Re: Tagging copyrighted text

[Marc.Ringuette@GS80.SP.CS.CMU.EDU]

I've been thinking more about how Dow Jones can catch customers which
leak its data.  Let's assume that I, as the thief, get 20
subscriptions, and that Dow Jones tries twiddling the value of various
bits of its output with some probability, differently for each
customer, to catch who's leaking its data.

My strategy as the thief is, whenever the newswire doesn't agree
between all of my subscriptions, they take a majority vote.

How many bits of random tag must Dow Jones add to its newsfeed to catch
a clique of size 20 out of its 10,000 customers?

I think I've narrowed the answer down to between 437 bits and 250 Gbits.

<sigh>


-- Marc Ringuette (mnr@cs.cmu.edu)