diffGHOST: Diffusion based Generative Hedged Oblivious Synthetic Trajectories

Florent Guépin *, Cheick Tidiani Cisse *, Denis Renaud *, François Bidet, Arnaud Legendre · 2026

Read diffGHOST: Diffusion based Generative Hedged Oblivious Synthetic Trajectories by Florent Guépin *, Cheick Tidiani Cisse *, Denis Renaud *, François Bidet, Arnaud Legendre (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-10647.txt.

Introduction

Florent Guépin and four colleagues present diffGHOST, a tool for making fake movement paths that guard privacy. Movement paths hold highly personal facts about people, such as where they worship or what politics they hold. Rules such as the European GDPR limit how such data can be shared. Their 2026 paper offers synthetic paths as a way to share useful movement facts without tying them to real people.

Past tools claimed privacy but gave no firm guarantee, and even top models proved open to privacy attacks. diffGHOST builds on a diffusion model that splits its latent space into segments. It spots and curbs memorization of key samples through those segments. The design hedges against leaks while keeping the fake paths useful for research, city planning, and health studies.

Catalog record

CREATORFlorent Guépin *, Cheick Tidiani Cisse *, Denis Renaud *, François Bidet, Arnaud Legendre
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE49 KB
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2605.10647

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

diffGHOST: Diffusion based Generative Hedged Oblivious Synthetic Trajectories Florent Guépin *, Cheick Tidiani Cisse *, Denis Renaud *, François Bidet, Arnaud Legendre 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. 1 1 footnotetext: These authors contributed equally to this work 1 Introduction Trajectory information about individuals is a valuable source of information, for a large variety of applications, ranging from research such as epidemiology studies [ 39 ] , to urban planning [ 26 ] or environmental studies [ 25 ] . However, those trajectory information contain intrinsically private information, such as religious practices or political affiliation information [ 12 ] . Thus, regulations such as the General Data Protection Regulation (GDPR) in Europe [ 10 ] have arise, and apply to human trajectories datasets, preventing stakeholders from valuable utilization of the underlying private information. Regulations introduced a notion of adversarial effort to produce information about a set of individuals, introducing a fundamental trade-off when publication such datasets: Between the quality of the valuable information referred to as the utility and the personal data protection guarantees, referred to as the privacy . Following this rising interest, mathematical guarantees have quickly emerged to provide insurance of privacy. The formalism that quantifies the sufficient level of noise to add in order to achieve a certain tradeoff, is called Differential Privacy (DP), and has been first introduced by Dwork et al. [ 9 ] . However, in practice, this doesn’t answer the fundamental trade-off privacy-utility as the level of noise one has to use to ensure privacy is too high to achieve satisfaction, especially for trajectories. A prolific literature exist to tighten the required magnitude of the noise [ 19 ] , with relaxed DP [ 8 ] , but the micro-information, however crucial for some usages is destroyed [ 37 ] . To answer this issue, production of synthetic mobility data has emerged as a promising solution to preserve micro-information while providing access to anonymized information [ 46 ] . This is because the synthetic trajectories would be synthesized by a model rather than pertain to individuals, and therefore would not be attributable to anyone specifically. In practice however, synthetic data has also been proven vulnerable to privacy attacks [ 15 , 28 ] . Even state of the art model, such as DiffTraj, based on diffusion model and introduced by [ 46 ] , falls vulnerable to privacy attacks as it relies solely on the generation by diffusion model [ 27 ] . Contributions. In this paper, we introduce a novel architecture called diffGHOST for Diffusion based Generative Hedged Oblivious Synthetic Trajectories. This model aims to produce synthetic datasets that provides a strong non-memorization guarantee. To do so, we introduce what is, to the best of our knowledge, the first conditional diffusion model based on latent space of VAE for trajectory generation. To build a strong privacy guarantee, we used a novel idea of segmentation of the distribution conditional to the latent space of a VAE, to later produce strong non-memorization guarantees at generation time. 2 Related Works Trajectory generation. Trajectory generation methods can be broadly divided into generative and non-generative approaches. Non-generative methods, such as perturbing real trajectories with random or Gaussian noise [ 41 ] or combining them [ 31 ] , are the most straightforward. While simple and sometimes privacy-preserving by construction, these approaches tend to distort the original distribution, producing unrealistic traces with limited utility for real applications. Non-generative solutions thus often fail to address the utility-privacy trade-off, as privacy can only be improved at the cost of utility. This limitation motivates the preference for generative approaches in the literature. These methods leverage deep neural networks to learn the underlying spatio-temporal distribution of real trajectories, from which new synthetic traces can be sampled. Liu et al. [ 24 ] first introduced the idea of trajectory generation based on Generative Adversarial Networks (GANs) [ 13 ] . Since then, a wide range of architectures have been explored: GAN-based [ 34 ] , VAE-based [ 6 ] , diffusion-based [ 46 ] , and transformer-based [ 20 ] solutions. Each comes with its own trade-offs. Diffusion models produce high-quality and diverse samples but may struggle to capture long-range dependencies. Transformer-based models address this through the attention mechanism, yet may lack diversity and typically require map-grid tokenization. Notably, [ 46 ] proposed conditioning generation on explicit features such as start/end points, duration, or average speed, enabling fine-grained control. In this work, we instead condition our model on discrete latent representations, a compact conditioning that implicitly captures multiple features and groups trajectories according to their spatio-temporal structure. Privacy Guarantees in trajectory generation. Discussion over the privacy guarantees in trajectory generation often falls into two distinct categories. The first one grounds the privacy analysis into the inherent property of model generation, using argument based on the anonymity by design offered by generative solutions, as described in multiple survey [ 29 , 4 ] . Following the argument made by( [ 7 ] and [ 1 ] ), we argue these privacy properties need to be formally evaluated and quantified. The other category grounds the privacy analysis on the formal guarantees provided by the addition of differential privacy into their models [ 42 ] . First, Buchholz et al. [ 4 ] unveiled that these formal guarantees suffers from problems in their proofs, leading to a false sense of privacy. Second, these solutions come at a great cost in utility as they add noise to the global distribution. By the capacity of our model to segment the original distribution, we unlock the capacity to provide, when needed, local K-anonymity, to tailor the noise addition to each segment’s memorization’s sensitivity and save the overall utility of the model. K-anonymity in tabular datasets has been introduced first by Sweeney et al. [ 38 ] . K-anonymity ensures that for every selection of attributes, at least K persons are concerned by this selection. This notion has been carefully adapted to trajectory information by Gruteser et al. [ 14 ] . It translates that for each couple of spatiotemporal points, there is at least K individuals can be described by these spatiotemporal selection. In our solution, we locally blur the spatiotemporal information of individuals evaluated memorized by the model, to achieve formal local K-anonymity, removing the memorization threat these segments had. 3 Background 3.1 Trajectories A trajectory, as the capture of individual’s mobility, is typically defined as a chronologically ordered sequence of event. Formally, with U the set of all individual in the considered dataset, we can define a trajectory as follows. An individual u\in U has

Read the complete text