Google COVID-19 Community Mobility Reports: Anonymization Process Description (version 1.1)

Aktay, Ahmet; Bavadekar, Shailesh; Cossoul, Gwen; Davis, John M.; Desfontaines, Damien; Fabrikant, Alex; Gabrilovich, Evgeniy; Gadepalli, Krishna; Gipson, Bryant; Guevara, Miguel Ángel; Kamath, Chaitanya; Kansal, Mansi; Lange, Ali; Mandayam, Chinmoy; Oplinger, Andrew; Pluntke, Christopher; Roessler, Thomas; Schlosberg, Arran; Shekel, Tomer; Vispute, Swapnil; Mia, Vu,; Wellenius, Gregory A.; Williams, Brian; Wilson, Royce J

doi:10.48550/arxiv.2004.04145

Cited by 49 publications

(56 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Differential privacy (DP) is the standard for privacypreserving statistical summaries (Dwork et al, 2006b). Companies such as Microsoft (Pereira et al, 2021), Google (Aktay et al, 2020), Apple (Tang et al, 2017), and government organizations such as the US Census (Abowd, 2018), have successfully applied DP in machine learning and data sharing scenarios. The popularity of DP is due to its strong mathematical guarantees.…”

Section: Introductionmentioning

confidence: 99%

An Analysis of the Deployment of Models Trained on Private Tabular Synthetic Data: Unexpected Surprises

Pereira,

Kshirsagar,

Mukherjee

et al. 2021

Preprint

View full text Add to dashboard Cite

Diferentially private (DP) synthetic datasets are a powerful approach for training machine learning models while respecting the privacy of individual data providers. The effect of DP on the fairness of the resulting trained models is not yet well understood. In this contribution, we systematically study the effects of differentially private synthetic data generation on classification. We analyze disparities in model utility and bias caused by the synthetic dataset, measured through algorithmic fairness metrics. Our first set of results show that although there seems to be a clear negative correlation between privacy and utility (the more private, the less accurate) across all data synthesizers we evaluated, more privacy does not necessarily imply more bias. Additionally, we assess the effects of utilizing synthetic datasets for model training and model evaluation. We show that results obtained on synthetic data can misestimate the actual model performance when it is deployed on real data. We hence advocate on the need for defining proper testing protocols in scenarios where differentially private synthetic datasets are utilized for model training and evaluation.

show abstract

Section: Introductionmentioning

confidence: 99%

An Analysis of the Deployment of Models Trained on Private Tabular Synthetic Data: Unexpected Surprises

Pereira,

Kshirsagar,

Mukherjee

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…In the resulting model ( 1)-( 2) that couples the commuting and non-commuting dynamics, the velocities (1), as well as the diffusion coefficients 2), with i ∈ {S, E, I, A, R}, are designed to take into account the heterogeneity of geographical areas, and are thus chosen dependent on the spatial location. Similarly, also the relaxation times τ i = τ i (x), i ∈ {S, E, I, A, R} are space dependent.…”

Section: Characterizing Commuter and Non Commuter Dynamicsmentioning

confidence: 99%

“…To model the effects of the lockdown imposed by the government from March 9, 2020 in north of Italy [61], from that day the transmission rate β A is reduced by 50%, also increasing the coefficients k I = k A = 80 as a result of the population becoming increasingly aware of the epidemic risks. In addition, the percentage of commuting individuals is reduced by 60% for each compartment according to mobility data tracked through GPS systems of mobile phones and made temporarily available by Google [1,56].…”

Section: Computational Setupmentioning

confidence: 99%

Spatial spread of COVID-19 outbreak in Italy using multiscale kinetic transport equations with uncertainty

Bertaglia,

Boscheri,

Dimarco

et al. 2021

Preprint

View full text Add to dashboard Cite

In this paper we introduce a space-dependent multiscale model to describe the spatial spread of an infectious disease under uncertain data with particular interest in simulating the onset of the COVID-19 epidemic in Italy. While virus transmission is ruled by a SEIAR type compartmental model, within our approach the population is given by a sum of commuters moving on a extra-urban scale and non commuters interacting only on the smaller urban scale. A transport dynamic of the commuter population at large spatial scales, based on kinetic equations, is coupled with a diffusion model for non commuters at the urban scale. Thanks to a suitable scaling limit, the kinetic transport model used to describe the dynamics of commuters, within a given urban area coincides with the diffusion equations that characterize the movement of non-commuting individuals. Because of the high uncertainty in the data reported in the early phase of the epidemic, the presence of random inputs in both the initial data and the epidemic parameters is included in the model. A robust numerical method is designed to deal with the presence of multiple scales and the uncertainty quantification process. In our simulations, we considered a realistic geographical domain, describing the Lombardy region, in which the size of the cities, the number of infected individuals, the average number of daily commuters moving from one city to another, and the epidemic aspects are taken into account through a calibration of the model parameters based on the actual available data. The results show that the model is able to describe correctly the main features of the spatial expansion of the first wave of COVID-19 in northern Italy.

show abstract

“…States commuter dataset (Machanavajjhala et al, 2008). More recently, in 2020, Google's COVID-19 Mobility Reports used 2.64-differential privacy for the daily reports a total of 79.22-differential privacy monthly (Aktay et al, 2020). In the same year, LinkedIn revealed their LinkedIn's Audience Engagement API that protected LinkedIn members' content engagement data, which used ( , δ)-differential privacy with daily values of (0.15, 10 −10 ) or (34.9, 7 × 10 −9 ) for monthly queries (Rogers et al, 2020).…”

Section: Setting the Privacy Loss Budgetmentioning

confidence: 99%

A Feasibility Study of Differentially Private Summary Statistics and Regression Analyses for Administrative Tax Data

Barrientos¹,

Williams²,

Snoke³

et al. 2021

Preprint

View full text Add to dashboard Cite

Federal administrative tax data are invaluable for research, but because of privacy concerns, access to these data is typically limited to select agencies and a few individuals. An alternative to sharing microlevel data are validation servers, which allow individuals to query statistics without accessing the confidential data. This paper studies the feasibility of using differentially private (DP) methods to implement such a server. We provide an extensive study on existing DP methods for releasing tabular statistics, means, quantiles, and regression estimates. We also include new methodological adaptations to existing DP regression algorithms for using new data types and returning standard error estimates. We evaluate the selected methods based on the accuracy of the output for statistical analyses, using real administrative tax data obtained from the Internal Revenue Service Statistics of Income (SOI) Division. Our findings show that a validation server would be feasible for simple statistics but would struggle to produce accurate regression estimates and confidence intervals. We outline challenges and offer recommendations for future work on validation servers. This is the first comprehensive statistical study of DP methodology on a real, complex dataset, that has significant implications for the direction of a growing research field.

show abstract

Google COVID-19 Community Mobility Reports: Anonymization Process Description (version 1.1)

Cited by 49 publications

References 0 publications

An Analysis of the Deployment of Models Trained on Private Tabular Synthetic Data: Unexpected Surprises

An Analysis of the Deployment of Models Trained on Private Tabular Synthetic Data: Unexpected Surprises

Spatial spread of COVID-19 outbreak in Italy using multiscale kinetic transport equations with uncertainty

A Feasibility Study of Differentially Private Summary Statistics and Regression Analyses for Administrative Tax Data

Contact Info

Product

Resources

About