Bug 1199184 - switch Tumbleweed to a 4096bit RSA signing key
switch Tumbleweed to a 4096bit RSA signing key
Status: CONFIRMED
Classification: openSUSE
Product: openSUSE Tumbleweed
Classification: openSUSE
Component: Basesystem
Current
Other Other
: P3 - Medium : Normal (vote)
: ---
Assigned To: Dominique Leuenberger
E-mail List
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2022-05-04 08:45 UTC by Marcus Meissner
Modified: 2022-06-14 09:18 UTC (History)
4 users (show)

See Also:
Found By: ---
Services Priority:
Business Priority:
Blocker: ---
Marketing QA Status: ---
IT Deployment: ---


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Marcus Meissner 2022-05-04 08:45:26 UTC
We currently still use a 2048 bit key for signing.

It would be better to switch to a 4096 RSA key.
Comment 1 Fabian Vogt 2022-05-04 08:53:33 UTC
AFAIK 2048 bit RSA is still considered secure. What about switching to something like ECDSA/EdDSA?
Comment 2 Marcus Meissner 2022-05-04 08:57:08 UTC
Is still considered secure, but other distros use longer keys and e.g. Dirk Mueller already argues on why openSUSE does not switch.

I am not yet familar how much could break with switching to elliptic curves though.
Comment 3 Dirk Mueller 2022-05-04 09:14:35 UTC
(In reply to Marcus Meissner from comment #2)
> Is still considered secure, but other distros use longer keys and e.g. Dirk
> Mueller already argues on why openSUSE does not switch.

I'm not arguing, I was asking what needs to be done to implement a longer key for ALP. 

Based on factory first we should try to roll it out in openSUSE first and see the downsides before doing anything on SLE. 

From a brief look at  https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-57pt1r5.pdf 

it appears that RSA2048 is the acceptable minimum, and other distributions are chosing larger keys. 

I don't really care which cryptographic method we chose, so elliptic curve is totally fine by me as well. I don't know the implications of that very thorughly though, more expertise is needed.
Comment 4 Fabian Vogt 2022-05-04 09:55:20 UTC
Looks like EdDSA support landed in RPM in 2020 and it works on TW (with sha256 only, https://github.com/rpm-software-management/rpm/issues/1877 is missing). It doesn't work on Leap though, so either that would have to be backported (https://github.com/rpm-software-management/rpm/pull/1202 at least) or we'd have to deal with RSA a bit longer.

IMO it's better to keep RSA 2048 for a a bit longer and then switch to ECC directly instead of switching to RSA 4096 now (and maybe switch to ECC in the future).
Comment 5 Marcus Meissner 2022-05-04 13:16:27 UTC
I filed an OBS ticket to make signing keytypes configurable via api calls, as currently OBS API would only creating 2048bit RSA keys.

https://jira.suse.com/browse/OBS-193
Comment 6 Stefan Seyfried 2022-06-09 05:33:17 UTC
Is there also a github issue or such that I could watch for the OBS implementation discussion?
(I want to change to a longer key in a private OBS instance and would hope that this discussion helps me achieve that seamlessly ;-))