I think the desired behavior depends on who is doing the scan, why the scan
is being done and what is being scanned. Unless I'm misunderstanding the
paranoid setting I agree with Renaud's assessment and in any case I prefer
the current behavior. This preference is based on the who, why and what I
mentioned above.

My use of nessus is as part of a continuous vulnerability assessment at a
university. We are constantly scanning the network for vulnerable systems.
At any given time there are over 8,000 systems on the network and we do not
want to process false positives -- they waste too much time. If we had to
ignore that open ssh plugin due to too many false positives then we lose out
on the true positives it would give.

That's my $0.02

Tim Doty


-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
On Behalf Of Richard van den Berg
Sent: Thursday, October 11, 2007 11:06 AM
To: [email protected]
Subject: Replacements too eager in backport.inc

Sorry Renaud for posting this here as well, but I'd like other people to
comment on this as well.

Today I found out why Nessus is so sparse with reporting OpenSSH issues. 
It is because backport.inc has entries like this:

backported_versions[i++]   = "OpenSSH_3.1p1";
real_versions[j++]         = "OpenSSH_3.9.99";

backported_versions[i++]   = "OpenSSH_3.4p1";
real_versions[j++]         = "OpenSSH_3.9.99";

backported_versions[i++]   = "OpenSSH_3.5p1";
real_versions[j++]         = "OpenSSH_3.9.99";

backported_versions[i++]   = "OpenSSH_3.6.1p2";
real_versions[j++]         = "OpenSSH_3.9.99";

backported_versions[i++]   = "OpenSSH_3.7.1p2";
real_versions[j++]         = "OpenSSH_3.9.99";

backported_versions[i++]   = "OpenSSH_3.8p1";
real_versions[j++]         = "OpenSSH_3.9.99";

backported_versions[i++]   = "OpenSSH_3.9p1";
real_versions[j++]         = "OpenSSH_3.9.99";

This effectively maps almost all openssh 3.x version strings to 3.9.99
unless you set report_paranoia to 2 (=Paranoid). There are similar entries
in backport.inc for Apache.

My argument is that backport.inc was meant to map vendor (Debian/SuSE/Red
Hat) version strings to real product version strings. 
So when you see "SSH-2.0-OpenSSH_4.3p2 Debian-9" you know that this is
Debian openssl-client package 4.3p2-9 which is really openssh 4.4p1. 
However, with the above examples, there is no indication whatsoever that
these versions are patched or even belong to a vendor that supplies a patch.

Renaud's answer is that changing backport.inc would create too many false
positives, and that you can always set report_paranoia=2 if you don't like
it. My reply is that now you get a lot of false negatives which is a lot
worse (and not what the report_paranoia was designed for).

What do you think?

Please read the full report and Renaud's answers before answering: 
http://bugs.nessus.org/show_bug.cgi?id=1643

Sincerely,

Richard van den Berg
_______________________________________________
Nessus mailing list
[email protected]
http://mail.nessus.org/mailman/listinfo/nessus

Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
Nessus mailing list
[email protected]
http://mail.nessus.org/mailman/listinfo/nessus

Reply via email to