>Example:
>
>Joe Blue on machine king.lame.net wants to let people finger him, but
>does not want people to beable to look at his system behind the
>firewall. He setups a "fake account" on the firewall and remotely
>updates the ".plan" file on that machine so people can: finger
>JoeBlow @
lame .
net without the packates getting into the internal net.
>
>I know it's a bad example, but you should beable to get the point.
Given that there's no inherent security risk in touching a network wire
to your computer, all risks arise out of the implementations of the
services (user- & kernel-level) offered to the network by a system.
Yes, it would be possible (and probably not too hard) to provide a
_simple_, safe user information server. The protocol very simple;
there's no need to get carried away with bells & whistles like the
standard fingerd. See the RFC. We'll ignore the content of the
information that Joe's giving out this way -- that's a different
topic.
What _I_ would like is a fingerd replacement that would give me, the
administrator of the system, various information about the firewall
without logging in. I tried one or another of the fingerd replacements
that allows you to specify fake users & run a given process, but I was
not satisfied with the security aspects and didn't actually install
it. It would, of course, be for inside-only connections, preferably
use some sort of real authentication, and be certifiably safe against
such string foolery as gets you in trouble with system() calls &c.
Seems to me I could use fwtk or tcp_wrapper as a starting point...
-- KH
|
|
