Abstract: Nowadays, tremendous amount of data is created and distributed to different repositories. With reduction in cost of storing information and existing infrastructures such as cloud computing and grid computing, there is the opportunity to extract knowledge or confidential data from that resources. In this topic the privacy preserving in Data mining works with Bi-Party Data Release Method. The exponential mechanism chooses a candidate that is close to optimum with respect to a utility function while preserving differential privacy. In the distributed setting, the candidates are owned by two parties and, therefore, a secure mechanism is required to compute the same output while ensuring that no extra information is leaked to any party. The proposed distributed exponential mechanism takes (candidate, score) pairs as inputs. The score is calculated using a utility function. The proposed distributed exponential mechanism is therefore independent of the choice of the utility function. In the case of vertically-partitioned data, we can use two types of utility functions: First, utility functions such as information gain, maximum function, and the widest (normalized) range of values that can be calculated locally by each party or Second, utility functions that cannot be computed locally. In the latter case, secure function evaluation techniques can be used by the parties to compute these utility functions. Once the scores of the candidates are computed using the utility functions in either case, they are ready to be used as inputs to execute the distributed exponential mechanism. The third party can request the data on the basis of anonymity and can view the data with the digital signatures provided by the both parties.
Keywords: Privacy Preserving, Anonymity, Utility Function
|
DOI:
10.17148/IJARCCE.2018.71017
