gabriel rosenkoetter on Tue, 25 Sep 2001 23:40:28 +0200 |
On Tue, Sep 25, 2001 at 09:08:15AM -0400, Mike Leone wrote: > If you're a programmer, perhaps not. Not everyone is, you know. :-) Too many > Linux users forget that point. Also, I'm coming from the (Net)BSD world where all serious users are programmers. > No, I'm not missing the point. I just want it to do things it's (apparently) > not designed to do. I don't want just anybody with a SSH-enabled telnet > client to be able to type in my IP address, and immediately get a system > login prompt. That's only 1 layer of security. This is not a problem that public/private key authentication of users is meant to solve. This is what limiting clients access based on their *host*'s public key (which you can probably do with a little hacking too) would be for, but not based on the user's key. Likewise, this is a problem solved very well by a good stateful firewall or, really, just by tcp_wrapping sshd. Note that a daemon does NOT have to be launched from /etc/inetd.conf in order to be tcp_wrapped... intelligent distros link all their daemons against libwrap, and OpenSSH's configure script will let you do the same. Any daemon linked against libwrap will honor access rules in /etc/host.{allow,deny}. If your problem is that you have users in public-access IP ranges, then what you're looking for is IPSec or another VPN solution. ssh is simply not designed to deal with this, and (imho) it shouldn't be. This is a problem *separate* from login authentication and session encryption. > Ideally, I want SSH to only respond if there is a public key installed on > the host ahead of time. And even then, I want it to ask for a valid host ID > and password. Swell, then fix it that way and submit your changes or pay somebody to do so. :^> Wishing something was true about a piece of software doesn't make it so, but making it happen is supremely easy with open source software. > Yes, I realize the first part is more of an authorization/authentication > mechanism better suited to a VPN, to allow access into the LAN at all. Bingo. This is the right way to do what you want. But tcp_wrappers will cut it for *just* want you want without much room to expand. > Was just trying to figure out if I could do both ways - public key > passphrase and ID/password - to 1 host at the same time. But it looks like > it's not to be. Well, not without a little elbow grease expenditure. But, as you suggested and I elaborate above, there are already ways to do what you want. (I think probably the tcp_wrappers solution would be most to your liking and involve the least addition cruft in between the world and your host.) -- ~ g r @ eclipsed.net Attachment:
pgprc82KrQhgh.pgp
|
|