ThesisPrivacy Preserving Data Collection and Analysis
The progress in hardware and communication technology enables data analyzers to compute with accuracy aggregate information, owing to the great volume of available data. The personal sensitive information that is binded with individual data, renders users reluctant in publishing it. The conflicting requirement of preserving individual confidentiality while at the same time granting partial access to an aggregate value over the data, has been addressed with Privacy Preserving Data Collection and Analysis protocols. However, in order to achieve individual privacy and efficiency, current cryptographic solutions assume honest-but-curious third parties or fully trusted key-dealers to distribute keys, thus restricting the security model and hindering its deployment in a dynamic environment.
In this dissertation we design and analyze several new Privacy Preserving Data Collection and Analysis protocols in order to strengthen the existing security model and to propose new features. We first propose a solution to the problem of privacy preserving clustering by exploiting the inherent properties of a specific similarity detection algorithm. Then, we design a solution that allows an energy supplier to learn more sophisticated statistics, as the time interval of maximum energy consumption without violating individuals’ privacy. Afterwards, we address the problem of data aggregation in a dynamic environment by relaxing existing trust assumptions. Finally, we strengthen the security requirements of existing protocols with a malicious Aggregator which will try to provide bogus results. We show the practicality of our solutions with prototype implementations. The security of our protocols is analyzed in the provably secure framework.