Hello All,

Among other good reasons to do full pcap dumps of network traffic,
we use this primarily for off-site analysis.  Currently, it's a
cumbersome process and I'm seeing if anyone has better stragegies. 
I have an idea, too, on a better process and am hoping to get some
feedback (^_^)

Generally, when I need to dig into a pcap for a specific issue, I
have the following information available:

--) Plugin ID number (e.g. 10101)
--) IP addr of host(s)

I use this to grep the KB files for start and stop times, reported
in UNIX epoch format.  I then use tcpslice to extract that range
and tcpdump/tethereal to filter a bit more (src/dst of scanner and
src/dst of target).  Sometimes this is sufficient, especially if
the time interval was short and there weren't overlapping checks. 
But, usually I'm not so lucky, so I have to figure out what is
noise and what isn't.  This isn't hard, just something that can't
be automated since I have incomplete data.  Also, since many
plugins use dependencies, that "noise" is actually another plugin
being fired and determining information.

So, how I see it, I can either attempt to get better information to
sort my data (e.g. TCP/UDP ports, strings in the plugin, etc) or
get better data to begin with.  I see the latter being a better
approach.  Unfortunately, though, the only way I see doing this is
modifying each plugin and performing pcap / result dumps in each
one.

The pcap dumps would have to be organized somehow to be meaningful
(perhaps saved on disk as <plugin-id><dst ip addr><counter>.pcap or
information written to some type of database).  It would seem to
get better information, if somehow it took into consideration all
the dependencies and captured it in all one file (or had
information linking them all together).  This would only work for
those plugins that create and interpret the network traffic (versus
local host checks via ssh or such, or those that reference KB
info).  This also doesn't address the overhead related to the
increase in disk read/writes, pcap overhead, or dropped packets,
which are real-world concerns.

Anyone have better approachs for off-site analysis?  Does the
in-the-plugin approach seem reasonable?  Or, is this a constraint
of off-site analysis, and one should only do on-site analysis only?
;-)  (I joke, since I've had too many clients where I needed to
analyze something way after we were done scanning, so I see
off-site analysis as necessary).

TIA,

Jon


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
_______________________________________________
Nessus mailing list
[email protected]
http://mail.nessus.org/mailman/listinfo/nessus

Reply via email to