Brief Announcement: Privacy Preserving Mining of Distributed Data Using a Trusted and Partitioned Third Party

摘要

We show in this paper how one can use the new PTTP architecture proposed by Sherman et al. to create simple algorithms like union/intersection over distributed database, without the need for strong cryptography techniques or the use of hash functions. Using PTTP also results in less communication rounds and in some cases also reduces the size of the messages. In most cases the new algorithms can also protect against malicious coalitions. That said, more research is still needed in order to shift more privacy preserving responsibilities from the two parts of the the PTTP back to the database holders, so that even coalition which involve both parts of the PTTP won't allow to reveal significant information. Finally, more research is also needed to show the utility of the PTTP architecture for other data mining tasks such as: clustering or decision trees construction.