Search papers, labs, and topics across Lattice.
This paper introduces cross-silo person-level differential privacy (XSP-DP), an information-theoretic framework that addresses the de-anonymization risk when a person's data is distributed across multiple data silos, each protected by local differential privacy. The authors establish a phase transition threshold at k* = 螛(log n / 蔚虏), beyond which adversaries can successfully identify individuals, highlighting the necessity of coordination among silos to mitigate risks. By demonstrating that individual outputs are uninformative yet collectively informative, the study reveals critical insights into the interplay between privacy guarantees and de-anonymization threats in multi-silo environments.
De-anonymization becomes inevitable once the number of data silos exceeds a critical threshold, underscoring the urgent need for coordinated privacy measures.
When a person's records appear in k independent data silos, each protected by (epsilon, delta)-differential privacy, standard composition yields a valid (k*epsilon, k*delta)-DP guarantee for the joint output. This worst-case bound, however, does not answer the concrete inference question: at what k can an adversary actually identify a target person? This paper develops the information-theoretic framework needed to answer that question. We introduce cross-silo person-level DP (XSP-DP), a Pufferfish-style privacy notion whose adjacency relation captures all records of a single person across all silos simultaneously, and verify that the standard basic composition bound carries over to this adjacency model. Within this framework we prove that de-anonymization undergoes a phase transition at k* = Theta(log n / epsilon^2) (population size n, per-silo RR parameter epsilon): a Fano lower bound shows any estimator fails for k << k*, while a matching maximum-likelihood upper bound shows the attack succeeds for k >> k*. An explicit XOR + randomized-response construction demonstrates information synergy: each silo's output is individually uninformative about the target, yet the joint mutual information is strictly positive. For non-coordinated binary randomized-response mechanisms, we prove that de-anonymization is inevitable once k exceeds the threshold, establishing that cross-silo coordination is necessary. These results provide a baseline threat model and Theta-level threshold for cross-silo inference attacks under local DP.
