DYNAMIC NON-COOPERATIVE COMPUTATION FOR PRIVACY PRESERVING DATA ANALYSIS

Data mining has been around for many years that can help organizations to discover actionable knowledge. When data is available with multiple data providers, it is useful practice to have privacy preserving collaborative data analysis in distributed environment. Many techniques came into existence in order to achieve this. Recently Kantarcioglu et al. proposed incentive compatible approach that motivates competing parties to provide genuine data instead of giving less than ideal or incompatible data. In this paper we implement a mechanism and built a prototype application that allows multiple parties to have secure communication in distributed environment. The parties are able to log into the system and provide data. It does mean that multiple parties can provide compatible data. The system verifies the data and comes to know whether it is genuine or not. When data provided is not compatible the system deducts the incentives and informs the party to provide data again. When valid data is provided, the incentives will be increased again. This technique along with secure multi-party computations makes the proposed system very useful for privacy preserving collaborative data analysis. The empirical results are encouraging.