Randomisation - IBM's answer to Web privacy

[bear]

On Tue, 4 Jun 2002, R. A. Hettinga wrote:

>http://www.theregister.co.uk/content/23/25551.html
>
>Randomisation - IBM's answer to Web privacy
>By ComputerWire
>Posted: 03/06/2002 at 08:27 GMT

<deleted>...

On further reflection, IBM's special sauce here may be a
different flavor entirely if the randomized data are linkable.

If you introduce a random error in measurement with a range
of say 5 years, and IBM does business with linkable customers
over a sequence of several orders, they can home in on the true
values for individual customers ages over time and literally
correct the database information based on variants recieved
with subsequent orders.  Let's say they'll get the customer's
age as 28 the first time -- and know it's between 23 and 33.
Then next time say they get 32.  Now they know it's 27 to 33.
The time after that, they get 27, and narrow it to 27 to 32.
Etc....  by the time the guy has placed 20 orders they're
probably going to know his age to within one year.


			Bear





---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at wasabisystems.com