Here are the papers related to today's talk on hiding sensitive rules in data mining:

1. The vision paper by Clifton 1996 is available at
http://www.cs.purdue.edu/homes/clifton/document/dmkd.pdf


About hiding association rules:

The formulation of the problem and (probably flawed) NP-hard proof:
2. Disclosure limitation of sensitive rules, by Atallah, Bertino, et al, 1999
http://citeseer.ist.psu.edu/atallah99disclosure.html

A set of (greedy, simple) heuristic santization algorithms:
3. Association rule hiding, by Verykios, Bertino, et al, 2004
http://ieeexplore.ieee.org/iel5/69/28407/01269668.pdf?tp=&arnumber=1269668&isnumber=28407


Then they argue that to flip 0/1 introduces wrong information, which is not good.
Instead, they suggest using "unknown" value in the following paper
(which seems more close to k-anonymity problem).
The style of algorithm is very similar to the 3rd paper.
4. Using unknowns to prevent discovery of association rules, SIGMOD record, 2001
http://portal.acm.org/citation.cfm?id=604271&dl=ACM&coll=portal

About controlling sample size to increase error probablity of learning method:
5. Protecting against data mining through samples, by Clifton, 1999
http://www.msci.memphis.edu/~linki/7118papers/Clifton99Protect.pdf

Ying