-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Nack
On 01/29/2010 05:51 AM, Sumit Bose wrote:
On Thu, Jan 28, 2010 at 02:52:21PM -0500, Stephen Gallagher wrote:
Nack. Please specify in the SSSDConfig API that the time is in minutes.
> fixed
I think it would be more correct to return EIO if we fail to get the
login delay or allowed attempts from confdb. EACCESS implies that we
explicitly failed.
> fixed
The DEBUG messages after failure to save SYSDB_LAST_LOGIN and
SYSDB_FAILED_LOGIN_ATTEMPTS are wrong.
> fixed
No it's not:
+ ret = sysdb_attrs_add_time_t(state->update_attrs, SYSDB_LAST_LOGIN,
+ time(NULL));
+ if (ret != EOK) {
+ DEBUG(3, ("sysdb_attrs_add_long failed, "
+ "but authentication is successful.\n"));
+ ret = EOK;
+ goto done;
+ }
And several other similar ones.
I'm not sure if this was intentional or not: if I'm reading this
correctly, we're still saving the last failed login time every time it
fails, even if we're already in the 5-minute waiting period. So,
effectively, it's going to deny forever unless they wait five minutes.
> no, you are wrong, if check_failed_login_attempts() fails or denies
> access nothing is changed or updated.
Yeah, I see it now. No problem.
I think we probably want to stop storing the last failed login time once
we hit our failed login counter, otherwise things might be frustrating
for the user.
On the other hand, this is a great way to defeat brute-force attacks, so
I'm not sure how we want to proceed on this.
> I think this would be an unexpected behaviour and it might be a bit
> annoying to a user who didn't look too closely on his watch and retried
> after 4 minutes and 55 seconds.
Agreed.
Also, would it be possible to return a message to the user that they've
passed the maximum attempts and that they need to wait five minutes?
(I'm thinking something similar to your "Warn the user if authentication
happens offline" patch.
> This would be possible, but wouldn't this disclose too much information
> to an attacker?
I'm not sure this is harmful information to disclose. The attacker now
knows that the previous passwords didn't work, and that they now need to
wait five minute more to try the next set of passwords that probably
won't work.
If anything, I think it would be better, because it might discourage the
attacker from hammering the site and wasting bandwidth and resources.
I'll open an RFE for this and we can discuss it separately. This is out
of scope for this patch.
- --
Stephen Gallagher
RHCE 804006346421761
Delivering value year after year.
Red Hat ranks #1 in value among software vendors.
http://www.redhat.com/promo/vendor/
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Fedora -
http://enigmail.mozdev.org/
iEYEARECAAYFAktm1yUACgkQeiVVYja6o6NF0QCdGUl45qiO4nN7EEnWcwsheDqd
g/sAoK8fnCwAUf7mHYV6v4/LZLOjnfWP
=nk9f
-----END PGP SIGNATURE-----