Privacy-Preserving Data Publishing And Utility Verification With Multiple Providers

A number of service providers may want to collaboratively publish their accumu-lated data which can be mined by the other research organizations.However,the data is sensitive as it involves the users' private information.Such that it is necessary to sanitize it before publishing for avoiding leaking privacy to the miners.While privacy protection is on the sacrifice of some data utility which has very significant effect on the mining result.Consequently,the miners or sometimes the providers want to ver-ify the data utility and determine whether it is consistent as claimed.Apparently,the verification cannot compromise the goal of privacy preservation.This paper firstly focuses on how to protect the privacy when publishing data in the context of multiple data providers with insider adversaries.The solutions in literature either rely on a trusted-third party(TTP)or is based on secure multi-party computation.Although the TTP-based schemes are very efficient,TTP does not always exist in real world.While the SMC-based ones do not rely on any TTP but consume much more time.A practical scheme which does not rely on any TTP is presented in this paper.The novel scheme consists of two phases in which the publisher respectively aggregate insensitive and sensitive data.Firstly,the data owners submit the insensitive data directly to the publisher who will carry out the sanitization.Then,to protect the sensitive data,the owners encrypt it and send it to m + 1 randomly chosen decipher who are responsible for decrypting it and mixing it.Finally,the publisher integrates the insensitive and sensitive data and publishes them.As for privacy-preserving data publishing schemes,the data providers or the min-ers may need to verify the data utility.It is difficult to target this because the utility is defined on both the sanitized data and the original data which cannot be leaked.In this paper,a novel scheme based on differential privacy is proposed to verify the utility of the data collaboratively published by multiple data owners.The scheme requires the publisher to provide encrypted statistical information of the original data besides the sanitized data.Then the owners determines whether the encrypted statistical dataset correctly involves their data.With verification-passed auxiliary dataset,the verifier fi-nally figures out the utility of the published data and determines whether it is consistent with the claimed one.Theoretical analysis demonstrates that the two-phase scheme implementing m-k-anonymity and the utility computing scheme will not compromise the individuals'privacy,while simultaneously a series of experiments on benchmarks illustrates that their actual time cost are reasonable.