---
title: LDP-Difftraj
canonical_url: "https://www.modelscope.ai/models/Saihong/LDP-Difftraj"
md_url: "https://www.modelscope.ai/models/Saihong/LDP-Difftraj.md"
repository: Saihong/LDP-Difftraj
last_updated: 2026-02-05
library_name:
  - pytorch
frameworks:
  - Pytorch
downloads: 0
stars: 0
---

# LDP-Difftraj

> LDP-Difftraj - An open-source model by Saihong on ModelScope. This repository provides model weights and generated results for LDP-DiffTraj, a personalized trajectory generation framework integrating local differential privacy (LDP) with diffusion-based…

- **Repository**: Saihong/LDP-Difftraj
- **Downloads**: 0
- **Stars**: 0
- **Last updated**: 2026-02-05

Source: https://www.modelscope.ai/models/Saihong/LDP-Difftraj

---

# LDP-DiffTraj

This repository provides **model weights and generated results** for **LDP-DiffTraj**, a personalized trajectory generation framework integrating **local differential privacy (LDP)** with **diffusion-based generative modeling**.

The model is designed for **privacy-preserving human mobility synthesis**, where each trajectory is associated with a heterogeneous privacy budget and generated under corresponding privacy strength.

## Contents

- **Model Weights**: Pretrained diffusion model checkpoints trained solely on **locally perturbed trajectories** (no clean data observed).
- **Generated Results**: Synthetic trajectories under multiple privacy levels and robustness configurations, used for utility evaluation.

## Usage

### Download via ModelScope SDK
```python
from modelscope import snapshot_download
model_dir = snapshot_download('Saihong/LDP-DiffTraj')
```