for example.

One day I was updating PRIV with a new version. I logged onto the SVM and there 
I was in CP READ. "Hmm", thinks I, "how do I shut this thing down?" So I did: 
SMSG * SHUTDOWN

You've guessed where this is going...about 20 seconds later, one of the 
operators comes running out of the Red Room (the main UofW data center; I was 
in a sysprog terminal room adjacent) yelling "System A just shut itself down!" 
and I said "Oops".

Ever since then, I've been very sensitive to control of privileged commands, 
and have never created a service that used the same syntax to shut it down as 
the system. (Although the problem that solves is actually fractionally 
different: had SHUTDOWN been the command to tell PRIV to end, it would have 
been fine; of course nobody could have been authorized to issue the real system 
SHUTDOWN via PRIV. But the point is to not train users to issue that command, 
lest they issue it in the wrong environment and it go to the system instead of 
the service.)

-----Original Message-----
From: IBM Mainframe Discussion List <[email protected]> On Behalf Of 
Brian Westerman
Sent: Tuesday, August 25, 2026 1:24 AM
To: [email protected]
Subject: Re: Odd idea?

So, when the original email was sent to the email address, I should include 
some encrypted code package, that they would have to combine with something 
(like maybe their mainframe password) which, when received could be decrypted 
and passed to SAF to verify them?   I don't see how that would be any more 
secure than what I already provide.  Now, if they had some software on their 
end that would generate some variable key then I think that would add some 
security, but that would mean they would have to have that generator.  Although 
I could probably code it so that they would need a passkey that they would need 
to touch to validate them.     

I still don't think it would be necessary, but I suppose it would be fairly 
simple to code.

On Mon, 24 Aug 2026 12:19:49 -0400, Phil Smith III <[email protected]> wrote:

>You're NOT encrypting the email--I thought I made that clear. You're adding a 
>keyed hash that lets you validate the command (and also prevents tampering in 
>flight).
>
>So if QUIESCE + <secret> = HMAC x, then the receiving end takes QUIESCE + 
><secret> and regenerates the same HMAC. If they don't compare, then either the 
>sender doesn't know the secret, doesn't know how to generate an HMAC, or the 
>command was altered after creation.
>
>Of course you have to share that secret somehow initially, but that's 
>presumably a one-time (or at least infrequent) operation, performed 
>out-of-band.
>
>-----Original Message-----
>From: IBM Mainframe Discussion List <[email protected]> On 
>Behalf Of Brian Westerman
>Sent: Monday, August 24, 2026 2:19 AM
>To: [email protected]
>Subject: Re: Odd idea?
>
>I don't trust things either, but encrypting the email is requiring a lot that 
>the user might not have access to or really even need to understand to use the 
>interface, and if things are set up correctly they would be largely 
>unnecessary.  If the site did not set things up correctly, that would show up 
>in the authentication processing and the email would be discarded anyway.  
>Encrypting it just adds another layer of complexity that doesn't actually seem 
>to make it any more secure. 
>
>Brian
>
>On Sun, 23 Aug 2026 14:11:35 -0400, Phil Smith III <[email protected]> wrote:
>
>>Well, it sounds like you're thinking this is "just" email, which is fine and 
>>certainly easy. The messageID is a thing, but can almost certainly be forged. 
>>The point of an HMAC isn't that you encrypt the whole email: you'd have a 
>>tool that knows a secret (password) and takes the command itself and creates 
>>the HMAC. So if the command in the email body looked like (say):
>>
>>command: submit some.library.jcl(member)
>>
>>then the user would past THAT into the tool, enter the password, and the tool 
>>would produce a string such as ABCDEF12345678 (the HMAC), which would be 
>>pasted into the body, perhaps:
>>
>>command: ABCDEF12345678 submit some.library.jcl(member)
>>
>>Extra steps for sure. A clever Outlook macro could automate this so it's a 
>>keystroke, though--user enters the command, hits keystroke, HMAC gets 
>>generated and inserted.
>>
>>Agreed, not clear it's necessary. Depends how well you can trust the 
>>environment, of course. With my security hat on, I *don't* trust things...
>>
>>-----Original Message-----
>>From: IBM Mainframe Discussion List <[email protected]> On 
>>Behalf Of Brian Westerman
>>Sent: Sunday, August 23, 2026 1:01 AM
>>To: [email protected]
>>Subject: Re: Odd idea?
>>
>>Hmmm.  I think that is probably overkill.  That would mean that I would need 
>>to make the user encrypt the command package and copy or attach it to the 
>>email.  I already check and keep the email messageid that looks like 
>>[email protected], and it would be pretty hard 
>>to spoof that and have it be valid.  But I'm willing to listen to another 
>>method if it would make things more secure, but encrypting the commands, 
>>while easy to handle, would be hard on the user to remember to do it and get 
>>it right.  I suppose we could assign them a key, but then if they lose the 
>>key, it would be just as bad as losing their password.  
>>
>>I think that a lot of the use will be more on the order of: 
>>
>>TASK A wants to do something, so an email gets sent to some user asking if 
>>it's okay to do "something", what is your response (yes or no).  The user 
>>types "yes" or "no" and sends the email.  
>>
>>I think there will be a lot of "user wants to know if a task executed yet, or 
>>if its running or what the status is", so they send the query to automation 
>>and it responds back with the status of the task or the output or whatever it 
>>was that they asked about it.
>>
>>I think that systems programmers will probably want to use it to see parts of 
>>syslog or check on things or start processes/tasks, etc.
>>
>>Brian
>>
>>
>>
>>On Sat, 22 Aug 2026 19:53:21 -0400, Phil Smith III <[email protected]> wrote:
>>
>>>RFCs.
>>>
>>>Really Clever(tm) spammers even fake some Received: headers, making 
>>>it harder to validate.
>>>
>>>How about an HMAC of the command itself? Still DOS-attackable but at 
>>>least you won't be executing bad commands.
>>>
>>>On August 22, 2026 7:41:59 PM Charles Mills <[email protected]> wrote:
>>>
>>>> Thanks for the link. Any umm, you know, documentation?
>>>>
>>>> CM
>>>>
>>>> On Sat, 22 Aug 2026 23:21:27 +0000, Seymour J Metz <[email protected]> wrote:
>>>>
>>>>> The From: header field is worthless, and the envelope from can 
>>>>> also be spoofed; include a sanity check on the Received header fields.
>>>>> See <https://mason.gmu.edu/~smetz3/source/unobfuscate.zip>
>>>>>
>>>>>
>>>>> --
>>>>> Shmuel (Seymour J.) Metz
>>>>> http://mason.gmu.edu/~smetz3
>>>>> עַם יִשְׂרָאֵל חַי
>>>>> נֵ֣צַח יִשְׂרָאֵ֔ל לֹ֥א יְשַׁקֵּ֖ר
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> ________________________________________
>>>>> From: IBM Mainframe Discussion List <[email protected]> on 
>>>>> behalf of Charles Mills <[email protected]>
>>>>> Sent: Saturday, August 22, 2026 7:02 PM
>>>>> To: [email protected] <[email protected]>
>>>>> Subject: Re: Odd idea?
>>>>>
>>>>>
>>>>> External Message: Use Caution
>>>>>
>>>>>
>>>>> I'm not an expert oeee-mail but on the negative side isn't the "from"
>>>>> pretty easy to spoof? There are reality checks you can do on the 
>>>>> IP addresses and so forth, but is it bulletproof enough?
>>>>>
>>>>> On the positive side, the email body is potentially a pretty rich 
>>>>> interface. You could accept a "command" with a syntax that allowed 
>>>>> all kinds of options: email always, or just on failure? Override 
>>>>> job class and scheduling? Etc.
>>>>>
>>>>> Charles
>>>>>
>>>>> On Sat, 22 Aug 2026 18:46:09 +1000, Peter Vlss <[email protected]> 
>>>>> wrote:
>>>>>
>>>>>> Hi Brian,
>>>>>>
>>>>>> A colple of quick thoughts:
>>>>>>
>>>>>> 1. Security.  How do you determint that the incoming email is legitimate?
>>>>>> 2. Yes, tell the user whether it was successful or not,  but give 
>>>>>> them a choice, make it a keyword or flag? If you don't tell them 
>>>>>> it was successful,  how do they know it even executed?
>>>>>> 3. Same goes for the output; give the user a choice.
>>>>>>
>>>>>> Regards,
>>>>>> Peter
>>>>>>
>>>>>>
>>>>>> On Sat, 22 Aug 2026, 17:09 BrianeWesterman, < 
>>>>>> [email protected]> wrote:
>>>>>>
>>>>>>> I was wondering.  Since IBM is unlikely to ever change CSSMTP to 
>>>>>>> create the ability to send email to the mainframe, what do you 
>>>>>>> think about adding the ability for sending email to our 
>>>>>>> automation system to have it execute functions and/or commands 
>>>>>>> on the mainframe?  We have supported using FTP to do that for 
>>>>>>> many years, and have provided trigger commands, but this would 
>>>>>>> be creating the ability to send an email to what would be the 
>>>>>>> "inbox" for talking with the Syzygy automation system and having 
>>>>>>> it confirm if the email sender was authorized to issue commands 
>>>>>>> and of what type, and then pass them on to the automation system on the 
>>>>>>> various LPARs that automation is executing on for processing.
>>>>>>>
>>>>>>> I have a prototype created that I have been testing for a while 
>>>>>>> and I am looking for volunteers to try it out.  The only 
>>>>>>> restriction is that you have to be running the Syzygy Automation 
>>>>>>> package, (shame on you if you aren't :).
>>>>>>>
>>>>>>> I'm trying to decide if the process needs to reflect back to the 
>>>>>>> user that the process was successful.  I current tell them if it 
>>>>>>> wasn't, and there is no reason that I couldn't tell them that it 
>>>>>>> was as well, but can't decide if that's necessary.
>>>>>>>
>>>>>>> I found that I could also ask automation to send the output of 
>>>>>>> jobs in the queue, but didn't know if that's something that 
>>>>>>> would be useful because the automation system already can send 
>>>>>>> email when any job ends that contains the output already.  So it 
>>>>>>> could be a "feature" I keep to the side to add later.
>>>>>>>
>>>>>>> Even if (for some odd reason) you don't run the Syzygy 
>>>>>>> Automation system, I would be interested in your thoughts on this 
>>>>>>> process.
>>>>>>>
>>>>>>> Brian
>>>>>>>
>>>>>>> ----------------------------------------------------------------
>>>>>>> -
>>>>>>> -
>>>>>>> ---- For IBM-MAIN subscribe / signoff / archive access 
>>>>>>> instructions, send email to [email protected] with the
>>>>>>> message: INFO IBM-MAIN
>>>>>>
>>>>>> -----------------------------------------------------------------
>>>>>> -
>>>>>> -
>>>>>> --- For IBM-MAIN subscribe / signoff / archive access 
>>>>>> instructions, send email to [email protected] with the
>>>>>> message: INFO IBM-MAIN
>>>>>
>>>>> ------------------------------------------------------------------
>>>>> -
>>>>> -
>>>>> -- For IBM-MAIN subscribe / signoff / archive access instructions, 
>>>>> send email to [email protected] with the message: INFO 
>>>>> IBM-MAIN
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> ------------------------------------------------------------------
>>>>> -
>>>>> -
>>>>> -- For IBM-MAIN subscribe / signoff / archive access instructions, 
>>>>> send email to [email protected] with the message: INFO 
>>>>> IBM-MAIN
>>>>
>>>> -------------------------------------------------------------------
>>>> -
>>>> -
>>>> - For IBM-MAIN subscribe / signoff / archive access instructions, 
>>>> send email to [email protected] with the message: INFO 
>>>> IBM-MAIN
>>>
>>>
>>>---------------------------------------------------------------------
>>>- For IBM-MAIN subscribe / signoff / archive access instructions, 
>>>send email to [email protected] with the message: INFO 
>>>IBM-MAIN
>>
>>----------------------------------------------------------------------
>>For IBM-MAIN subscribe / signoff / archive access instructions, send 
>>email to [email protected] with the message: INFO IBM-MAIN
>>
>>----------------------------------------------------------------------
>>For IBM-MAIN subscribe / signoff / archive access instructions, send 
>>email to [email protected] with the message: INFO IBM-MAIN
>
>----------------------------------------------------------------------
>For IBM-MAIN subscribe / signoff / archive access instructions, send 
>email to [email protected] with the message: INFO IBM-MAIN
>
>----------------------------------------------------------------------
>For IBM-MAIN subscribe / signoff / archive access instructions, send 
>email to [email protected] with the message: INFO IBM-MAIN

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions, send email to 
[email protected] with the message: INFO IBM-MAIN

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to