Results 1 
4 of
4
Protection of Database Security via Collaborative Inference Detection 1 Abstract
"... Malicious users can exploit the correlation among data to infer sensitive information from a series of seemingly innocuous data accesses. Thus, we develop an inference violation detection system to protect sensitive data content. Based on data dependency, database schema and semantic knowledge, we c ..."
Abstract

Cited by 3 (0 self)
 Add to MetaCart
Malicious users can exploit the correlation among data to infer sensitive information from a series of seemingly innocuous data accesses. Thus, we develop an inference violation detection system to protect sensitive data content. Based on data dependency, database schema and semantic knowledge, we constructed a semantic inference model (SIM) that represents the possible inference channels from any attribute to the preassigned sensitive attributes. The SIM is then instantiated to a semantic inference graph (SIG) for querytime inference violation detection. For a single user case, when a user poses a query, the detection system will examine his/her past query log and calculate the probability of inferring sensitive information. The query request will be denied if the inference probability exceeds the prespecified threshold. For multiuser cases, the users may share their query answers to increase the inference probability. Therefore, we develop a model to evaluate collaborative inference based on the query sequences of collaborators and their tasksensitive collaboration levels. Experimental studies reveal that information authoritativeness, communication fidelity and honesty in collaboration are three key factors that affect the level of achievable collaboration. An example is given to illustrate the use of the proposed technique to prevent multiple collaborative users from deriving sensitive information via inference. 1 This research is supported by NSF grant number IIS03113283 1.
Sisterhood of Classifiers: A Comparative Study of Naive Bayes and Noisyor Networks
"... Classification is a task central to many machine learning problems. In this paper we examine two Bayesian network classifiers, the naive Bayes and the noisyor models. They are of particular interest because of their simple structures. We compare them on two dimensions: expressive power and ability ..."
Abstract

Cited by 1 (0 self)
 Add to MetaCart
Classification is a task central to many machine learning problems. In this paper we examine two Bayesian network classifiers, the naive Bayes and the noisyor models. They are of particular interest because of their simple structures. We compare them on two dimensions: expressive power and ability to learn. As it turns out, naive Bayes, noisyor, and logistic regression classifiers all have equivalent expressiveness. We show mathematical derivations of how to transform a classifer in one model into the other two. These classifiers differ on their ability to learn though. We conducted an experiment confirming the intuition that naive Bayes performs better than noisyor when the data fits its independence assumptions, and vice versa. However, we still do not have a clear set of criteria for determining under exactly what conditions would each classifier excel. Further study of the strenghts and weaknesses of each classifier should provide deeper insight on how to improve the current models. One possible extension would be to combine the naive Bayes and noisyor model so that the network will more closely depict the actual relationship between the attributes. 1
Evidence and Scenario Sensitivities in Naive Bayesian Classifiers
, 2008
"... Empirical evidence shows that naive Bayesian classifiers perform quite well compared to more sophisticated classifiers, even in view of inaccuracies in their parameters. In this paper, we study the effects of such parameter inaccuracies by investigating the sensitivity functions of a naive Bayesian ..."
Abstract
 Add to MetaCart
Empirical evidence shows that naive Bayesian classifiers perform quite well compared to more sophisticated classifiers, even in view of inaccuracies in their parameters. In this paper, we study the effects of such parameter inaccuracies by investigating the sensitivity functions of a naive Bayesian network. We show that, as a consequence of the network’s independence properties, these sensitivity functions are highly constrained. We further investigate whether the patterns of sensitivity that follow from these functions support the observed robustness of naive Bayesian classifiers. In addition to standard sensitivities given available evidence, we also study the effect of parameter inaccuracies in view of scenarios of additional evidence. We show that standard sensitivity functions suffice to describe such scenario sensitivities.
A Fast Way to Produce Optimal FixedDepth Decision Trees
"... Decision trees play an essential role in many classification tasks. In some circumstances, we only want to consider fixeddepth trees. Unfortunately, finding the optimal depthd decision tree can require time exponential in d. This paper presents a fast way to produce a fixeddepth decision tree that ..."
Abstract
 Add to MetaCart
Decision trees play an essential role in many classification tasks. In some circumstances, we only want to consider fixeddepth trees. Unfortunately, finding the optimal depthd decision tree can require time exponential in d. This paper presents a fast way to produce a fixeddepth decision tree that is optimal under the Naïve Bayes (NB) assumption. Here, we prove that the optimal depthd feature essentially depends only on the posterior probability of the class label given the tests previously performed, but not on either the identity nor the outcomes of these tests. We can therefore precompute, in a fast preprocessing step, which features to use at the final layer. This results in a speedup of O(n / log n), where n is the number of features. We apply this technique to learning fixeddepth decision trees from standard datasets from the UCI repository, and find this model improves the computational cost significantly. Surprisingly, this approach still yields relatively high classification accuracy, despite the NB assumption. 1