Visualizing differentially private mechanisms with physical cards

Card-based cryptography studies the problem of implementing cryptographic algorithms in a visual way using physical cards to demonstrate their security properties for those who are unfamiliar with cryptography. In this paper, we initiate the study of card-based implementations of differentially private mechanisms, which are a standard privacy-enhancing technique to publish statistics of databases. We start with giving the definition of differential privacy of card-based protocols. As a feasibility result, we present three kinds of protocols using standard binary cards for computing the sum of parties' binary inputs, $f(x_1,\dots ,x_n)={\sum }_{i=1}^n{x}_i$ for

, under differential privacy. Our first protocol follows the framework of output perturbation, which provides differential privacy by adding noise to exact aggregation results. The protocol needs only two shuffles, and the overheads in the number of cards and the error bound are independent of the number n of parties. Our second and third protocols are based on Randomized Response, which adds noise to each input before aggregation. Compared to the first protocol, they improve the overheads in the number of cards and the error bound in terms of differential privacy parameters. Furthermore, we show that our card-based noise generation method can be used to compute general functions beyond the binary sum under differential privacy. To address a technical challenge of generating non-uniform noise using a finite number of cards, we propose a novel differentially private mechanism based on the hypergeometric distribution, which we believe may be of independent interest beyond applications to card-based cryptography.