English

Breaking the $n^{1.5}$ Additive Error Barrier for Private and Efficient Graph Sparsification via Private Expander Decomposition

Data Structures and Algorithms 2025-07-03 v1

Abstract

We study differentially private algorithms for graph cut sparsification, a fundamental problem in algorithms, privacy, and machine learning. While significant progress has been made, the best-known private and efficient cut sparsifiers on nn-node graphs approximate each cut within O~(n1.5)\widetilde{O}(n^{1.5}) additive error and 1+γ1+\gamma multiplicative error for any γ>0\gamma > 0 [Gupta, Roth, Ullman TCC'12]. In contrast, "inefficient" algorithms, i.e., those requiring exponential time, can achieve an O~(n)\widetilde{O}(n) additive error and 1+γ1+\gamma multiplicative error [Eli{\'a}{\v{s}}, Kapralov, Kulkarni, Lee SODA'20]. In this work, we break the n1.5n^{1.5} additive error barrier for private and efficient cut sparsification. We present an (ε,δ)(\varepsilon,\delta)-DP polynomial time algorithm that, given a non-negative weighted graph, outputs a private synthetic graph approximating all cuts with multiplicative error 1+γ1+\gamma and additive error n1.25+o(1)n^{1.25 + o(1)} (ignoring dependencies on ε,δ,γ\varepsilon, \delta, \gamma). At the heart of our approach lies a private algorithm for expander decomposition, a popular and powerful technique in (non-private) graph algorithms.

Keywords

Cite

@article{arxiv.2507.01873,
  title  = {Breaking the $n^{1.5}$ Additive Error Barrier for Private and Efficient Graph Sparsification via Private Expander Decomposition},
  author = {Anders Aamand and Justin Y. Chen and Mina Dalirrooyfard and Slobodan Mitrović and Yuriy Nevmyvaka and Sandeep Silwal and Yinzhan Xu},
  journal= {arXiv preprint arXiv:2507.01873},
  year   = {2025}
}

Comments

ICML 2025

R2 v1 2026-07-01T03:43:32.370Z