Improving Opinion Analysis Through Statistical Disclosure Control in eVoting Scenarios.

This work addresses the problem of Statistical Disclosure Control (SDC) on an electronic voting scenario. Electoral datasets containing voting choices linked to voters demographic profile information, can be used to perform fine-grained analysis of citizen opinion. However, it is strongly required to protect voters' privacy. Traditional SDC techniques study methods to met some predefined privacy criteria, assuming a trustworthy owner that knows the values of the confidential attributes. Unfortunately, this assumption cannot be made in our scenario, since its dataset contains secret voting choices, which are unknown until they are properly anonymized and decrypted. We propose a protocol and a system architecture to perform SDC in datasets with encrypted attributes, while minimizing the amount of information an attacker can learn about the secret data. The protocol enables the release of electoral datasets, which allow governments and third parties to gain more insight into citizen opinion, and improve decision making processes and public services.