Late to my own party (again)...
On 05/16/26 06:03 pm, Steven Levine wrote:
In<list-2450420@2rosenthals.com>, on 05/14/26
at 11:40 AM, "Lewis G Rosenthal"<ecs-isp@2rosenthals.com> said:
Hi,
validity duration from 398 days to 47 days. The first phase has started
and validity is now 200 days. This will again change to 100 days by
March 2027 and finally to 47 days by March 2029.
The good news is that there is plenty of time to prepare for the full
switchover to 47 day lifetimes and for the effects of the interim life
time changes to be evaluated. I have no idea how big CRLs tend to be
these days, but reducing their size cannot be a bad thing.
I would have to suspect that some commercial CAs will provide semi-automated methods for updating. As I do when building new pci-ids and usb-ids packages, I have scripts which go out and download the requisite files, massage the content as necessary, and then build the RPMs. I don't run them on a set schedule, though I could.
In the case of a Starfield Tech cert, this becomes a little more involved due to the current login process on the site (any such script would have to mimic a user's actions in a browser, much like a bot, but this could be done with some careful analysis of the steps involved - a POST and a GET are just what they are, assuming one knows the data to POST and GET). TFA becomes a bit trickier to navigate, as do silly CAPTCHA things (neither needed just yet, but I expect they are coming).
Of course, downloading the new certs is probably the least of the annoyance, so even if I just "did it" every month, and stored the new cert zip file in a known location, my to-be-written script could just check for a new package as a cron job and if new, "do the needful," as they say. Not fully automatic, but automatic enough. (I log into multiple online banking systems every day, so I could just as easily add logging into my GoDaddy account to my daily routine.)
The whole argument about shorter cert lives being more secure is a tough
one for me, given the availability of OCSP stapling and other
validation/revocation methods. Oh, well.
As others have mentioned both OCSP and OCSP stapling seem to be going
away. It appears that neither really is widely in use.
Well, for a while, neither was DNSSEC, but now that seems to be seeing a bit of momentum, again.
For Apache, there is mod_md, which improves upon the OCSP stapling framework.
I'm just not convinced that reducing certificate lifetime is a viable security approach. The keys could be compromised the day of issuance (a continuous tap on a network, for example), in which case, the cert could be compromised quite literally before it was even installed, so a 47-day or a 400-day lifetime would make no difference.
I recall some years ago, you and I traded some interesting arguments concerning (forced) password rotation as it pertained to Yahoo email accounts, I believe. My argument was that if the password had already been compromised, the mail and address book would likely have already been accessed and stolen, so changing the password would do little to address what had already happened. Your argument was that by changing the password more frequently, it would theoretically reduce the window for such breaches (IIRC). I see this artificially-curtailed cert lifetime as something along the same lines: it's a shortsighted approach to a bigger problem.
Realtime validation of a given cert is the best way to ensure timeliness of the response. That reads as an oxymoron, but it is true nonetheless. Barring realtime validation, timestamped validation within a narrow window would be the next best choice from my perspective (though neither is foolproof: one would have to report the keys stolen in order for the OCSP server to invalidate the spoiled cert; no reporting means no revocation).
Otherwise, this is all just like me having to change the combination on my front door lock every 6 weeks vs every day vs once every 10 years: the combination may still be stolen by someone looking over my shoulder at the time I change it. After that, it's all just more form over substance.
Finally, this article pretty much sums up what is supposed to happen in the event of compromised keys:
https://www.encryptionconsulting.com/education-center/handle-breached-certificate-and-key/
The question to ask, however, is "what if someone *isn't* watching and *never* becomes aware of a breach?" References to "...as soon as a breach is detected..." imply that there will always be some internal practice to monitor and detect such breaches, but we all know that in practice, this is harder than it looks (*a* breach does not necessarily mean that the private key to a given cert was compromised, so how to tell for certain? every breach? how would one know if five different breaches had occurred or only one over a longer period of time?). So, again, making life more complicated for all of us is a Pyrrhic victory, AFAIC.
--
Lewis
-------------------------------------------------------------
Lewis G Rosenthal, CNA, CLP, CLE, CWTS, EA
Rosenthal & Rosenthal, LLC www.2rosenthals.com
visit my IT blog www.2rosenthals.net/wordpress
-------------------------------------------------------------
|