Analyzing Randomized Response Mechanisms Under Differential Privacy

The randomized response technique was first introduced by Warner in 1965 [27] as a technique to survey sensitive questions. Since it is considered to protect the respondent's privacy, many variants and applications have been proposed in the literature. Unfortunately, the randomized response and its variants have not been well evaluated from the privacy viewpoint historically. In this paper, we evaluate them by using differential privacy. Specifically, we show that some variants have a tradeoff between the privacy and utility, and that the "negative" survey technique obtains negative results.