diffGHOST: Diffusion based Generative Hedged Oblivious Synthetic Trajectories

2026-05-11

AI Summary

Existing trajectory generation models often rely on implicit privacy assumptions, rendering them vulnerable to leakage of individuals’ sensitive information. To address this issue, this work proposes a conditional diffusion-based trajectory generation method that, for the first time, integrates latent space segmentation with conditional diffusion models. By explicitly identifying and mitigating memorization effects of critical samples in the latent space, the approach provides direct control over privacy risks. The proposed method not only offers provable privacy guarantees but also significantly enhances the utility of synthesized trajectories, achieving a markedly superior trade-off between privacy and utility compared to current state-of-the-art techniques.

Abstract

Trajectories are nowadays valuable information for a wide range of applications. However they are also inherently sensitive, as they contain highly personal information about individuals. Facing this challenge, synthesizing mobility trajectories has emerged as a promising solution to leverage mobility information while preserving privacy. State-of-the-art models, often rely on the false assumptions of generative models implicit privacy and fails to provide privacy guarantees while preserving trajectories utility. Here, we introduce diffGHOST, a conditional diffusion model based on latent space segmentation, designed to answer this challenge. Thus, this paper propose a methodology that identify and mitigate memorization of critical samples using condition segments of a learn latent space.

Problem

• trajectory synthesis
• privacy preservation
• generative models
• mobility data
• memorization

Innovation

• diffusion model
• trajectory synthesis
• privacy preservation
• latent space segmentation
• memorization mitigation