Privacy protection in government data sharing: an improved LDP-based approach

Governments own various types and large amounts of individual data. One governmental department manages specific areas of data. To develop smart government, data need to be shared among the governmental departments. However, how to prevent potential attackers from getting private information in data sharing is a challenging problem. To protect private information while sharing statistics among government departments, an improved LDP-based (local differential privacy) approach is proposed. This approach combines the data binning technique with the count mean sketch (CMS) algorithm. Equi-width binning is adopted to divide the data records into smaller data domains to overcome the problem of large statistical errors in the current privacy protection algorithms with large data domain size and small amounts of data. Then, the proposed algorithm is compared with the CMS and HCMS algorithms from different aspects such as frequency estimation, data size, privacy budget, and data domain size. Experimental results show that the proposed algorithm effectively reduces statistical errors and enhances the utility of data after privacy protection with both various distributions and data domain sizes.