Research On Metric For Strength Of Privacy Protection

As the privacy protection issue in data sharing has been continuously attracted by researchers,a variety of different privacy protection methods are proposed.The protection effects provided by different privacy protection methods are different.The unified metric of protection is the basis of privacy protection effect evaluation.Aiming at the problem that the existed privacy protection strength metrics methods are not applicable to metric the provenance and stream histogram privacy protection strength,a provenance privacy protection metric method based on min-entropy and a stream histogram privacy protection metric method based on Bayesian are proposed.The main contributions are as follows:(1)Aiming at the problem that the attacker has a single background knowledge and poor adaptability to the diversified structure of the existed privacy protection strength measurement method,a min-entropy-based metric provenance privacy protection is proposed.The function of node module is described by labeling method.The sub-structure features of l-prehop matrix and l-posthop matrix describing module are proposed to obtain the feature information of inferred provenance which is constructed by hidden provenance and background knowledge;further,we establish the relationship between the module in original provenance and the feature information by reclassifying the module according to the feature information.Finally,the protection strength provided by provenance privacy protection algorithm is measured by introducing the min-entropy mechanism.(2)Aiming at the problem that evaluating the overall distribution of data and strong correlation between background knowledge and privacy information in traditional privacy protection strength measurement method,a Bayesian-based stream histogram privacy protection metric method is proposed.By analyzing the association between background knowledge and sliding window,the association between background knowledge and published result is established.The concept of the associated histogram is proposed to surmount the association between histograms which have state information of the same user.Meanwhile,the measurement mechanism of privacy leakage of associated histograms oriented to sliding window to realize the measurement of privacy leakage of associated histograms;further,we set weights for error probability of attacks after acquiring background knowledge and privacy leakage to realize the measure for stream histogram protection algorithm.Theoretical analysis and experimental results show that the proposed metric can effectively measure the privacy protection strength of provenance and stream histogram privacy protection method.