Publishing Node Strength Distribution With Node Differential Privacy

Liu, Ganghong; Ma, Xuebin; Li, Wuyungerile

doi:10.1109/access.2020.3040077

Cited by 8 publications

(4 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…when spatio-temporal data is published [22]). As a result, there exist two concievable routes to perform private learning on such data: edge-level DP to protect the connections to other individuals in the graph and prevent unique identification of users like in [55,73,78,92,100] and node-level DP to protect the data of each individual itself (as well as the outgoing edges) like in [8,37,70,131]. Furthermore, Sajadmanesh et al [98] utilise locally differentially private GNNs in the context of social networks.…”

Section: Application Areas For Dp On Graphsmentioning

confidence: 99%

Differentially Private Guarantees for Analytics and Machine Learning on Graphs: A Survey of Results

Mueller,

Usynin,

Paetzold

et al. 2024

JPC

View full text Add to dashboard Cite

We study the applications of differential privacy (DP) in the context of graph-structured data and discuss the formulations of DP applicable to the publication of graphs and their associated statistics as well as machine learning on graph-based data, including graph neural networks (GNNs). Interpreting DP guarantees in the context of graph-structured data can be challenging, as individual data points are interconnected (often non-linearly or sparsely). This connectivity complicates the computation of individual privacy loss in differentially private learning. The problem is exacerbated by an absence of a single, well-established formulation of DP in graph settings. This issue extends to the domain of GNNs, rendering private machine learning on graph-structured data a challenging task. A lack of prior systematisation work motivated us to study graph-based learning from a privacy perspective. In this work, we systematise different formulations of DP on graphs, discuss challenges and promising applications, including the GNN domain. We compare and separate works into graph analytics tasks and graph learning tasks with GNNs. We conclude our work with a discussion of open questions and potential directions for further research in this area.

show abstract

Section: Application Areas For Dp On Graphsmentioning

confidence: 99%

Differentially Private Guarantees for Analytics and Machine Learning on Graphs: A Survey of Results

Mueller,

Usynin,

Paetzold

et al. 2024

JPC

View full text Add to dashboard Cite

show abstract

“…when spatio-temporal data is published [84]). As a result, there exist two sensible routes to perform private learning on such data: edge-level DP to protect the connections to other individuals in the graph and prevent unique identification of users like in [28,20,29] and node-level DP to protect the data of each individual itself (as well as the outgoing edges) like in [27,32,33,47,56,57]. Numerous works have previously been employed to allow private release of social graphs or their associated statistics [32,42,50].…”

Section: Social Networkmentioning

confidence: 99%

SoK: Differential Privacy on Graph-Structured Data

Mueller¹,

Usynin²,

Paetzold³

et al. 2022

Preprint

View full text Add to dashboard Cite

In this work, we study the applications of differential privacy (DP) in the context of graph-structured data. We discuss the formulations of DP applicable to the publication of graphs and their associated statistics as well as machine learning on graph-based data, including graph neural networks (GNNs). The formulation of DP in the context of graph-structured data is difficult, as individual data points are interconnected (often non-linearly or sparsely). This connectivity complicates the computation of individual privacy loss in differentially private learning. The problem is exacerbated by an absence of a single, well-established formulation of DP in graph settings. This issue extends to the domain of GNNs, rendering private machine learning on graph-structured data a challenging task. A lack of prior systematisation work motivated us to study graph-based learning from a privacy perspective. In this work, we systematise different formulations of DP on graphs, discuss challenges and promising applications, including the GNN domain. We compare and separate works into graph analysis tasks and graph learning tasks with GNNs. Finally, we conclude our work with a discussion of open questions and potential directions for further research in this area.

show abstract

“…Any addition or deletion of one node has little effect on the query results. Liu et al [22] studied node strength distribution based on node differential privacy, reducing sensitivity by limiting weight and degree. Qian et al [23] proposed a privacy node strength publishing method based on edge differential privacy.…”

Section: Related Workmentioning

confidence: 99%

“…􏼈 􏼉 [32,33], and the value of privacy budget ε is between 0.5 and 1.5 [22,23]. Due to the randomness of Laplacian noise and in order to better measure the algorithm performance, for each algorithm, we use the mean error of 100 runs.…”

Section: Data Sets and Settingsmentioning

confidence: 99%

Publishing Triangle Counting Histogram in Social Networks Based on Differential Privacy

Tang

et al. 2021

Security and Communication Networks

View full text Add to dashboard Cite

The continuous expansion of the number and scale of social networking sites has led to an explosive growth of social network data. Mining and analyzing social network data can bring huge economic value and social benefits, but it will result in privacy leakage and other issues. The research focus of social network data publishing is to publish available data while ensuring privacy. Aiming at the problem of low data availability of social network node triangle counting publishing under differential privacy, this paper proposes a privacy protection method of edge triangle counting. First, an edge-removal projection algorithm TSER based on edge triangle count sorting is proposed to obtain the upper bound of sensitivity. Then, two edge triangle count histogram publishing methods satisfying edge difference privacy are given based on the TSER algorithm. Finally, experimental results show that compared with the existing algorithms, the TSER algorithm can retain more triangles in the original graph, reduce the error between the published data and the original data, and improve the published data availability.

show abstract

Publishing Node Strength Distribution With Node Differential Privacy

Cited by 8 publications

References 17 publications

Differentially Private Guarantees for Analytics and Machine Learning on Graphs: A Survey of Results

Differentially Private Guarantees for Analytics and Machine Learning on Graphs: A Survey of Results

SoK: Differential Privacy on Graph-Structured Data

Publishing Triangle Counting Histogram in Social Networks Based on Differential Privacy

Contact Info

Product

Resources

About