[109196] in Cypherpunks
Re: birthdate statistics?
daemon@ATHENA.MIT.EDU (Osama Bin Laden)
Sat Mar 13 17:31:05 1999
Date: Sun, 14 Mar 1999 00:06:26 +0200 (EET)
From: Osama Bin Laden <waste@zor.hut.fi>
To: David Honig <honig@sprynet.com>
cc: cypherpunks@toad.com
In-Reply-To: <3.0.5.32.19990313080018.0086f5c0@m7.sprynet.com>
Reply-To: Osama Bin Laden <waste@zor.hut.fi>
On Sat, 13 Mar 1999, David Honig wrote:
>
> If you learn someone's birthdate, have you obtained log2 (365 1/4) bits?
> If you learn that someone was born in the first half-year, is this 1 bit?
>
> I.E., Are birthdates truly uniformly distributed?
Actually, no they are not. There are some high birth peaks,
but they vary by area. There are some indications that the
conception rates are much higher on days that traditionally
have parties and celebrations.
Of course, this is very much regionally and culturally dependant.
> Anyone thought about the applicability of cryptanalysis to
> identifying people from census info? About identifying
> Finns from their soon-to-be public health records.
> (No, I don't want to do this, but how would you reason
> about defenses against this kind of data mining?)
Duh. I doubt there's any reason to apply cryptanalysis to
something that isn't encrypted. More like statistics and
probabilities.
Of course, to identify some 'anonymous' person on the basis of
what you know of him, you likely need to get to lots of databases
and cross-reference them. Phonebooks, tax records, etc. The police,
FBI, NSA, etc, are likely to have lots of experience in it.
