Johnsten, T. and Raghavan, V. (1999). Impact of decision-region based classification algorithms on database security. In Proceedings of the Thirteenth Annual IFIP WG 11.3 Working Conference on Database Security. Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
Unknown - (Correct)
....with the rise of data mining technology, this has become a recognizable problem. What can we do about this, in particular when we don t know what the adversary will be looking for (In cases where we know the inference we must keep secret, like the example above, other techniques are available. [Johnsten and Raghavan, 1999]) We can ensure that any results will be suspect. If we can convince the adversary that any guess (inferred rule) gained from data mining is no more likely to be a good guess than a random guess, the adversary will not be able to trust any data mining results. How do we do this Inferences from ....
Johnsten, T. and Raghavan, V. (1999). Impact of decision-region based classification algorithms on database security. In Proceedings of the Thirteenth Annual IFIP WG 11.3 Working Conference on Database Security.
Using Sample Size to Limit Exposure to Data Mining - Clifton (2000) (7 citations) (Correct)
....(e.g. inference rules involving a single independent variable) This limits our space of possible classifiers. As we will show; these two variations lead to di#erent solutions. 1 In cases where we know the inference we must keep secret, like the example above, other techniques are available. JR99, ABE 99] 2 1.1 What this work will lead to The eventual goal of this work is to provide a tool usable by security administrators. This tool would enable the administrator to determine: The allowable size of a sample, given constraints on the quality of what an adversary would be ....
....is a worst case error: We have a rule that gives exactly the opposite of the proper result. Although the probability of this happening may seem small ( 05 or .1) the result is still significant. 4.2. 1 Expected deviation of sample Note that we can do better if we know what we want to protect [JR99, ABE 99] Here we show how to evaluate expected rule performance based on the expected deviation of the sample from the real data. Given a single category rule, e.g. a rule of the form Ph.D. and Comp.Sci. # P rofessor our goal is to bound the error estimate for such a single rule; i.e. ....
Tom Johnsten and Vijay Raghavan. Impact of decision-region based classification algorithms on database security. In Proceedings of the Thirteenth Annual IFIP WG 11.3 Working Conference on Database Security, Seattle, Washington, July 26--28 1999.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC