|
BTW after reading the full 15 page paper, I learned that
the technique isn't as effective as you might think from the
short article. Essentially it lets you classify all remote
computers into about 50 groups. It's as though there were
only 50 different fingerprints shared by everyone in the world.
If the suspects's fingerprints are different from those at
the scene of the crime you could say they didn't commit
the crime. If they do match, they would also match
100,000,000 other people besides the suspect so
it's not that useful.
So it's not too much more useful than standard techniques
we've used for a long time - classifying users by what browser
they use, what operating system, etc. It just provides one
more criteria to distinguish between groups of users.
It's a criteria we might use in Strongbox, though.
Currently Strongbox is only about 15,000 times accurate
than something like Pennywize. Adding the skew factor
multiplies that by 50, so it'll be 750,000 times more accurate.
Actually that's not entirely true though, after making various
adjustments it'll be maybe 300,000 times as accurate.
|