m privacy for collaborative data publishing ppt
Posts: 14,118
Threads: 61
Joined: Oct 2014
The problem of collaborative publication of data to anonymize horizontally divided data in multiple data providers. We consider a new type of "insider trading" by the collusion of data providers who can use their own data records (a subset of the general data) in addition to external background knowledge to infer data records from other providers of data. The document addresses this new threat and makes several contributions. First, we introduce the notion of m-privacy, which ensures that anonymous data satisfies a privacy restriction given against any group of up to m collusion data providers. Second, we present heuristic algorithms that exploit the monotony of the group of equivalence of privacy restrictions and adaptive management techniques for the efficient verification of m-privacy given a set of records. Finally, we present an anonymization algorithm with knowledge of the data provider with adaptive privacy verification strategies to ensure the high utility and privacy of anonymous data with efficiency. Experiments on real-life data sets suggest that our approach achieves better or comparable utility and efficiency than existing and basic algorithms while providing a guarantee of privacy.