diffGHOST: Diffusion Model for Privacy-Preserving Synthetic Trajectories
A new conditional diffusion model named diffGHOST has been developed by researchers, focusing on latent space segmentation to create synthetic mobility trajectories that ensure privacy. This model tackles the shortcomings of leading generative models that lack adequate privacy protections while still delivering useful trajectory data. By utilizing condition segments within a learnable latent space, diffGHOST effectively recognizes and reduces the memorization of essential samples. This research has been made available on arXiv in the Computer Science > Artificial Intelligence category.
Key facts
- diffGHOST is a conditional diffusion model for synthetic trajectory generation.
- It uses latent space segmentation to preserve privacy.
- The model mitigates memorization of critical samples.
- State-of-the-art models often lack privacy guarantees.
- The paper is on arXiv under Computer Science > Artificial Intelligence.
- Trajectories contain highly personal information.
- Synthesizing mobility trajectories is a promising privacy solution.
- diffGHOST aims to balance privacy and utility.
Entities
Institutions
- arXiv