EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Federated Illusion: Multi-Level Geometric Privacy Audit for Federated Graph Unlearning

Haoke Han, Yan Huang, Zhenzhen Xie, Junjie Pang

Information · 27.04.2026 · 17 (5) · 424

DOI: 10.3390/info17050424 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Information, online erschienen am 27.04.2026. Verfasst von Haoke Han und 3 weiteren Autor:innen. Die Arbeit nennt 27 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Machine unlearning in federated graph learning must satisfy the multi-level indistinguishability requirement of the deletion of a target node being undetectable at the level of the global model, of the unlearning client’s local model, and of every non-target client’s local model. Approximate unlearning methods that pass confidence-based audits may still leave geometric traces through embedding drift at one or more of these K+1 levels. We formalize this requirement, introduce a five-model threat taxonomy, and extend the Hub–Ripple embedding drift audit to global, local, and cross-client levels. Across 31,900 trials spanning five graph benchmarks, five federated unlearning methods, and four supplementary ablations (K-value, cross-edge handling, control sampling, and DP-SGD defense), we find that all approximate methods fail the following multi-level requirement: the Confidence–Embedding Gap persists at 0.12 (versus 0.35 centralized), cross-client leakage correlates with shared cross-edge count (r=0.56, p

Zitieren

APA

Han, H., Huang, Y., Xie, Z., & Pang, J. (2026). Federated Illusion: Multi-Level Geometric Privacy Audit for Federated Graph Unlearning. Information, 17(5), 424. https://doi.org/10.3390/info17050424

Vancouver

Han H, Huang Y, Xie Z, Pang J. Federated Illusion: Multi-Level Geometric Privacy Audit for Federated Graph Unlearning. Information. 2026;17(5):424. doi:10.3390/info17050424

BibTeX
@article{Han2026,
  author = {Han, Haoke and Huang, Yan and Xie, Zhenzhen and Pang, Junjie},
  title = {Federated Illusion: Multi-Level Geometric Privacy Audit for Federated Graph Unlearning},
  journal = {Information},
  year = {2026},
  volume = {17},
  number = {5},
  pages = {424},
  doi = {10.3390/info17050424}
}
Kostenlos starten →