Pith. sign in

REVIEW 4 major objections 7 minor 56 references

PrivDPR: Synthetic Graph Publishing with Deep PageRank under Differential Privacy

T0 review · 4 major / 7 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that a deep PageRank network with weight-normalized layers can synthesize graphs under node-level differential privacy by trading network depth for privacy noise.

desk verdict Interesting and fixable: the depth-sensitivity mechanism is real, but the privacy proof has a factor-of-2 gap and the utility claims outrun the evidence. read the letter →

arxiv 2501.02354 v1 pith:YSPRYIUL submitted 2025-01-04 cs.DB cs.CR

Sen Zhang , Haibo Hu , Qingqing Ye , Jianliang Xu This is my paper · ORCID
classification cs.DBcs.CR
keywords differentialprivacysyntheticgraphpublishingPageRanknode-levelgradientperturbationweightnormalizationsynthesis
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proposes PrivDPR, a method for publishing synthetic graphs under node-level differential privacy. Instead of clipping and noising every gradient, PrivDPR adds Gaussian noise only to the gradient of the node embedding matrix in a deep PageRank network, and uses weight normalization to bound that gradient. The bounding argument shows the gradient norm falls as $(1/s)^{L+1}$ with the number of layers $L$, so the desired sensitivity can be preset and extra layers absorb the cost of splitting the privacy budget across iterations. The paper proves that the released graph satisfies $(\varepsilon,\delta)$-node-level DP and reports experiments where its structural utility stays competitive even at $\varepsilon = 0.1$.

What carries the argument

The central objects are the deep PageRank objective and the Cauchy-Schwarz upper bound $\mathcal{L}(v_i,v_j;\Theta)$ from Lemma 4.1, which replaces the non-decomposable PageRank loss with an edge-wise training loss. Weight normalization $W/(\|sW\|_2)$ with $s>1$ is the mechanism that makes each layer shrink the gradient bound by a factor $1/s$, so the full bound $M(1/s)^{L+1}$ follows from the chain-rule/Lipschitz estimate of the network. The layer count $L$ is then chosen from the log-inequality $\log\left(\frac{S_\nabla}{BM}\right)/\log\left(\frac{1}{s}\right)-1 \le L$, turning network depth into the privacy knob.

What would settle it

Take a fixed graph, fix $s$ and the batch size, train PrivDPR with $L$ chosen from Eq. (12), and measure the empirical $\ell_2$ norm of the gradient of the edge-wise loss with respect to $V$ over training; if it does not stay at or below the preset $S_\nabla$, the sensitivity bound in Theorem 4.2 is false. Alternatively, compute the PageRank vectors of the published graph and of the input graph; if their Spearman correlation is near zero while the training loss has converged, the surrogate objective is not preserving the structure the method claims.

Watch

Extended reading notes

Core claim

At the center of the paper is Theorem 4.2, which bounds the $\ell_2$ norm of $\partial \mathcal{L}(v_i,v_j;\Theta)/\partial V$ by $M(1/s)^{L+1}$ for a weight-normalized deep PageRank with normalization $W/\|sW\|_2$ and sigmoid activations. Choosing the number of layers according to $\log\left(\frac{S_\nabla}{BM}\right)/\log\left(\frac{1}{s}\right)-1 \le L$ therefore makes the batch gradient's sensitivity at most the preset value $S_\nabla$, and the same inequality with $S_\nabla$ replaced by $S_\nabla/T$ counters the noise increase from splitting $\varepsilon$ across $T$ gradient updates. Theorem 5.1 then assembles node-level DP from the per-iteration Gaussian mechanism, sequential composition, and post-processing. The paper's claim is that this depth-for-privacy trade-off, not a stronger composition calculus, is what lets a private deep graph generator work under small privacy budgets.

Load-bearing premise

The training minimizes an upper-bound surrogate for the PageRank objective, and the paper explicitly says the proof that this bound approximates the objective is future work; if that surrogate does not preserve PageRank structure, the synthetic graph's utility claim fails even though the DP guarantee may hold.

Editorial extensions

If this is right

  • Node-level synthetic graph release with $(\varepsilon,\delta)$-DP is achievable at privacy budgets as small as $\varepsilon=0.1$ without moments accountant style composition.
  • Adding hidden layers lowers the required Gaussian noise, so a user can preset a target sensitivity and size the network accordingly.
  • Because DP is preserved under post-processing, the published graph can feed arbitrary downstream queries without extra privacy cost.
  • The training loop runs in time linear in the number of nodes for fixed batch and walk parameters, so the method is claimed to scale to large graphs.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The privacy theorem only guarantees that the released node embedding matrix is close in distribution; whether that embedding captures PageRank structure rests on the unproven approximation ratio of the upper-bound loss, so a natural test is to compare node rankings from the synthetic graph with true PageRank values.
  • The exponential drop $(1/s)^{L+1}$ suggests the gradient can vanish for large $L$; the method may inherit vanishing-gradient behavior and need residual connections or skip layers, which the paper does not discuss.
  • One could apply the same weight-normalization-and-depth argument to other random-walk embedding objectives, not just PageRank, to obtain node-level DP for embeddings more generally.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. PrivDPR proposes a node-level differentially private synthetic graph publishing method based on a deep PageRank objective. The algorithm learns a node embedding matrix V by adding Gaussian noise to the gradient of V during training, while using weight normalization to bound the per-edge gradient norm by M(1/s)^{L+1}. The paper's central theoretical claims are that increasing the number of network layers can meet any preset sensitivity S∇ and mitigate privacy budget splitting, and that the final synthetic graph satisfies (ε,δ)-node-level DP. The empirical section compares PrivDPR with GAP, DPGGAN, DPGVAE, and a non-private variant on five datasets across eight structural metrics and two downstream tasks.

Significance. If the privacy proof were fully correct, the depth-versus-sensitivity trade-off would be a useful addition to the DP graph synthesis toolbox, and the paper's explicit targeting of node-level DP (rather than the weaker edge-level DP) is valuable. The manuscript is clearly written, the experiments are broad, and the authors provide a public code repository. However, the load-bearing privacy analysis contains a factor-of-2 sensitivity error, the data-dependent random-walk batch construction is not accounted for in the proof, and the absolute utility numbers are modest; the current version therefore does not yet substantiate the claimed privacy guarantee or the abstract's 'high data utility' statement.

major comments (4)
  1. [Sec. 4.3.1, Eq. (11); Sec. 5.1, Theorem 5.1] The sensitivity of the batch gradient sum is underestimated by a factor of 2. Theorem 4.2 gives ||∂L(vi,vj;Θ)/∂V||_2 ≤ ρ = M(1/s)^{L+1} for each edge, and Eq. (11) sets Bρ ≤ S∇ by claiming that the worst-case norm of the sum over B affected gradients is at most Bρ. Under the bounded node DP adopted in Definition 3, two neighboring graphs both contribute B edges to the batch: for each edge, the gradient in G and the gradient in G′ each have norm at most ρ, so by the triangle inequality the per-edge difference is at most 2ρ and the sensitivity of the sum is at most 2Bρ, not Bρ. Consequently, the Gaussian noise in Theorem 5.1 is scaled too small for the stated (ε,δ)-node-level DP guarantee. The error is fixable by requiring S∇ ≥ 2Bρ (or increasing L by one layer), but as written the central privacy claim is not established.
  2. [Sec. 5.1, Theorem 5.1; Algorithm 2, lines 4–7] The privacy proof does not account for the data-dependent random-walk batch construction. In Algorithm 2, the batch E_B is generated by performing random walks on the input graph (lines 5–7), so the distribution over batches is itself a function of the private graph. Theorem 5.1 only analyzes the Gaussian mechanism for a fixed batch, i.e., it shows that the noisy sum of gradients is DP conditioned on a batch. However, a mixture of DP mechanisms with data-dependent mixture weights is not necessarily DP. To complete the proof, the authors need either to make the batch selection data-independent or to bound the privacy loss of the random-walk sampling itself, for example by a subsampling argument with data-independent probabilities. Without this, the per-iteration DP claim and the composition argument in Theorem 5.1 are not fully justified.
  3. [Sec. 4.2, Lemma 4.1 and following paragraph] The approximation ratio of the surrogate objective is left open. The paper minimizes the edge-wise upper bound L(vi,vj) from Lemma 4.1 in place of the PageRank objective in Eq. (3), and it explicitly states that the proof of the approximation ratio is 'currently a subject for future research'. If this upper bound is loose or does not preserve the minimizer structure of PageRank, then the learned embedding V may not reflect the PageRank properties that motivate the method. The DP guarantee is independent of this issue, but the utility claim and the name 'deep PageRank' depend on it. The authors should either provide a bound on the approximation ratio or clearly qualify the theoretical connection to PageRank.
  4. [Sec. 6, Tables 2–5 and Figure 2] The reported utilities appear too low to support the abstract's claim of 'high data utility'. For example, the mean relative error for triangle count is around 0.99 across all datasets in Tables 2 and 3, the KS distance on Cora and Citeseer is around 0.5–0.6, and the link-prediction AUC of PrivDPR is only about 0.50–0.60 in Tables 4 and 5, which is close to random guessing. While PrivDPR often outperforms the baselines, the absolute numbers suggest that the synthetic graphs preserve only weak structural fidelity. The utility claims should be reworded, and the discussion should address why the absolute errors are this large.
minor comments (7)
  1. [Sec. 6, Baselines] The sentence 'For a fair comparison, we configure GAP to generate synthetic graphs using the same generation method as PrivDPR' is duplicated in Section 6; please remove the duplicate.
  2. [Theorem 4.2] The phrase 'maximum number of layers' should be 'minimum number of layers', since the inequality provides a lower bound on L for a preset sensitivity S∇.
  3. [Sec. 4.3.2] The text says 'we evenly divide privacy parameters ε and σ'; this should likely read 'ε and δ', since σ is computed from ε/T and δ/T via the Gaussian mechanism.
  4. [Eq. (3)] The typesetting of Eq. (3) is difficult to parse; please display the PageRank objective with unambiguous parentheses and define all terms, for example the placement of (1−γ)/N inside the outer parentheses.
  5. [Algorithm 2, line 12 and Appendix A] Algorithm 2 says 'Sample graphs from VV^T to generate score matrix S', but the reconstruction procedure in Appendix A builds a transition count matrix from random walks over the learned embeddings; please reconcile the description so that the reader can reproduce the reconstruction.
  6. [Notation throughout] The batch size is denoted ♭ in Algorithm 2 but B in Eqs. (6)–(7) and Theorem 4.2; please use a consistent symbol.
  7. [Sec. 6.3 and Tables 4–5] The non-private baseline DPR (No DP) is described in Section 6 but does not appear in the link prediction or node classification tables; please either include it or state why it is omitted.

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity; the central DP sensitivity derivation is self-contained. The score reflects only minor non-load-bearing self-citations and non-circular caveats.

full rationale

No circular step is present in the derivation chain. Theorem 4.2 bounds the per-edge gradient norm, ||∂L(vi,vj;Theta)/∂V||_2 <= M(1/s)^(L+1), from Lemma 4.4 and the explicit weight-normalization choice W_WN(W)=W/||sW||_2; this is an input to the privacy proof, not a restatement of the (epsilon,delta) target. Eq. (7) calibrates Gaussian noise to a preset S_∇, and Eq. (11) shows how choosing L makes B M (1/s)^(L+1) <= S_∇; this is a genuine sensitivity analysis rather than a fitted prediction. Theorem 5.1 then invokes only standard sequential composition and post-processing. The authors' self-citations ([11], [44]-[47], [52]) appear only as related-work references and are not load-bearing premises. Two caveats should be weighed separately rather than as circularity: the paper explicitly states 'The proof of this upper bound on the approximation ratio is currently a subject for future research' (Section 4.2), leaving the surrogate objective's fidelity to PageRank unproven; and Eq. (11)'s worst-case bound B M (1/s)^(L+1) bounds the norm of one batch sum, whereas the sensitivity of the difference of two neighboring batch sums under bounded node DP would be at most 2 B M (1/s)^(L+1), so the noise calibration may be off by a factor of two. Both are correctness/utility concerns, not cases where an equation reduces by construction to its own input.

Assumptions & free parameters 8 free parameters · 6 assumptions · 0 invented entities

The central claim rests on the DP composition framework, the Sigmoid boundedness, spectral normalization, and the unproven surrogate objective. The many hyperparameters (s, d, S∇, r, random-walk settings, learning rate, epochs) are chosen by hand or tuned on the evaluation datasets, making the utility claims dependent on those choices.

free parameters (8)
  • Weight normalization parameter s = 8
    Set to 8 in experiments based on robustness of MRE/KSD across s in {2,4,6,8} (Section 6.1, Table 2).
  • Hidden dimension d = 64
    Set to 64 based on robustness across d in {64,128,256,512} (Section 6.1, Table 3).
  • Preset sensitivity S∇ = 5
    Chosen as 'determined based on DPGGAN' (Section 6 parameter settings). Controls the target sensitivity that Theorem 4.2 must achieve.
  • Embedding dimension r = 128
    Set to 128, common in network embedding baselines (Section 6 parameter settings).
  • Random walk parameters Rwn and Rwl = Rwn=2, Rwl=16
    Set to one-fifth of recommended DeepWalk values (Section 6).
  • Batch size ♭ = 16
    Arbitrary choice in parameter settings; affects T and thus the per-iteration budget split.
  • Learning rate η = 1e-3
    Consistent with DPGGAN and DPGVAE settings.
  • Epochs nepochs = 5
    Maximum epochs; privacy budget split over T = nepochs * floor(N/♭).
assumptions (6)
  • standard math Standard DP composition and Gaussian mechanism theorems
    Theorems 2.1-2.3 and the sequential composition used in Theorem 5.1.
  • standard math Cauchy-Schwarz inequality used in Lemma 4.1
    Basis for the edge-wise upper-bound objective L(vi,vj).
  • domain assumption Activation function has bounded derivative and output in [0,1]
    Lemma 4.3 and Lemma 4.4 rely on Sigmoid activation to bound f(v;Θ) by 1 and the Lipschitz constant by 1; the paper claims ReLU can be substituted, but ReLU is not upper-bounded by 1.
  • domain assumption Weight normalization yields spectral norm exactly 1/s
    Section 4.3.1 defines W_WN(W) = W/(∥sW∥_2) and uses ∥W_WN∥_2 = 1/s throughout Theorem 4.2.
  • domain assumption Random walk batch sensitivity is bounded by B times the per-edge gradient bound
    The privacy proof assumes the worst-case effect of a node change is captured by bounding each of B edge gradients; the missing factor of 2 for bounded-DP differences is a flaw in this assumption.
  • ad hoc to paper The edge-wise upper bound approximates the PageRank objective
    Section 4.2 explicitly defers the proof of the approximation ratio to future work, so the surrogate objective is an unproven modeling assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PrivDPR: Synthetic Graph Publishing with Deep PageRank under Differential Privacy." pith.science (2026). https://pith.science/paper/YSPRYIUL

@misc{pith2026250102354,
  author       = {Pith},
  title        = {Pith review of: PrivDPR: Synthetic Graph Publishing with Deep PageRank under Differential Privacy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YSPRYIUL}},
  note         = {Machine review of arXiv:2501.02354}
}
read the original abstract

The objective of privacy-preserving synthetic graph publishing is to safeguard individuals' privacy while retaining the utility of original data. Most existing methods focus on graph neural networks under differential privacy (DP), and yet two fundamental problems in generating synthetic graphs remain open. First, the current research often encounters high sensitivity due to the intricate relationships between nodes in a graph. Second, DP is usually achieved through advanced composition mechanisms that tend to converge prematurely when working with a small privacy budget. In this paper, inspired by the simplicity, effectiveness, and ease of analysis of PageRank, we design PrivDPR, a novel privacy-preserving deep PageRank for graph synthesis. In particular, we achieve DP by adding noise to the gradient for a specific weight during learning. Utilizing weight normalization as a bridge, we theoretically reveal that increasing the number of layers in PrivDPR can effectively mitigate the high sensitivity and privacy budget splitting. Through formal privacy analysis, we prove that the synthetic graph generated by PrivDPR satisfies node-level DP. Experiments on real-world graph datasets show that PrivDPR preserves high data utility across multiple graph structural properties.

Figures

Figures reproduced from arXiv: 2501.02354 by the authors.

Figure 1
Figure 1. Framework of our proposed PrivDPR 4.3 Gradient Perturbation 4.3.1 How to Resist High Sensitivity? To yield a private em￾bedding matrix V, a naive method is to first clip 𝜕L (𝑣𝑖 ,𝑣𝑗 ;𝚯) 𝜕V and then inject noise into this gradient. This results in the following expression: e∇VL ← 1 𝐵 © ­ « ∑︁ (𝑣𝑖 ,𝑣𝑗 ) ∈𝐸𝐵 𝐶𝑙𝑖𝑝  𝜕L (𝑣𝑖 , 𝑣𝑗 ; 𝚯) 𝜕V  + N  S 2 ∇ 𝜎 2 I  ª ® ¬ , (6) where the sensitivity of Í (𝑣𝑖 ,𝑣𝑗 ) ∈𝐸𝐵 𝐶𝑙𝑖𝑝  𝜕L (… view at source ↗
Figure 2
Figure 2. Privacy budget on Cora Xiao et al. encode the graph structure through private edge count￾ing queries under the hierarchical random graph model and report improved results compared to the dK-series approach [42]. Chen et al. employ the exponential mechanism to sample an adjacency matrix after clustering the input graph [3]. Proserpio et al. suggest down-weighting the edges of a graph non-uniformly to mitigate high gl… view at source ↗
Figure 3
Figure 3. Privacy budget on Citeseer [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Privacy budget on p2p 0.1 0.2 0.4 0.8 1.6 3.2 ǫ (a) Chicago, TC 0 5 10 MRE GAP DPGGAN DPGVAE DPR (No DP) PrivDPR 0.1 0.2 0.4 0.8 1.6 3.2 ǫ (b) Chicago, WC 0 20 40 60 MRE 0.1 0.2 0.4 0.8 1.6 3.2 ǫ (c) Chicago, CC 0 1000 2000 3000 4000 MRE 0.1 0.2 0.4 0.8 1.6 3.2 ǫ (d) C…
Figure 5
Figure 5. Figure 5: Privacy budget on Chicago [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 51 canonical work pages

  1. [1]

    Martin Abadi, Andy Chu, Ian Goodfellow, H Brendan McMahan, Ilya Mironov, Kunal Talwar, and Li Zhang. 2016. Deep learning with differential privacy. In ACM SIGSAC Conference on Computer and Communications Security . 308–318

    work page 2016

  2. [2]

    Aleksandar Bojchevski, Oleksandr Shchur, Daniel Zügner, and Stephan Günne- mann. 2018. NetGAN: Generating graphs via random walks. In International Conference on Machine Learning . 610–619

    work page 2018

  3. [3]

    Rui Chen, Benjamin C Fung, Philip S Yu, and Bipin C Desai. 2014. Correlated network data publication via differential privacy. The VLDB Journal 23, 4 (2014), 653–676

    work page 2014

  4. [4]

    Ameya Daigavane, Gagan Madan, Aditya Sinha, Abhradeep Guha Thakurta, Gaurav Aggarwal, and Prateek Jain. 2021. Node-level differentially private graph neural networks. arXiv preprint arXiv:2111.15521 (2021)

    arXiv 2021

  5. [5]

    Lun Du, Xu Chen, Fei Gao, Qiang Fu, Kunqing Xie, Shi Han, and Dongmei Zhang

    work page

  6. [6]

    Cynthia Dwork. 2006. Differential privacy. In International Colloquium on Au- tomata, Languages, and Programming . 1–12

    work page 2006

  7. [7]

    Cynthia Dwork, Frank McSherry, Kobbi Nissim, and Adam Smith. 2006. Cali- brating noise to sensitivity in private data analysis. In Theory of Cryptography Conference. 265–284

    work page 2006

  8. [8]

    Cynthia Dwork and Aaron Roth. 2014. The algorithmic foundations of differential privacy. Foundations and Trends in Theoretical Computer Science. 9, 3–4 (2014), 211–407

    work page 2014

Show all 56 references
  1. [9]

    Marek Eliáš, Michael Kapralov, Janardhan Kulkarni, and Yin Tat Lee. 2020. Dif- ferentially private release of synthetic graphs. In Annual ACM-SIAM Symposium on Discrete Algorithms. 560–578

    2020

  2. [10]

    Alessandro Epasto, Vahab Mirrokni, Bryan Perozzi, Anton Tsitsulin, and Peilin Zhong. 2022. Differentially private graph learning via sensitivity-bounded per- sonalized pagerank. In International Conference Neural Information Processing Systems. 22617–22627

    2022

  3. [11]

    Jie Fu, Qingqing Ye, Haibo Hu, Zhili Chen, Lulu Wang, Kuncan Wang, and Xun Ran. 2024. DPSUR: Accelerating differentially private stochastic gradient descent using selective update and release. In Proceedings of the VLDB Endowment . 1200– 1213

    2024

  4. [12]

    Tianchong Gao and Feng Li. 2019. Preserving persistent homology in differ- entially private graph publications. In IEEE INFOCOM Conference on Computer Communications. 2242–2250

    2019

  5. [13]

    Anupam Gupta, Aaron Roth, and Jonathan Ullman. 2012. Iterative constructions and private data release. In Theory of Cryptography Conference . 339–356

    2012

  6. [14]

    Michael Hay, Chao Li, Gerome Miklau, and David Jensen. 2009. Accurate es- timation of the degree distribution of private networks. In IEEE International Conference on Data Mining . 169–178

    2009

  7. [15]

    Michael Hay, Kun Liu, Gerome Miklau, Jian Pei, and Evimaria Terzi. 2011. Privacy- aware data management in information networks. In ACM SIGMOD Conference on Management of Data . 1201–1204

    2011

  8. [16]

    Haibo Hu, Jianliang Xu, Xizhong Xu, Kexin Pei, Byron Choi, and Shuigeng Zhou

  9. [17]

    Xun Jian, Yue Wang, and Lei Chen. 2021. Publishing graphs under node differen- tial privacy. IEEE Transactions on Knowledge and Data Engineering 35, 4 (2021), 4164–4177

    2021

  10. [18]

    Fan Jiang, Carson K Leung, and Adam GM Pazdor. 2016. Big data mining of social networks for friend recommendation. In IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining . 921–922

    2016

  11. [19]

    Zach Jorgensen, Ting Yu, and Graham Cormode. 2016. Publishing attributed social graphs with formal privacy guarantees. In ACM SIGMOD Conference on Management of Data. 107–122

    2016

  12. [20]

    Thomas N Kipf and Max Welling. 2016. Variational graph auto-encoders. arXiv preprint arXiv:1611.07308 (2016)

    2016 arXiv

  13. [21]

    Yi-An Lai, Chin-Chi Hsu, Wen Hao Chen, Mi-Yen Yeh, and Shou-De Lin. 2017. PRUNE: Preserving proximity and global ranking for network embedding. In International Conference Neural Information Processing Systems . 5257–5266

    2017

  14. [22]

    Jure Leskovec, Daniel Huttenlocher, and Jon Kleinberg. 2010. Signed networks in social media. In International Conference on Human Factors in Computing Systems . 1361–1370

    2010

  15. [23]

    Jure Leskovec, Jon Kleinberg, and Christos Faloutsos. 2007. Graph Evolution: Densification and shrinking diameters.ACM Transactions on Knowledge Discovery From Data 1, 1 (2007), 2–es

    2007

  16. [24]

    Ashwin Machanavajjhala, Daniel Kifer, Johannes Gehrke, and Muthuramakrish- nan Venkitasubramaniam. 2007. l-diversity: Privacy beyond k-anonymity. ACM Transactions on Knowledge Discovery from Data 1, 1 (2007), 3–es

    2007

  17. [25]

    Mir and Rebecca N

    Darakhshan J. Mir and Rebecca N. Wright. 2009. A differentially private graph estimator. In IEEE International Conference on Data Mining Workshops . 122–129

    2009

  18. [26]

    Peter J Mucha and Mason A Porter. 2010. Communities in multislice voting networks. Chaos: An Interdisciplinary Journal of Nonlinear Science 20, 4 (2010)

    2010

  19. [27]

    Iyiola E Olatunji, Thorben Funke, and Megha Khosla. 2023. Releasing graph neural networks with differential privacy guarantees. Transactions on Machine Learning Research (2023), 2835–8856

    2023

  20. [28]

    Lawrence Page, Sergey Brin, Rajeev Motwani, and Terry Winograd. 1998. The PageRank citation ranking: Bring order to the web . Technical Report. Stanford InfoLab

    1998

  21. [29]

    Bryan Perozzi, Rami Al-Rfou, and Steven Skiena. 2014. DeepWalk: Online learning of social representations. In ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 701–710

    2014

  22. [30]

    Davide Proserpio, Sharon Goldberg, and Frank McSherry. 2014. Calibrating data to sensitivity in private data analysis. In Proceedings of the VLDB Endowment . 637–648

    2014

  23. [31]

    Luca Rendsburg, Holger Heidrich, and Ulrike Von Luxburg. 2020. NetGAN without GAN: From random walks to low-rank approximations. In International Conference on Machine Learning . 8073–8082

    2020

  24. [32]

    Sina Sajadmanesh, Ali Shahin Shamsabadi, Aurélien Bellet, and Daniel Gatica- Perez. 2023. GAP: Differentially private graph neural networks with aggregation perturbation. In USENIX Security Symposium. 3223–3240

    2023

  25. [33]

    Alessandra Sala, Xiaohan Zhao, Christo Wilson, Haitao Zheng, and Ben Y Zhao

  26. [34]

    Prithviraj Sen, Galileo Namata, Mustafa Bilgic, Lise Getoor, Brian Galligher, and Tina Eliassi-Rad. 2008. Collective classification in network data. AI Magazine 29, 3 (2008), 93–106

    2008

  27. [35]

    Dushyant Sharma, Rishabh Shukla, Anil Kumar Giri, and Sumit Kumar. 2019. A brief review on search engine optimization. In International Conference on Cloud Computing, Data Science and Engineering . 687–692

    2019

  28. [36]

    Haipei Sun, Xiaokui Xiao, Issa Khalil, Yin Yang, Zhan Qin, Hui Wang, and Ting Yu

  29. [37]

    Latanya Sweeney. 2002. k-anonymity: A model for protecting privacy. Inter- national Journal of Uncertainty, Fuzziness and Knowledge-Based Systems 10, 05 (2002), 557–570

    2002

  30. [38]

    Jian Tang, Meng Qu, Mingzhe Wang, Ming Zhang, Jun Yan, and Qiaozhu Mei. 2015. LINE: Large-scale information network embedding. In International Conference on World Wide Web. 1067–1077

    2015

  31. [39]

    Cunchao Tu, Xiangkai Zeng, Hao Wang, Zhengyan Zhang, Zhiyuan Liu, Maosong Sun, Bo Zhang, and Leyu Lin. 2018. A unified framework for community detection and network representation learning. IEEE Transactions on Knowledge and Data Engineering 31, 6 (2018), 1051–1065

    2018

  32. [40]

    Yue Wang and Xintao Wu. 2013. Preserving differential privacy in degree- correlation based graph generation. Transactions on data privacy 6, 2 (2013), 127

    2013

  33. [41]

    Zihang Xiang, Tianhao Wang, and Di Wang. 2024. Preserving node-level privacy in graph neural networks. InIEEE Symposium on Security and Privacy. 4714–4732

    2024

  34. [42]

    Qian Xiao, Rui Chen, and Kian-Lee Tan. 2014. Differentially private network data release via structural inference. In ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 911–920

    2014

  35. [43]

    Carl Yang, Haonan Wang, Ke Zhang, Liang Chen, and Lichao Sun. 2021. Se- cure deep graph generation with link differential privacy. In International Joint Conference on Artificial Intelligence. 3271–3278

    2021

  36. [44]

    Qingqing Ye, Haibo Hu, Kai Huang, Man Ho Au, and Qiao Xue. 2023. Stateful Switch: Optimized time series release with local differential privacy. In IEEE INFOCOM Conference on Computer Communications . 1–10

    2023

  37. [45]

    Qingqing Ye, Haibo Hu, Ninghui Li, Xiaofeng Meng, Huadi Zheng, and Haotian Yan. 2021. Beyond Value Perturbation: Local differential privacy in the temporal setting. In IEEE INFOCOM Conference on Computer Communications . 1–10

    2021

  38. [46]

    Qingqing Ye, Haibo Hu, Xiaofeng Meng, and Huadi Zheng. 2019. PrivKV: Key- value data collection with local differential privacy. InIEEE Symposium on Security and Privacy. 317–331

    2019

  39. [47]

    Qingqing Ye, Haibo Hu, Xiaofeng Meng, Huadi Zheng, Kai Huang, Chengfang Fang, and Jie Shi. 2021. PrivKVM*: Revisiting key-value statistics estimation with local differential privacy. IEEE Transactions on Dependable and Secure Computing 20, 1 (2021), 17–35

    2021

  40. [48]

    Jiaxuan You, Rex Ying, Xiang Ren, William Hamilton, and Jure Leskovec. 2018. GraphRNN: Generating realistic graphs with deep auto-regressive models. In International Conference on Machine Learning . 5708–5717

    2018

  41. [49]

    Quan Yuan, Zhikun Zhang, Linkang Du, Min Chen, Peng Cheng, and Mingyang Sun. 2023. PrivGraph: Differentially private graph data publication by exploiting community information. In USENIX Security Symposium. 3241–3258

    2023

  42. [50]

    Muhan Zhang and Yixin Chen. 2018. Link prediction based on graph neural networks. In International Conference Neural Information Processing Systems . 5165–5175

    2018

  43. [51]

    Qiuchen Zhang, Hong kyu Lee, Jing Ma, Jian Lou, Carl Yang, and Li Xiong. 2024. DPAR: Decoupled graph neural networks with node-level differential privacy. In Proceedings of the ACM on Web Conference . 1170–1181. PrivDPR: Synthetic Graph Publishing with Deep PageRank under Diff...

    2024

  44. [52]

    Yuemin Zhang, Qingqing Ye, Rui Chen, Haibo Hu, and Qilong Han. 2023. Trajec- tory data collection with local differential privacy. In Proceedings of the VLDB Endowment. 2591–2604. Appendix A Graph Synthesis In what follows, we describe the specific details of generating graphs...

    2023

  45. [2011]

    InACM SIGCOMM Internet Measurement Conference

    Sharing graphs using differentially private graph models. InACM SIGCOMM Internet Measurement Conference. 81–98

  46. [2014]

    In IEEE International Conference on Data Engineering

    Private search on key-value stores with hierarchical indexes. In IEEE International Conference on Data Engineering . 628–639

  47. [2019]

    InACM SIGSAC Conference on Computer and Communications Security

    Analyzing subgraph statistics from extended local views with decentralized differential privacy. InACM SIGSAC Conference on Computer and Communications Security. 703–717

  48. [2022]

    In ACM International Conference on Web Search and Data Mining

    Understanding and improvement of adversarial training for network embedding from an optimization perspective. In ACM International Conference on Web Search and Data Mining . 230–240

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.