Statistics has always seemed to me stuck in a place that resembles physics before Newton. Lots of pieces that work on their own, but no unifying theory. This drives me crazy, and is probably a reason that there is so much hating on statistics. I discovered Kolmogorov Complexity reading a paper on vacation last week, and wonder if that could be used as a basis for a unified theory of statistics. Here’s a reasonable starting point:
Kolmogorov Complexity – A Primer
which leads to
Information Distance — A Primer
I had an interesting talk with a work buddy about his Master’s program which is Information Security. That led to a rambling discussion on what “trust” and “identity” mean. So what if access to a system was based on an Eliza-like conversation with the computer. There would be an initial “get acquainted” session, and then the “login” conversation would be shorter.
- Store key hit timing and error/correction patterns as well as the words.
- Use a plagerism test to look for cut-and-paste replies from recorded sessions
- Login discussions could be like Capchas in that they could be half test, half learning.
Mostly scrambling to finish a paper, but this looks interesting…
So I have this project that might be a really interesting, though somewhat dark excursion into CSCW. I’m not quite sure what to do with it, but when I hear things like this on Morning Edition, I really wonder if I just shouldn’t figure out a way to go all in. The question is – is a solution that winds up with fewer people getting killed in a war always better?
And here’s something that matters for this: http://securityledger.com/mobile-phone-use-patterns-the-new-fingerprint/. This is the paper the article references: http://www.nature.com/srep/2013/130325/srep01376/fig_tab/srep01376_F1.html
A few more useful links: