Search papers, labs, and topics across Lattice.
This paper introduces a differentially private (DP) pipeline for simulating disease spread on sensitive human contact networks. The pipeline uses node-level DP to compute network summary statistics, fits a statistical network model (SBM or ERGM) to generate synthetic networks, and then simulates disease spread on these networks using an agent-based model. Experiments on egocentric sexual network data from the ARTNet study demonstrate that the DP noise is small compared to sampling and model misspecification errors, suggesting the feasibility of privacy-preserving epidemiologic insights.
You can protect individual privacy in disease transmission models without significantly compromising the accuracy of epidemiological insights.
Epidemiologic studies of infectious diseases often rely on models of contact networks to capture the complex interactions that govern disease spread, and ongoing projects aim to vastly increase the scale at which such data can be collected. However, contact networks may include sensitive information, such as sexual relationships or drug use behavior. Protecting individual privacy while maintaining the scientific usefulness of the data is crucial. We propose a privacy-preserving pipeline for disease spread simulation studies based on a sensitive network that integrates differential privacy (DP) with statistical network models such as stochastic block models (SBMs) and exponential random graph models (ERGMs). Our pipeline comprises three steps: (1) compute network summary statistics using \emph{node-level} DP (which corresponds to protecting individuals' contributions); (2) fit a statistical model, like an ERGM, using these summaries, which allows generating synthetic networks reflecting the structure of the original network; and (3) simulate disease spread on the synthetic networks using an agent-based model. We evaluate the effectiveness of our approach using a simple Susceptible-Infected-Susceptible (SIS) disease model under multiple configurations. We compare both numerical results, such as simulated disease incidence and prevalence, as well as qualitative conclusions such as intervention effect size, on networks generated with and without differential privacy constraints. Our experiments are based on egocentric sexual network data from the ARTNet study (a survey about HIV-related behaviors). Our results show that the noise added for privacy is small relative to other sources of error (sampling and model misspecification). This suggests that, in principle, curators of such sensitive data can provide valuable epidemiologic insights while protecting privacy.
