A Study of Negative Surveys with Background Knowledge

Negative survey is a privacy-preserving model for collecting sensitive data. In real-world applications, we usually have some background knowledge about the survey or the participants. Moreover, it is also reasonable to collect some non-sensitive information about the survey or participants and use this information as background knowledge in the negative survey to estimate more accurate aggregated results. However, most of previous works have not investigated the usage of background knowledge in negative surveys. In this paper, we investigate the usage of a special kind of background knowledge in the negative survey. Specifically, the background knowledge is the range of the sum of the aggregated results for multiple categories, which is presented as an interval. We introduce a method for estimating aggregated results from the data collected from the negative survey with the background knowledge. Experimental results show that more accurate aggregated results could be estimated by using the background knowledge and our method is effective.