Re: IDS is dead, etc

[Bennett Todd <bet () rahul net>]

2003-08-07T16:49:10 Barry Fitzgerald:
> Oh yes, and someone (perhaps tongue-in-cheek) mentioned that a
> properly configured firewall removes the need for an NIDS.

Perhaps you're referring to my comment:

        2003-08-06T14:57:53 Bennett Todd:
        > 2003-08-06T07:39:28 Paul Schmehl:
        > > Why would you want to know about Nimda attacks
        > > against your servers?
        >
        > (or more generally, attacks that won't succeed)
        >
        > Some people _don't_ care. They need to disable the
        > sigs they don't care about, or configure their IDS
        > to only match those sigs against servers for which
        > they're relevent.
        >
        > The limiting case of this argument says that given
        > a really perfectly implemented firewall, you don't
        > need an IDS at all. Some folks don't.

> I have to chime in and say that I couldn't possibly disagree more.

Understandable. I really shouldn't have included that remark; or
else I should have expanded on it. I didn't say "properly configured
firewall", I said "really perfectly implemented firewall", and I
meant something different by that, although I neglected to explain.

A perfectly implemented firewall allows no protocols through for
which there are vulnerable implementations inside. That means it's
impossible to implement a perfect firewall if you're going to allow
Windows users to have internet access. You can come moderately
close, with a hideous amount of work, but you'll still be very
exposed, and an IDS will be critical reinforcement of your flawed
security.

But given suitable systems configuration, it is possbile to have a
perfect firewall, and if you do then an IDS is just an educational
tool, and would probably be most useful in concert with a honeypot.

-Bennett

Attachment: _bin
Description: