Spectral Graph Clustering under Differential Privacy: Balancing Privacy, Accuracy, and Efficiency
2510.07136v1
cs.IT, cs.CR, cs.LG, cs.SI, math.IT
2025-10-10
Авторы:
Mohamed Seif, Antti Koskela, H. Vincent Poor, Andrea J. Goldsmith
Abstract
We study the problem of spectral graph clustering under edge differential
privacy (DP). Specifically, we develop three mechanisms: (i) graph perturbation
via randomized edge flipping combined with adjacency matrix shuffling, which
enforces edge privacy while preserving key spectral properties of the graph.
Importantly, shuffling considerably amplifies the guarantees: whereas flipping
edges with a fixed probability alone provides only a constant epsilon edge DP
guarantee as the number of nodes grows, the shuffled mechanism achieves
(epsilon, delta) edge DP with parameters that tend to zero as the number of
nodes increase; (ii) private graph projection with additive Gaussian noise in a
lower-dimensional space to reduce dimensionality and computational complexity;
and (iii) a noisy power iteration method that distributes Gaussian noise across
iterations to ensure edge DP while maintaining convergence. Our analysis
provides rigorous privacy guarantees and a precise characterization of the
misclassification error rate. Experiments on synthetic and real-world networks
validate our theoretical analysis and illustrate the practical privacy-utility
trade-offs.