Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Parameter Privacy-Preserving Data Sharing: A Particle-Belief MDP Formulation

Haokun Yu, Jingyuan Zhou, Kaidi Yang

TL;DR

This work addresses parameter privacy in data sharing for continuous-state dynamical systems by formulating a privacy-utility trade-off using the leakage $I^{\boldsymbol{\pi}}(\Theta; Y_{1:T},W_{1:T})$ and a distortion constraint. It develops a belief-MDP framework and introduces a scalable particle-belief MDP via Sequential Monte Carlo, paired with a regime-adaptive Gaussian-mixture upper bound on mutual information to enable tractable optimization. A practical RL-based solver combining an MGF encoder and a DDPG learner optimizes continuous data-sharing policies. Empirical results on a mixed-autonomy platoon show that the learned policy significantly hinders inference of human-driving parameters while preserving data usability and system performance, highlighting the approach’s real-world applicability for privacy-preserving data sharing in networked control systems.

Abstract

This paper investigates parameter-privacy-preserving data sharing in continuous-state dynamical systems, where a data owner designs a data-sharing policy to support downstream estimation and control while preventing adversarial inference of a sensitive parameter. This data-sharing problem is formulated as an optimization problem that trades off privacy leakage and the impact of data sharing on the data owner's utility, subject to a data-usability constraint. We show that this problem admits an equivalent belief Markov decision process (MDP) formulation, which provides a simplified representation of the optimal policy. To efficiently characterize information-theoretic privacy leakage in continuous state and action spaces, we propose a particle-belief MDP formulation that tracks the parameter posterior via sequential Monte Carlo, yielding a tractable belief-state approximation that converges asymptotically as the number of particles increases. We further derive a tractable closed-form upper bound on particle-based MI via Gaussian mixture approximations, which enables efficient optimization of the particle-belief MDP. Experiments on a mixed-autonomy platoon show that the learned continuous policy substantially impedes inference attacks on human-driving behavior parameters while maintaining data usability and system performance.

Parameter Privacy-Preserving Data Sharing: A Particle-Belief MDP Formulation

TL;DR

This work addresses parameter privacy in data sharing for continuous-state dynamical systems by formulating a privacy-utility trade-off using the leakage and a distortion constraint. It develops a belief-MDP framework and introduces a scalable particle-belief MDP via Sequential Monte Carlo, paired with a regime-adaptive Gaussian-mixture upper bound on mutual information to enable tractable optimization. A practical RL-based solver combining an MGF encoder and a DDPG learner optimizes continuous data-sharing policies. Empirical results on a mixed-autonomy platoon show that the learned policy significantly hinders inference of human-driving parameters while preserving data usability and system performance, highlighting the approach’s real-world applicability for privacy-preserving data sharing in networked control systems.

Abstract

This paper investigates parameter-privacy-preserving data sharing in continuous-state dynamical systems, where a data owner designs a data-sharing policy to support downstream estimation and control while preventing adversarial inference of a sensitive parameter. This data-sharing problem is formulated as an optimization problem that trades off privacy leakage and the impact of data sharing on the data owner's utility, subject to a data-usability constraint. We show that this problem admits an equivalent belief Markov decision process (MDP) formulation, which provides a simplified representation of the optimal policy. To efficiently characterize information-theoretic privacy leakage in continuous state and action spaces, we propose a particle-belief MDP formulation that tracks the parameter posterior via sequential Monte Carlo, yielding a tractable belief-state approximation that converges asymptotically as the number of particles increases. We further derive a tractable closed-form upper bound on particle-based MI via Gaussian mixture approximations, which enables efficient optimization of the particle-belief MDP. Experiments on a mixed-autonomy platoon show that the learned continuous policy substantially impedes inference attacks on human-driving behavior parameters while maintaining data usability and system performance.
Paper Structure (25 sections, 65 equations, 2 figures, 2 tables)

This paper contains 25 sections, 65 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Problem Statement
  3. Belief MDP Formulation
  4. Particle-Belief Formulation and Upper Bound of MI in Continuous Spaces
  5. Particle formulation of belief state
  6. Asymptotic convergence of Bellman equation
  7. Upper bound on particle MI under Gaussian policy
  8. Particle filter RL-based solution method
  9. Experiments
  10. Experimental setup
  11. Performance comparison
  12. Conclusion
  13. Proof for Simplification of MI in Section \ref{['sec:Problem Statement']}
  14. Proofs in Sec. \ref{['sec: A Computationally Efficient Privacy-Preserving Policy']}
  15. Proof for Theorem \ref{['th:simplified policy Theorem']}
  16. ...and 10 more sections

Figures (2)

  • Figure 1: Privacy performance. For each $\theta^\ast$, the left figure shows the evolution of posterior probability, while the right figure reports the misdetection instances over time.
  • Figure 2: Data-usability: true states versus posterior mean estimates.

Theorems & Definitions (1)

  • Remark 1: Other cost terms