Crypto-Assisted Graph Degree Sequence Release under Local Differential Privacy

TL;DR

This paper introduces CADR-LDP, a privacy-preserving framework that accurately releases graph degree sequences under local differential privacy by optimizing parameter selection and edge addition, outperforming existing methods.

Contribution

The paper proposes CADR-LDP, a novel framework combining encryption and private mechanisms to improve degree sequence release accuracy under local differential privacy constraints.

Findings

CADR-LDP satisfies $$-node local differential privacy.

Experimental results show superior performance over existing methods.

Optimal--Selection reduces communication cost.

Abstract

Given a graph $G$ defined in a domain $\mathcal{G}$, we investigate locally differentially private mechanisms to release a degree sequence on $\mathcal{G}$ that accurately approximates the actual degree distribution. Existing solutions for this problem mostly use graph projection techniques based on edge deletion process, using a threshold parameter $\theta$ to bound node degrees. However, this approach presents a fundamental trade-off in threshold parameter selection. While large $\theta$ values introduce substantial noise in the released degree sequence, small $\theta$ values result in more edges removed than necessary. Furthermore, $\theta$ selection leads to an excessive communication cost. To remedy existing solutions' deficiencies, we present CADR-LDP, an efficient framework incorporating encryption techniques and differentially private mechanisms to release the degree sequence. In CADR-LDP, we first use the crypto-assisted Optimal-$\theta$-Selection method to select the optimal parameter with a low communication cost. Then, we use the LPEA-LOW method to add some edges for each node with the edge addition process in local projection. LPEA-LOW prioritizes the projection with low-degree nodes, which can retain more edges for such nodes and reduce the projection error. Theoretical analysis shows that CADR-LDP satisfies $\epsilon$-node local differential privacy. The experimental results on eight graph datasets show that our solution outperforms existing methods.