Hi, I need to implement a company policy of requiring a minimum key size that is estimated to be secure for the next X years (see keylength.com estimates). For that we would need to reject authentication with keys that do not meet the key length size which is necessarily higher than the minimum hard-coded in the most recent OpenSSH release today. I have already read the whole thread "Settable minimum RSA key sizes on the client end for legacy devices." ( https://marc.info/?t=157733023300001&r=1&w=2) so I believe I understand the arguments made there about *lowering* the minimum. However, I'm talking about *increasing* the minimum at runtime. a) Is there an argument to be made against a feature like this? b) Can this be done today? I've given some thought to using AuthorizedKeysCommand in combination with a hard-coded non-existent value for AuthorizedKeysFile. Setting AuthorizedKeysFile to /dev/null and setting up AuthorizedKeysCommand to look at the user's actual authorized_keys file, filtering out keys not fulfilling the policy sounds like it could work? -- Irfan