Pith. sign in

REVIEW 3 major objections 6 minor 47 references

Scalable Distributed Reproduction Numbers of Network Epidemics with Differential Privacy

T0 review · 3 major / 6 minor · reviewed 2026-08-09 · deepseek-v4-flash

Pith's one-line read The paper defines a local effective reproduction number for each node whose threshold at 1 exactly tells whether that node's infection level is increasing, decreasing, or flat, and shows the same threshold aggregates to clusters and…

desk verdict Useful cluster-level aggregation and a coherent DP pipeline, but the privacy guarantee rests on an asserted sensitivity k and the abstract overpromises equilibrium results the body never proves. read the letter →

arxiv 2501.18862 v2 pith:56HMG5EM submitted 2025-01-31 eess.SY cs.SY

classification eess.SYcs.SY MSC 92D3068P27
keywords distributedreproductionnumbersnetworkepidemicmodelseffectivenumberclusteraggregationdifferentialprivacyboundedGaussianmechanismshufflemodelmobilitydata
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

Network-level reproduction numbers summarize the whole epidemic but hide where it is spreading. This paper proposes a per-node local effective reproduction number, $\bar R^t_i = \sum_j s_i(t)\beta_{ij}x_j(t)/(\gamma_i x_i(t))$, and proves that comparing it to 1 gives an exact, real-time sign detector for that node's infection: above 1 means rising, below 1 means falling, equal means flat. It then defines cluster-level versions that aggregate from local numbers at any scale, so regional and national monitors can use the same threshold logic. Because these numbers encode how often people interact, the paper adds a privacy layer: each local authority perturbs its aggregated vector with a bounded Gaussian mechanism, a shuffler anonymizes the vectors inside each cluster, and the resulting cluster numbers carry quantified first- and second-moment accuracy. A COVID-19 mobility example suggests the privatized cluster numbers remain useful, with roughly 5 to 9 percent average error at strong privacy.

What carries the argument

The carrying object is the local distributed effective reproduction number vector, whose ith row has entries $\bar R^t_{ij}=s_i(t)\beta_{ij}x_j(t)/(\gamma_i x_i(t))$; its row sum is the local ERN $\bar R^t_i$. This vector is a diagonal similarity transform of the next-generation-matrix-based pseudo-ERN matrix, so its spectral radius matches the network-level reproduction number while its row sums encode local thresholds. For privacy, the machinery is the bounded Gaussian mechanism applied to the pre-aggregated local vectors $\zeta_i$, with the shuffler model amplifying the local $\epsilon_0$-differential privacy to central $(\epsilon,\delta)$-differential privacy, and Theorem 4 gives the accuracy of the resulting private cluster distributed ERNs.

What would settle it

Recompute the local aggregated ERN vector $\zeta_i$ after adding and after removing one individual's daily visits in the Section V mobility data, using the same transmission-rate estimator; if the largest $\ell^2$ change exceeds $k=10^{-5}$, the differential-privacy guarantee claimed for interaction frequencies does not hold on that data.

Watch

Extended reading notes

Core claim

The paper's central claim is that the sign of $\dot x_i(t)$ is fully determined by the local effective reproduction number $\bar R^t_i$, the row sum of the local distributed ERN matrix whose entries are $\bar R^t_{ij}=s_i(t)\beta_{ij}x_j(t)/(\gamma_i x_i(t))$. Theorem 1 states $\bar R^t_i>1$ iff $x_i$ is increasing, $\bar R^t_i<1$ iff it is decreasing, and $\bar R^t_i=1$ iff it is flat, provided all infection proportions are positive. The same construction is lifted to clusters: the cluster effective reproduction number $\bar R^t_{\chi_q}=\sum_{i\in\chi_q}\gamma_i x_i\bar R^t_i/\sum_{i\in\chi_q}\gamma_i x_i$ has the corresponding threshold for the sum of infected proportions in the cluster, and finer cluster values aggregate into coarser ones by a weighted sum. On the privacy side, the paper claims that a bounded Gaussian randomizer applied to each entity's local aggregated ERN vector, followed by a shuffler at each cluster, yields $(\epsilon,\delta)$-differential privacy for the cluster distributed ERN matrix, with explicit formulas for the mean and variance of the private values.

Load-bearing premise

The privacy guarantee rests on the asserted bound that a single person's mobility record changes any local aggregated reproduction-number vector by at most $k=10^{-5}$ in $\ell^2$ distance, a value the paper states but does not derive from the transmission-rate estimation procedure.

Editorial extensions

If this is right

  • A local authority can tell whether its own region is expanding, contracting, or flat by comparing its local effective reproduction number to 1, using only its own transmission row and public infection proportions.
  • Cluster effective reproduction numbers can be aggregated hierarchically from local numbers, so county-, state-, and national-level monitors can follow the same outbreak at different resolutions without pooling raw mobility data.
  • If every local ERN is below 1, the network-level effective reproduction number is below 1, so local thresholds alone certify that the overall epidemic is declining.
  • The private cluster distributed ERNs preserve threshold-scale information: in the paper's COVID-19 mobility experiment, estimates at $\epsilon=1$ track the nonprivate values with errors of roughly 5 to 9 percent.
  • The bounded Gaussian mechanism leaves zero entries at zero, so privacy noise does not fabricate transmission channels that do not exist in the network.

Reading between the lines

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

  • The threshold identity suggests a decentralized control rule: each region could push its own $\bar R^t_i$ below 1 by reducing contact rates or shortening infection windows, without needing a central optimizer; the paper lists control as future work, so this is an extrapolation.
  • The privacy guarantee would be auditable if a calibration study computed the actual maximum $\ell^2$ change in $\zeta_i$ caused by adding or removing one person's mobility record through the paper's transmission-rate estimator, rather than asserting the $k=10^{-5}$ bound.
  • A streaming deployment that publishes private cluster ERNs on every day would need a composition analysis, because the paper analyzes a single release; repeated releases would consume the privacy budget over time.
  • The cluster threshold is defined with respect to a chosen partition and a weighted sum of infected proportions, so CERN values from different partitions should not be interpreted as the same epidemic quantity.
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

3 major / 6 minor

Summary. The paper defines local distributed effective reproduction numbers (LERNs) for each node of a network SIS/SIR model, proves that each LERN exceeds one exactly when the node's infected proportion is increasing, and extends this to cluster-level effective reproduction numbers (CERNs) that threshold the summed infection rate in a cluster. It then develops a differential privacy framework in which local authorities add bounded Gaussian noise to local aggregated effective reproduction number vectors, a shuffler amplifies privacy, and central aggregators form privatized cluster-level reproduction numbers. The framework is validated on SafeGraph mobility data for 1,023 US regions grouped into 100 clusters.

Significance. If its claims hold, the paper contributes a scalable, interpretable monitoring tool: each region or cluster can share a scalar threshold signal without releasing raw mobility flows, and the local-to-cluster aggregation is exact. The privacy framework is a sensible application of existing mechanisms (bounded Gaussian noise plus shuffling), and the empirical evaluation on real mobility data is a strength. However, the central threshold property is largely a rearrangement of the model equations: since \bar R_i^t = (\dot x_i + \gamma_i x_i)/(\gamma_i x_i), the statement that \bar R_i^t > 1 iff \dot x_i > 0 is immediate from the definition. The paper also overclaims in the abstract by promising conditions for existence, uniqueness, and stability of equilibria, which do not appear in the body. Most importantly, the differential privacy guarantee depends on an asserted, not derived, sensitivity bound for real mobility data, and the main accuracy theorem contains an apparent algebraic error.

major comments (3)
  1. [Section V-C and Definition 13 / Remark 11] The differential privacy guarantee is calibrated to the adjacency parameter k = 10^{-5}, asserted in Section V-C as "the maximum variation in the distributed ERNs that a single mobile data point can cause when it changes by its maximum possible amount." No derivation, formula, or code is provided for this bound. Because Definition 3 defines adjacency by \|\zeta - \zeta'\|_2 \le k, the sensitivity of the identity mapping is k by construction (Remark 11), so the load-bearing question is whether any two local aggregated ERN vectors arising from real neighboring mobility databases are within l2 distance k. If a single mobile record can change \zeta_i by more than k, the noise variance in (15) is under-calibrated and the claimed (\epsilon_0,\delta)-differential privacy, and hence the shuffle amplification in Lemma 2, does not hold for the actual data. Please provide a derivation from the transmission-rate formula in [16] or an explicit conservative upper bound.
  2. [Section IV-D, Theorem 4] The stated first and second moments of the private cluster distributed effective reproduction numbers are inconsistent with their definition in (14). According to (14), \tilde R^t_{\chi_q,\chi_r} = (\sum_{k\in\chi_q} \tilde R^t_{k,\chi_r})/(\sum_{k\in\chi_q} \gamma_k x_k(t)), so its expectation should be a single weighted average over i \in \chi_q with denominator \sum_{i\in\chi_q} \gamma_i x_i(t). The formula given in Theorem 4 instead sums over all clusters q=1,\dots,m in the numerator and divides by \sum_{i\in(\cup\chi_q)} \gamma_i x_i(t), which appears to be an algebraic error. This is load-bearing because Theorem 4 is the formal accuracy guarantee for Problem 6.
  3. [Abstract and Section I] The abstract states that the derived conditions are "used to derive new conditions for the existence, uniqueness, and stability of equilibrium states of the underlying epidemic model," but the manuscript contains no such theorem. Theorem 2 and Corollary 1 relate the local effective reproduction numbers to the spectral radius of the next-generation matrix, which is a threshold condition for the global reproduction number, not an existence, uniqueness, or stability result for equilibria. The introduction's contribution list repeats this claim. The abstract and contributions should be revised to match the actual content, or the missing analysis should be added.
minor comments (6)
  1. [Definition 12, equation (9)] The last row of the cluster distributed effective reproduction number matrix shows \bar R^t_{\chi_m,\chi_1} twice; the final entry should be \bar R^t_{\chi_m,\chi_m}.
  2. [Equation (14)] The numerator is written as \sum_{k\in\chi_q} \tilde R^k_{,\chi_r}, which appears to be a typo for \sum_{k\in\chi_q} \tilde R^t_{k,\chi_r}.
  3. [Proof of Theorem 3] The sentence "Consider |\chi_q| = m \le n clusters" uses m for both the size of the cluster and the total number of clusters; please use |\chi_q| for the cardinality to avoid confusion.
  4. [Theorem 4 statement] The notation \tilde R_{\chi_q,\chi_r} in the theorem drops the time superscript t used elsewhere; please keep notation consistent.
  5. [Section V-B] The phrase "capturing causal relationships" is too strong for the described empirical analysis, which is correlational; please rephrase.
  6. [Section V-C, Figure 10] The RMSE values are based on 100 samples with no confidence intervals; as the percentage errors are small, reporting standard errors or confidence intervals would strengthen the accuracy claim.

Circularity Check

3 steps flagged · score 6.0 of 10

The central threshold theorems are algebraic restatements of the epidemic dynamics, and the privacy sensitivity is a definitional equality rather than a derived data-dependent bound.

  1. self definitional [Section III-B, Definition 9 (Eq. (6)) and Theorem 1]
    "¯Rt i = Pn j=1 ¯Rt ij = Pn j=1 Rt ijIij(t) ... ¯Rt i > 1 gives ... Pn j=1 si(t)βij xj(t) xi(t)γi > 1, which leads to dxi(t) dt = Pn j=1 si(t)βijxj(t) − γixi(t) > 0."

    By Eq. (6), \bar R_i^t is defined as \sum_j s_i(t)\beta_{ij}x_j(t)/(\gamma_i x_i(t)), while the SIS/SIR equation (1) gives \dot x_i(t)=\sum_j s_i(t)\beta_{ij}x_j(t)-\gamma_i x_i(t). Therefore \bar R_i^t-1=\dot x_i(t)/(\gamma_i x_i(t)), so all three threshold statements in Theorem 1 are algebraic rearrangements of the state equation. The local ERN was introduced precisely from the two terms of \dot x_i, so the theorem restates the definition rather than deriving an independent condition.

  2. self definitional [Section III-C, Definition 11 and Theorem 3]
    "If x(t) ≫ 0, then the cluster effective reproduction number (CERN) of χq is given by ¯Rt χq = P i∈χq γixi(t) ¯Rt iP i∈χq γixi(t) ... ¯Rt χq > 1 if and only if P i∈χq ˙xi(t) > 0."

    Substituting \bar R_i^t=(\sum_j s_i\beta_{ij}x_j)/(\gamma_i x_i) into Definition 11 gives \bar R^t_{\chi_q}=1+(\sum_{i\in\chi_q}\dot x_i)/(\sum_{i\in\chi_q}\gamma_i x_i). Thus the Theorem 3 threshold conditions on \sum_{i\in\chi_q}\dot x_i are identities following from the definition of the CERN as a weighted average of LERNs. The proof is algebraic manipulation of the defining ratio, not an independent dynamical derivation.

1 more flagged steps
  1. self definitional [Section IV-B, Definition 13 and Remark 11; Section V-C]
    "By using Definitions 3 and 13, we can see that the sensitivity of the identity mapping acting on a local aggregated ERN vector is ∆2ζi = k, where k is the user-specified adjacency parameter."

    Definition 3 fixes adjacency by ||ζ−ζ'||2 ≤ k, and Definition 13 defines the L2-sensitivity as the maximum of exactly that same norm over adjacent pairs. Hence ∆2ζi = k is true by construction for any data and carries no information about the epidemic inputs. The load-bearing real-data assertion in Section V-C that k=10−5 bounds the change caused by 'a single mobile data point' is stated without derivation, so the privacy mechanism is calibrated to a definitional equality plus an unverified constant.

full rationale

The two threshold theorems at the heart of the reproduction-number contribution are identities. Because the local ERN is defined as the ratio of the inflow and outflow terms of the SIS/SIR equation, Theorem 1 is a restatement of \dot x_i in a new variable; the cluster CERN is a weighted average of those local ratios, making Theorem 3 the same kind of identity. The privacy sensitivity is also a definitional tautology: the adjacency relation is defined by an ℓ2 bound, so the identity mapping's sensitivity over that relation equals that bound by construction. The real question of whether a single mobile record moves a local aggregated ERN vector by no more than k is asserted, not derived. These self-definitional steps are genuine and affect the paper's central 'threshold' claims. However, not all content is circular: Theorem 2's bridge between local thresholds and the spectral radius of the pseudo-ERN matrix uses a nontrivial row-stochastic similarity argument, and the bounded Gaussian and shuffle privacy mechanisms are stated in the paper with their conditions, so the formal DP analysis does not hide its argument entirely in a self-citation. The SafeGraph and CSSE experiments provide external, if non-circular, validation. The score of 6 reflects that several 'predictions' reduce by construction, with partial independent content remaining in the spectral and privacy machinery.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

The central mathematical claims rest on standard SIS/SIR dynamics and Perron-Frobenius theory, plus the public-data assumption. The privacy claims rest entirely on the authors' own bounded Gaussian mechanism [42] and the shuffle amplification bound [43], neither re-derived here. The data application introduces two hand-set numbers: the adjacency parameter k and the [0,14] cap. No new physical entities, forces, or conserved quantities are introduced; the local authority, shuffler, and aggregator roles are procedural components of the communication framework.

free parameters (2)
  • Adjacency parameter k = 1e-5
    Set by the authors as the maximum change in distributed ERNs caused by a single mobile data point; it controls the noise scale and the formal DP guarantee, but no derivation is provided.
  • Truncation upper bound U=14 = 14
    All local distributed ERNs are projected into [0,14] before analysis; this sets the bounded Gaussian domain and introduces bias in the private estimates.
assumptions (6)
  • domain assumption The spreading graph G=(V,E,B) is strongly connected (Assumption 1).
    Used to make the relevant matrices irreducible so that the Perron-Frobenius monotonicity argument in Theorem 2 applies.
  • domain assumption The infected proportions satisfy x(t) >> 0 for all t considered.
    Required for infection ratios x_j/x_i and for the row-stochastic similarity argument; in the data section it is enforced by assuming at least one person is infected in every region.
  • domain assumption The susceptible and infected state vectors and recovery rates are public, and each local authority knows only its own row of the transmission matrix (Assumption 2).
    Defines the privacy threat model: only transmission rates are sensitive, which justifies privatizing the local aggregated ERN vectors.
  • standard math The network-level reproduction number is rho(Gamma^{-1}B), with threshold at 1 for the SIS/SIR models.
    Imported from [7], [8], [31] and used as the external benchmark in Theorem 2 and Remark 8.
  • standard math The bounded Gaussian mechanism from [42] satisfies eps0-differential privacy under the sigma condition in (15).
    The paper builds Mechanism 1 on this published result, whose authors overlap with the present paper, and does not reprove or machine-check it.
  • standard math The shuffle amplification bound from [43] applies to the cluster shuffler (Lemma 2).
    Used to convert local eps0-DP into amplified (eps,delta)-DP; the proof is omitted and cited to [43].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Scalable Distributed Reproduction Numbers of Network Epidemics with Differential Privacy." pith.science (2026). https://pith.science/paper/56HMG5EM

@misc{pith2026250118862,
  author       = {Pith},
  title        = {Pith review of: Scalable Distributed Reproduction Numbers of Network Epidemics with Differential Privacy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/56HMG5EM}},
  note         = {Machine review of arXiv:2501.18862}
}
read the original abstract

Reproduction numbers are widely used for the estimation and prediction of epidemic spreading processes over networks. However, conventional reproduction numbers of an overall network do not indicate where an epidemic is spreading. Therefore, we propose a novel notion of local distributed reproduction numbers to capture the spreading behaviors of each node in a network. We first show how to compute them and then use them to derive new conditions under which an outbreak can occur. These conditions are then used to derive new conditions for the existence, uniqueness, and stability of equilibrium states of the underlying epidemic model. Building upon these local distributed reproduction numbers, we define cluster distributed reproduction numbers to model the spread between clusters composed of nodes. Furthermore, we demonstrate that the local distributed reproduction numbers can be aggregated into cluster distributed reproduction numbers at different scales. However, both local and cluster distributed reproduction numbers can reveal the frequency of interactions between nodes in a network, which raises privacy concerns. Thus, we next develop a privacy framework that implements a differential privacy mechanism to provably protect the frequency of interactions between nodes when computing distributed reproduction numbers. Numerical experiments show that, even under differential privacy, the distributed reproduction numbers provide accurate estimates of the epidemic spread while also providing more insights than conventional reproduction numbers.

Figures

Figures reproduced from arXiv: 2501.18862 by the authors.

Figure 1
Figure 1. (Bottom) shows the corresponding network-level effective reproduction number, i.e., Rt . The effective repro￾duction number Rt > 1 until roughly timestep 25. However, the infected proportions of most communities, including com￾munities 1, 2, 3, 4, and 9 have already significantly decreased by timestep 25. Therefore, if we aim to analyze a single entity or a subnetwork of connected entities, then the network-level re… view at source ↗
Figure 2
Figure 2. Structure of the network framework: The local author￾ities of the same cluster are responsible for collecting and reporting their local distributed reproduction numbers to the shuffler and then to the aggregator of the same cluster. The shuffler is responsible for anonymizing the local distributed reproduction numbers and randomly shuffling them to elim￾inate any usefulness in their order. The aggregator decodes and… view at source ↗
Figure 3
Figure 3. Local distributed ERNs. The network on the left depicts a spreading network with three nodes i, j, k, where the network is modeled by the transmission rates within and between the entities, and the recovery rate within entities, and the network on the right depicts how to model the system using local distributed ERNs. and s(t)β γ as the basic and effective reproduction numbers, respectively. Based on these terms, we… view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Cluster distributed effective reproduction numbers. The network on the left depicts a spreading network with six nodes f, g, h, i, j, and k. The network can be partitioned via a finer partition πX such that Xπ = {χp, χq, χr} and its coarser partition πXˆ such that Xˆ π…
Figure 5
Figure 5. Figure 5: Construction of the private cluster distributed ERN matrix of a disease spreading network with seven entities. The seven entities are managed by their own local authorities, labeled from 1 to 7. In addition, these seven authorities are organized into three clusters: χ1…
Figure 6
Figure 6. Figure 6: The centroids of the jurisdictions of local authorities [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]
Figure 8
Figure 8. Figure 8: Cluster distributed ERNs between region χ1 (Detroit area) and χ2 (Miami area) over time [PITH_FULL_IMAGE:figures/full_fig_p018_8.png]
Figure 9
Figure 9. Figure 9: Cluster distributed ERNs between regions [PITH_FULL_IMAGE:figures/full_fig_p018_9.png]
Figure 7
Figure 7. Figure 7: The infected portions of region χ1 (Detroit area), χ2 (Miami area), χ3 (Delta Junction) over time [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 10
Figure 10. Figure 10: The root mean squared error (RMSE) of the private [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]
Figure 11
Figure 11. Figure 11: Original values of the cluster distributed ERNs and their private values with [PITH_FULL_IMAGE:figures/full_fig_p020_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 47 canonical work pages

  1. [16]

    A high-resolution flux-matrix model describes the spread of diseases in a spatial network and the effect of mitigation strategies,

    G. Le Treut, G. Huber, M. Kamb, K. Kawagoe, A. McGeever, J. Miller, R. Pnini, B. Veytsman, and D. Yllanes, “A high-resolution flux-matrix model describes the spread of diseases in a spatial network and the effect of mitigation strategies,” Sci. Rep., vol. 12, no. 1, p. 15946, 2022

  2. [1]

    Reproduction numbers of infectious disease models,

    P. van den Driessche, “Reproduction numbers of infectious disease models,” Infec. Dise. Model. , vol. 2, no. 3, pp. 288–303, 2017

  3. [2]

    The effect of interventions on COVID-19,

    K. Soltesz, F. Gustafsson, T. Timpka, J. Jald ´en, C. Jidling, A. Heimerson, T. B. Sch ¨on, A. Spreco, J. Ekberg, ¨O. Dahlstr ¨om et al., “The effect of interventions on COVID-19,” Nature, vol. 588, no. 7839, pp. E26–E28, 2020. 20 (a) The original value of the cluster distributed ERNs from the region χ1 (Detroit region) to the region χ2 (Miami region) is ...

  4. [3]

    A framework for counterfactual analysis, strategy evaluation, and control of epidemics using reproduction number estimates,

    B. She, R. L. Smith, I. Pytlarz, S. Sundaram, and P. E. Par ´e, “A framework for counterfactual analysis, strategy evaluation, and control of epidemics using reproduction number estimates,” PLoS Comp. Bio. , vol. 20, no. 11, p. e1012569, 2024

  5. [4]

    On a networked SIS epidemic model with cooperative and antagonistic opinion dynamics,

    B. She, J. Liu, S. Sundaram, and P. E. Par ´e, “On a networked SIS epidemic model with cooperative and antagonistic opinion dynamics,” IEEE Trans, on Contr. of Netw. Syst. , vol. 9, pp. 1154 – 1165, 2022

  6. [5]

    Nonsmooth convex optimization to estimate the Covid-19 reproduction number space-time evolution with robustness against low quality data,

    B. Pascal, P. Abry, N. Pustelnik, S. Roux, R. Gribonval, and P. Flandrin, “Nonsmooth convex optimization to estimate the Covid-19 reproduction number space-time evolution with robustness against low quality data,” IEEE Trans. on Sig. Proc. , vol. 70, pp. 2859–2868, 2022

  7. [6]

    Convex optimization of the basic repro- duction number,

    K. D. Smith and F. Bullo, “Convex optimization of the basic repro- duction number,” IEEE Trans. on Autom. Contr. , vol. 68, no. 7, pp. 4398–4404, 2022

  8. [7]

    On the dynamics of deterministic epidemic propagation over networks,

    W. Mei, S. Mohagheghi, S. Zampieri, and F. Bullo, “On the dynamics of deterministic epidemic propagation over networks,” Annu. Rev. in Control, vol. 44, pp. 116–128, 2017

Show all 47 references
  1. [8]

    Modeling, estimation, and analysis of epidemics over networks: An overview,

    P. E. Par ´e, C. L. Beck, and T. Bas ¸ar, “Modeling, estimation, and analysis of epidemics over networks: An overview,” Annu. Rev. in Control , vol. 50, pp. 345–360, 2020

  2. [9]

    Analysis, prediction, and control of epidemics: A survey from scalar to dynamic network models,

    L. Zino and M. Cao, “Analysis, prediction, and control of epidemics: A survey from scalar to dynamic network models,” IEEE Cir. and Syst. Mag., vol. 21, no. 4, pp. 4–23, 2021

  3. [10]

    Analysis and control of epidemics: A survey of spreading processes on complex networks,

    C. Nowzari, V . M. Preciado, and G. J. Pappas, “Analysis and control of epidemics: A survey of spreading processes on complex networks,” IEEE Control Syst. Magazine , vol. 36, no. 1, pp. 26–46, 2016

  4. [11]

    The N-intertwined SIS epidemic network model,

    P. Van Mieghem, “The N-intertwined SIS epidemic network model,” Computing, vol. 93, no. 2, pp. 147–169, 2011

  5. [12]

    Analysis and control of a continuous-time bi-virus model,

    J. Liu, P. E. Par ´e, A. Nedi ´c, C. Tang, C. Beck, and T. Bas ¸ar, “Analysis and control of a continuous-time bi-virus model,” IEEE Trans. Autom. Control, vol. 64, no. 12, pp. 4891–4906, 2019

  6. [13]

    Modeling COVID-19 scenarios for the United States,

    I. C.-. F. Team, “Modeling COVID-19 scenarios for the United States,” Nature Medicine, vol. 27, no. 1, pp. 94–105, 2021

  7. [14]

    A multi-city epidemic model,

    J. Arino and P. Van den Driessche, “A multi-city epidemic model,” Mathematical Population Studies, vol. 10, no. 3, pp. 175–193, 2003

  8. [15]

    Assessing the role of contact tracing in a suspected H7N2 influenza a outbreak in humans in wales,

    K. T. Eames, C. Webb, K. Thomas, J. Smith, R. Salmon, and J. M. F. Temple, “Assessing the role of contact tracing in a suspected H7N2 influenza a outbreak in humans in wales,” BMC infectious diseases , vol. 10, pp. 1–6, 2010

  9. [17]

    Multiscale mobility networks and the spatial spread- ing of infectious diseases,

    D. Balcan, V . Colizza, B. Gonc ¸alves, H. Hu, J. J. Ramasco, and A. Vespignani, “Multiscale mobility networks and the spatial spread- ing of infectious diseases,” Proceedings of the National Academy of Sciences, vol. 106, no. 51, pp. 21 484–21 489, 2009

  10. [18]

    Locally differentially private analysis of graph statistics,

    J. Imola, T. Murakami, and K. Chaudhuri, “Locally differentially private analysis of graph statistics,” in Proc. 30th USENIX security symposium (USENIX Security 21) , 2021, pp. 983–1000

  11. [19]

    Private analysis of graph structure,

    V . Karwa, S. Raskhodnikova, A. Smith, and G. Yaroslavtsev, “Private analysis of graph structure,” ACM Trans. Database Syst., vol. 39, no. 3, oct 2014

  12. [20]

    Publishing graph degree distribution with node differential privacy,

    W.-Y . Day, N. Li, and M. Lyu, “Publishing graph degree distribution with node differential privacy,” in Proc. 2016 Int. Conf. on Mana of Data , ser. SIGMOD ’16. New York, NY , USA: Association for Computing Machinery, 2016, p. 123–138

  13. [21]

    Differentially private graph publishing with degree distribution preservation,

    S. Zhang, W. Ni, and N. Fu, “Differentially private graph publishing with degree distribution preservation,” Computers & Security , vol. 106, p. 102285, 2021

  14. [22]

    Edge differential privacy for algebraic connectivity of graphs,

    B. Chen, C. Hawkins, K. Yazdani, and M. Hale, “Edge differential privacy for algebraic connectivity of graphs,” in Proc. of the 60th IEEE Conf. on Dec. and Cont. (CDC) . IEEE, 2021, pp. 2764–2769

  15. [23]

    The algorithmic foundations of differential privacy,

    C. Dwork, A. Roth et al. , “The algorithmic foundations of differential privacy,” Foundations and Trends® in Theoretical Computer Science , vol. 9, no. 3–4, pp. 211–407, 2014. 21

  16. [24]

    Differential privacy in control and network systems,

    J. Cort ´es, G. E. Dullerud, S. Han, J. Le Ny, S. Mitra, and G. J. Pappas, “Differential privacy in control and network systems,” in Proc. of the 2016 IEEE Conf. on Deci. and Contr. (CDC) . IEEE, 2016, pp. 4252– 4272

  17. [25]

    Differentially private smart metering: Implementation, analytics, and billing,

    M. Hale, P. Barooah, K. Parker, and K. Yazdani, “Differentially private smart metering: Implementation, analytics, and billing,” in Proceedings of the 1st ACM International Workshop on Urban Building Energy Sensing, Controls, Big Data Analysis, and Visualization, ser. UrbSys’1...

  18. [26]

    Differentially private formation control,

    C. Hawkins and M. Hale, “Differentially private formation control,” in in Proc. of the 59th IEEE Conf. on Deci. and Cont. (CDC) , 2020, pp. 6260–6265

  19. [27]

    Node and edge differential privacy for graph laplacian spectra: Mechanisms and scaling laws,

    C. Hawkins, B. Chen, K. Yazdani, and M. Hale, “Node and edge differential privacy for graph laplacian spectra: Mechanisms and scaling laws,” IEEE Transactions on Network Science and Engineering , 2023

  20. [28]

    Distributed reproduction numbers of networked epidemics,

    B. She, P. E. Par ´e, and M. Hale, “Distributed reproduction numbers of networked epidemics,” in Proc. of the 2023 Amer. Contr. Conf. (ACC) . IEEE, 2023, pp. 4302–4307

  21. [29]

    Differentially private computation of basic reproduction numbers in networked epidemic models,

    B. Chen, B. She, C. Hawkins, A. Benvenuti, B. Fallin, P. E. Par ´e, and M. Hale, “Differentially private computation of basic reproduction numbers in networked epidemic models,” in Proc. of the 2024 Amer. Contr. Conf. (ACC). IEEE, 2024, pp. 4422–4427

  22. [30]

    Modeling, estimation, and analysis of epidemics over networks: An overview,

    P. E. Par ´e, C. L. Beck, and T. Bas ¸ar, “Modeling, estimation, and analysis of epidemics over networks: An overview,” Annual Reviews in Control , vol. 50, pp. 345–360, 2020

  23. [31]

    The construction of next-generation matrices for compartmental epidemic models,

    O. Diekmann, J. Heesterbeek, and M. G. Roberts, “The construction of next-generation matrices for compartmental epidemic models,” J. of the Roy. Soci. Inter., vol. 7, no. 47, pp. 873–885, 2010

  24. [32]

    Peak infection time for a networked SIR epidemic with opinion dynamics,

    B. She, H. C. Leung, S. Sundaram, and P. E. Par ´e, “Peak infection time for a networked SIR epidemic with opinion dynamics,” in In Proc. 60th IEEE Conf. on Dec. and Contr. IEEE, 2021, pp. 2104–2109

  25. [33]

    Shortest paths and distances with differential privacy,

    A. Sealfon, “Shortest paths and distances with differential privacy,” in Proc. of the 35th ACM SIGMOD-SIGACT-SIGAI Symp. on Prin. of Data. Syst., 2016, pp. 29–41

  26. [34]

    Accurate estimation of the degree distribution of private networks,

    M. Hay, C. Li, G. Miklau, and D. Jensen, “Accurate estimation of the degree distribution of private networks,” in Proc. 2009 Ninth IEEE Int. Conf. on Data Mining , 2009, pp. 169–178

  27. [35]

    Differentially private data analysis of social networks via restricted sensitivity,

    J. Blocki, A. Blum, A. Datta, and O. Sheffet, “Differentially private data analysis of social networks via restricted sensitivity,” arXiv preprint arXiv:1208.4586, 2013

  28. [36]

    High- resolution measurements of face-to-face contact patterns in a primary school,

    J. Stehl ´e, N. V oirin, A. Barrat, C. Cattuto, L. Isella, J.-F. Pinton, M. Quaggiotto, W. Van den Broeck, C. R ´egis, B. Lina et al. , “High- resolution measurements of face-to-face contact patterns in a primary school,” PloS One, vol. 6, no. 8, p. e23176, 2011

  29. [37]

    What can we learn privately?

    S. P. Kasiviswanathan, H. K. Lee, K. Nissim, S. Raskhodnikova, and A. Smith, “What can we learn privately?” in Proc. of the 49th Annual IEEE Symp. on Found. of Comp. Sci. , 2008, pp. 531–540

  30. [38]

    Differential privacy: An economic method for choosing epsilon,

    J. Hsu, M. Gaboardi, A. Haeberlen, S. Khanna, A. Narayan, B. C. Pierce, and A. Roth, “Differential privacy: An economic method for choosing epsilon,” in Proc. of the 2014 IEEE 27th Comp. Sec. Found. Symp. IEEE, 2014, pp. 398–410

  31. [39]

    The truncated normal distribution,

    J. Burkardt, “The truncated normal distribution,” Department of Scien- tific Computing Website, Florida State University , vol. 1, p. 35, 2014

  32. [40]

    On the dynamics of deterministic epidemic propagation over networks,

    W. Mei, S. Mohagheghi, S. Zampieri, and F. Bullo, “On the dynamics of deterministic epidemic propagation over networks,” Ann. Rev. in Cont. , vol. 44, pp. 116–128, 2017

  33. [41]

    R. S. Varga, Matrix Iterative Analysis . Springer Science & Business Media, 2009, vol. 27

  34. [42]

    The bounded Gaussian mechanism for differen- tial privacy,

    B. Chen and M. Hale, “The bounded Gaussian mechanism for differen- tial privacy,” Journal of Privacy and Confidentiality , vol. 14(1), 2024

  35. [43]

    Distributed differential privacy via shuffling,

    A. Cheu, A. Smith, J. Ullman, D. Zeber, and M. Zhilyaev, “Distributed differential privacy via shuffling,” in Advances in Cryptology – EURO- CRYPT 2019, Y . Ishai and V . Rijmen, Eds. Cham: Springer International Publishing, 2019, pp. 375–403

  36. [44]

    SafeGraph Social Distancing Metrics,

    Safe Graph, “SafeGraph Social Distancing Metrics,” 2021, available at https://docs.safegraph.com/docs/social-distancing-metrics

  37. [45]

    An interactive web-based dashboard to track COVID-19 in real time,

    E. Dong, H. Du, and L. Gardner, “An interactive web-based dashboard to track COVID-19 in real time,” The Lancet Infectious Diseases , vol. 20, no. 5, pp. 533–534, 2020

  38. [46]

    SafeGraph Open Census Data ,

    Safe Graph, “SafeGraph Open Census Data ,” 2020, available at https: //docs.safegraph.com/docs/open-census-data

  39. [47]

    Effective Reproductive Num- ber estimation for initial stage of COVID-19 pandemic in Latin Amer- ican Countries,

    Y . Caicedo-Ochoa, D. E. Rebell ´on-S´anchez, M. Pe ˜naloza-Rall´on, H. F. Cort´es-Motta, and Y . R. M´endez-Fandi˜no, “Effective Reproductive Num- ber estimation for initial stage of COVID-19 pandemic in Latin Amer- ican Countries,” International Journal of Infectious Disease...

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.