Oct 20, 2026, 6:22 PMeMudhra says they used an HSM for the keys:
https://bugzilla.mozilla.org/show_bug.cgi?id=2083449#c11
> The hardware used for each of the issuers is in the attached file. These HSMs are FIPS 140-2 Level 3 compliant.
> The Ultimaco devices were used to generate the ECC, RSA key pairs for the issuers.
Oct 20, 2026, 6:31 PMtl;dr
b) an approved deviation to our usual KGC to test a wrap/unwrap was introduced as these keys were meant for export for this instance
c) a flaw in the code resulted in the second pair of keys being created onboard which were weaker
d) these second pair of weaker keys were used by mistake to create the csr (because the key labels were similar) and signed by our root
Oct 20, 2026, 6:24 PM@q3k fucking whaaaat
Oct 20, 2026, 6:26 PM> b) an approved deviation to our usual KGC to test a wrap/unwrap was introduced as these keys were meant for export for this instance
*extremely USCSB narrator voice* “an approved deviation”
@q3k @gsuberland the missile knows where it is
Oct 20, 2026, 6:25 PM@q3k did they get them from the same shop as those ledger wallets the other week?
@q3k "HSM" being a fancy term for what usually turns out to be a 19" rack mounted Linux box with a PCIe card in it, inside a cage, still faithfully executing whatever nonsense you type after "openssl".
Oct 20, 2026, 6:30 PM@jripley this bad boy can fit some many 4's
@secretasianman @q3k I don't think having an HSM is actually relevant. It's pretty much _mandatory_ for a CA to use one (in whatever form it may be). It's just a protected environment to run your scripts in, with guarantees on random number generators and such, but if you feed it a bad script, it's still going to give you a bad result.
@jripley @q3k this is what happens when you walk into a subway station and the joke has already run through
Editor's note: https://chaos.social/@jaseg/117474189466127816
Editor's metanote: I see you're on this thread, so I retract my snark and apologize, have a nice day
@secretasianman It's the phrase "Necessary but not sufficient". An HSM is _necessary_ to implement a secure CA, but it is _not sufficient_. Like making a virtual machine of your entire development setup, and executing it inside an HSM, does not magically make it secure.
(This sounds contrived but I once had an awkward conversation at a former employer with this exact example, which they still didn't understand)
@secretasianman @q3k @jripley The battery backup is actually also "use resistant": To swap the battery, you need to get this very specfic cylindrical lithium cell, then take off the red/white plastic cover, and then figure out again what the polarity was, since the polarity is pretty much only on this plastic cover, and not marked on the PCB. I then tried the documentation, but ended up having up relying on a product photo like q3k posted above.
@root42 I can't see any of this being Utimaco's fault. A tool's only as good as its user. If you ask it to do something dumb, it'll faithfully execute it. The point of an HSM is (barring extreme incompetence) you can execute that without outside observation, interference, or fear of storage being extracted. The rest is up to you.
@jripley Yes, I figured that out after reading more posts, too. It's basically "just" a glorified TPM. :) If you don't use it, you are screwed... But it shows that there are definitely gaps where users or integrators can make mistakes that break all the advantages of having an HSM. The whole idea should be that you CAN'T do any non-secure encryption. But obviously you can circumvent it in software anyway...
@q3k if only there was a way to not have to do fucking around on the prod hardware. but the small indie CA sadly can't afford a second HSM to do testing on
Oct 20, 2026, 6:36 PM@q3k "weaker keys" 💀🥀🥀
Oct 20, 2026, 6:57 PM@q3k tbh I think it’s a design flaw in a HSM’s programming interface if the device ever allows you to “import” a key created outside of it except for backup/restore and clustering functionality to prevent this exact type of error.
Oct 20, 2026, 6:58 PM@jaseg Yeah, I was under the impression that the whole point of an HSM is to not allow you to do that!
Oct 20, 2026, 6:59 PM@q3k I dont think their APIs are necessarily designed for misuse resistance. This sort of thing doesn’t show up in FIPS 140-2 security evaluation.
@q3k @jaseg typically there are flags you can set on a key, sometimes only at key creation time, that control exportability
The options tend to be “No export at all”, “Export only to HSMs from the same manufacturer (via key attestation from the receiving HSM)” and “No export at all”.
Typically there are also custody flags (i.e. you can know “Key was generated on this HSM”, “Key was imported from a same-manufacturer HSM with attestation” or “Key was imported yolo”, “Key has been exported with attestation”, “Key has been exported yolo”)
But like. I think quite often the keys have less strict controls set on them because nobody wants to end up stuck in a 30 year relationship with a HSM vendor.
Oct 20, 2026, 9:28 PM@q3k there is "weaker", and there is "private keys fit one byte".
What the actual four…
