Mutual Information Bounds in the Shuffle Model
By: Pengcheng Su, Haibo Cheng, Ping Wang
Potential Business Impact:
Keeps your private messages secret from others.
The shuffle model enhances privacy by anonymizing users' reports through random permutation. This paper presents the first systematic study of the single-message shuffle model from an information-theoretic perspective. We analyze two regimes: the shuffle-only setting, where each user directly submits its message ($Y_i=X_i$), and the shuffle-DP setting, where each user first applies a local $\varepsilon_0$-differentially private mechanism before shuffling ($Y_i=\mathcal{R}(X_i)$). Let $\boldsymbol{Z} = (Y_{σ(i)})_i$ denote the shuffled sequence produced by a uniformly random permutation $σ$, and let $K = σ^{-1}(1)$ represent the position of user 1's message after shuffling. For the shuffle-only setting, we focus on a tractable yet expressive \emph{basic configuration}, where the target user's message follows $Y_1 \sim P$ and the remaining users' messages are i.i.d.\ samples from $Q$, i.e., $Y_2,\dots,Y_n \sim Q$. We derive asymptotic expressions for the mutual information quantities $I(Y_1;\boldsymbol{Z})$ and $I(K;\boldsymbol{Z})$ as $n \to \infty$, and demonstrate how this analytical framework naturally extends to settings with heterogeneous user distributions. For the shuffle-DP setting, we establish information-theoretic upper bounds on total information leakage. When each user applies an $\varepsilon_0$-DP mechanism, the overall leakage satisfies $I(K; \boldsymbol{Z}) \le 2\varepsilon_0$ and $I(X_1; \boldsymbol{Z}\mid (X_i)_{i=2}^n) \le (e^{\varepsilon_0}-1)/(2n) + O(n^{-3/2})$. These results bridge shuffle differential privacy and mutual-information-based privacy.
Similar Papers
Bayesian Advantage of Re-Identification Attack in the Shuffle Model
Cryptography and Security
Protects secrets by scrambling messages before sharing.
Augmented Shuffle Protocols for Accurate and Robust Frequency Estimation under Differential Privacy
Cryptography and Security
Protects private data from hackers and cheaters.
Decomposition-Based Optimal Bounds for Privacy Amplification via Shuffling
Cryptography and Security
Protects your data better when shared online.