Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Private Agent-Based Modeling

Ayush Chopra, Arnau Quera-Bofarull, Nurullah Giray-Kuru, Michael Wooldridge, Ramesh Raskar

TL;DR

This paper addresses privacy challenges in agent-based modeling (ABM) where microdata about agents and interactions are sensitive. It introduces a privacy-preserving ABM framework based on secure multi-party computation (MPC) to perform simulation, calibration, and analysis without centralizing agents' attributes or interactions. The approach uses GMW-style additive secret sharing to enable secure aggregation and gradient computations, integrating differentiable ABMs to support gradient-based calibration and ML pipelines. A case study on an Oxford, UK epidemiological ABM demonstrates that policy-relevant insights can be obtained while preserving individual privacy, illustrating end-to-end privacy protection for ABMs.

Abstract

The practical utility of agent-based models in decision-making relies on their capacity to accurately replicate populations while seamlessly integrating real-world data streams. Yet, the incorporation of such data poses significant challenges due to privacy concerns. To address this issue, we introduce a paradigm for private agent-based modeling wherein the simulation, calibration, and analysis of agent-based models can be achieved without centralizing the agents attributes or interactions. The key insight is to leverage techniques from secure multi-party computation to design protocols for decentralized computation in agent-based models. This ensures the confidentiality of the simulated agents without compromising on simulation accuracy. We showcase our protocols on a case study with an epidemiological simulation comprising over 150,000 agents. We believe this is a critical step towards deploying agent-based models to real-world applications.

Private Agent-Based Modeling

TL;DR

This paper addresses privacy challenges in agent-based modeling (ABM) where microdata about agents and interactions are sensitive. It introduces a privacy-preserving ABM framework based on secure multi-party computation (MPC) to perform simulation, calibration, and analysis without centralizing agents' attributes or interactions. The approach uses GMW-style additive secret sharing to enable secure aggregation and gradient computations, integrating differentiable ABMs to support gradient-based calibration and ML pipelines. A case study on an Oxford, UK epidemiological ABM demonstrates that policy-relevant insights can be obtained while preserving individual privacy, illustrating end-to-end privacy protection for ABMs.

Abstract

The practical utility of agent-based models in decision-making relies on their capacity to accurately replicate populations while seamlessly integrating real-world data streams. Yet, the incorporation of such data poses significant challenges due to privacy concerns. To address this issue, we introduce a paradigm for private agent-based modeling wherein the simulation, calibration, and analysis of agent-based models can be achieved without centralizing the agents attributes or interactions. The key insight is to leverage techniques from secure multi-party computation to design protocols for decentralized computation in agent-based models. This ensures the confidentiality of the simulated agents without compromising on simulation accuracy. We showcase our protocols on a case study with an epidemiological simulation comprising over 150,000 agents. We believe this is a critical step towards deploying agent-based models to real-world applications.
Paper Structure (22 sections, 27 equations, 5 figures, 1 table, 6 algorithms)

This paper contains 22 sections, 27 equations, 5 figures, 1 table, 6 algorithms.

Table of Contents

  1. Introduction
  2. Agent-based models
  3. Simulation of ABMs
  4. Calibration of ABMs
  5. Analysis of ABMs
  6. Characterizing Privacy
  7. Threat Model
  8. Secure Multi-party Computation
  9. The GMW protocol
  10. Additive secret sharing example
  11. Private Agent-based Models
  12. Secure Simulation
  13. Secure Calibration
  14. Secure Analysis
  15. Case Study: Privacy-preserving Epidemiology
  16. ...and 7 more sections

Figures (5)

  • Figure 1: Diagram illustrating the SecureSimulation protocol for ABM parameters $\boldsymbol{\theta}$.
  • Figure 2: Infection curves for different levels of compliance: 0% (blue), 25% (green), 50% (orange), 75% (red). The number of infections has been normalized to the number of agents $N$. These plots are generated without releasing the infection status or compliance decision of any individual agent.
  • Figure 3: Left: Probability density plot for the trained normalizing flow (blue) against the prior distribution (orange). Ground-truth value is marked as a dashed black line. Right: Results from simulating $\beta$ samples from the trained flow (blue) and prior (orange) compared to the ground-truth data (black). The number of infections has been normalized to the number of agents $N$.
  • Figure 4: Age (left) and ethnicity (right) histogram of the infected population. These statistics are computed without leaking the infection or demographic properties of any agent.
  • Figure 5: Geographical distribution of infections by ZIP code sector within the city of Oxford. These statistics are computed without leaking the infection status or geo-location of any individual agent.