...in that the Bad Guys can join the network and learn the realtime "shield nutation frequency" (just love that TrekBabble™), providing them with realtime calibration.

Not if my understanding of how this works is correct. If everyones 'last hop latency' on the graph is the EXACT same, they can never get better than 'note before last' resolution of location- which you can often get with Traceroute and Google+Google Earth.

To get better resolution, here is what you have to have (butchering formalism here due to HTML inadequacies):

an 'attack path' o....z, n

such that:
every element of {o, ..., z, n} are nodes on the internet graph,
o being is a beaconing node
z is the the last node in the 'attacking' net. IE, the 'next to last hop'
n is any 'end user' behind the path o....z, ie, n is the terminal node of the location finding attack. The end user.

Where all latencies (ol......zl) are determined AND
the latency of end user (nl) is known and is at least partially distinct (as in a partial ordering exists ) from every other end user latency that is on the path o.....z+n

NOW, suppose every nl resolves to the same value as every other nl that is on the attack path o....z, n. Every end user behind the 'next to last hop' looks like every other end user behind the 'next to last hop'.

Then the BEST location resolution possible will be the loaction of Z. Which you and I often get for fun with minimally clever use of network tools.

Now, I still have not read the patent, so I could be completely silly.

DL