On Tue, Mar 11, 2008 at 9:05 PM, Tim Shoppa <[EMAIL PROTECTED]> wrote:
> Are there any currently executing NTP surveys, like Guyton's (1994),
> Minar's (1999), Mura/Torres (2005)?
>

I wrote some code months ago to do a random IP scanning NTP survey.
This seemed like a great idea for getting around the fact that most
servers don't respond to ntpdc or ntpq, and a good way to produce
usable estimates for the actual number of public NTP servers out there
in a non-linear IP space.

However, I realized after I wrote the code that it behaves very much
like a worm to intrusion detection systems, and in fact in testing
triggered a heuristic rule on our own IDS.

My initial rate-limited test scan of 100K IP addresses had interesting
results, but not quite enough to be statistically significant (only
one stratum-1 server was found, for example.) I was unnwilling to go
further and scan wide swaths of the Internet, lest my ISP notice the
scanning and decide to cut us off as a pre-emtpive security measure.

Then I realized that measuring the number, type, and quality of
publicly accessible servers, while interesting, is not nearly as
interesting as also measuring the number type and quality of clients
per server. My method couldn't do that, and I couldn't think of a way
to do it, so I let the whole idea slide until I came up with
something.

-- 
RPM
_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers

Reply via email to