Privacy-Preserving Collaborative Data Anonymization with Sensitive Quasi-Identifiers

Abstract

Collaborative anonymization deals with a group of respondents in a distributed environment. Unlike in centralized settings, no respondent is willing to reveal his or her records to any party due to the privacy concerns. This creates a challenge for anonymization, and it requires a level of trust among respondents. In this paper, we study a collaborative anonymization protocol that aims to increase the confidence of respondents during data collection. Unlike in existing works, our protocol does not reveal the complete set of quasi-identifier (QID) to the data collector (e.g., agency) before and after the data anonymization process. Because QID can be both sensitive values and identifying values, we allow the respondents to hide sensitive-QID attributes from other parties. Our protocol ensures that the desired protection level (i.e., k-anonymity) can be verified before the respondents submit their records to the agency. Furthermore, we allow honest respondents to indict a malicious agency if it modifies the intermediate results or not following the protocol faithfully.