Turning Black Box into White Box: Dataset Distillation Leaks
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866918363950743552 |
|---|---|
| author | Chen, Huajie Zhu, Tianqing Zhong, Yuchen Zhang, Yang Wang, Shang He, Feng Zhang, Lefeng Shen, Jialiang Wang, Minghao Zhou, Wanlei |
| author_facet | Chen, Huajie Zhu, Tianqing Zhong, Yuchen Zhang, Yang Wang, Shang He, Feng Zhang, Lefeng Shen, Jialiang Wang, Minghao Zhou, Wanlei |
| contents | Dataset distillation compresses a large real dataset into a small synthetic one, enabling models trained on the synthetic data to achieve performance comparable to those trained on the real data. Although synthetic datasets are assumed to be privacy-preserving, we show that existing distillation methods can cause severe privacy leakage because synthetic datasets implicitly encode the weight trajectories of the distilled model, they become over-informative and exploitable by adversaries. To expose this risk, we introduce the Information Revelation Attack (IRA) against state-of-the-art distillation techniques. Experiments show that IRA accurately predicts both the distillation algorithm and model architecture, and can successfully infer membership and recover sensitive samples from the real dataset. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_01053 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Turning Black Box into White Box: Dataset Distillation Leaks Chen, Huajie Zhu, Tianqing Zhong, Yuchen Zhang, Yang Wang, Shang He, Feng Zhang, Lefeng Shen, Jialiang Wang, Minghao Zhou, Wanlei Cryptography and Security Artificial Intelligence Machine Learning Dataset distillation compresses a large real dataset into a small synthetic one, enabling models trained on the synthetic data to achieve performance comparable to those trained on the real data. Although synthetic datasets are assumed to be privacy-preserving, we show that existing distillation methods can cause severe privacy leakage because synthetic datasets implicitly encode the weight trajectories of the distilled model, they become over-informative and exploitable by adversaries. To expose this risk, we introduce the Information Revelation Attack (IRA) against state-of-the-art distillation techniques. Experiments show that IRA accurately predicts both the distillation algorithm and model architecture, and can successfully infer membership and recover sensitive samples from the real dataset. |
| title | Turning Black Box into White Box: Dataset Distillation Leaks |
| topic | Cryptography and Security Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2603.01053 |