Bug#710853: openssh-server: ssh server keys creation
Colin Watson <[email protected]> Tue, 2 Dec 2025 10:58:47 +0000
| Newsgroups | gmane.linux.debian.devel.ssh |
|---|---|
| Message-ID | <aS7GZ1KdLO0U1Udp__43938.4024635487$1764673292$gmane$org@riva.ucam.org> |
Control: retitle -1 openssh-server: fine-tune server key creation On Mon, Jun 03, 2013 at 02:32:06AM +0200, Christoph Anton Mitterer wrote: >With respect to the creation of SSH server keys in postinst, may I suggest the >following: >- not create ssh1 keys at all... actually I've never seen them auto-created, > but code seems to be there > This is mainly for security reasons... if someone really want's ssh1, he shoul > manually create the keys. Done in 1:7.1p2-2 (see #811265). >- specify bit sizes > Also for security reasons, use the highest bit sizes possible for the respective > algorithm,... it should have basically no performance impact, and if someone > really thinks he wants a weaker key,.. he still can manually create it > That is > rsa2: -b 4096 > dsa: -b 1024 > ecdsa: -b 521 (no typo) ssh-keygen's defaults are fine, especially nowadays that it defaults to 3072 bits for RSA keys. For the same sorts of reasons that I laid out in https://bugs.debian.org/1094246#10, I don't think it's necessary to override them here. >- use the FQDN as comment > I always found it handy to have the full hostname on the server keys as comment, i.e. > -C "$(hostname -f)" > without username, as e.g. root@$(hostname -f), would be the personal key of the user > root. This seems probably reasonable. The only thing I was wondering was whether there were any (minor) privacy implications to recording that information? I guess not but I'm not certain.