Link Inference Attack on Privacy-Preserving Knowledge Graphs
A topology-based link inference attack recovers hidden knowledge-graph relations that edge deletion was assumed to protect.
The paper shows that deleting sensitive relations before publishing a knowledge graph does not stop their recovery, because hidden edges leave structural traces. A topology-based link inference attack using all public graph structure reaches average precision 0.949 and ROC-AUC 0.999. In a more realistic setting that uses only some semantic information, it recovers up to 74% of hidden edges. Further analysis finds privacy risk is uneven and that particular structural patterns make some relations much easier to infer.
- Deleting sensitive relations leaves exploitable structural traces
- Full topology yields AP 0.949 and ROC-AUC 0.999
- Limited-semantics setting recovers up to 74% of hidden edges
- Inference risk varies sharply across entity structures
From the original article
1 min read at arxiv.orgKnowledge Graphs (KGs) are widely used to store and share structured information across sensitive domains such as healthcare, fi- nance, and social networks. A common privacy practice is to delete sen- sitive relations before publishing the graph, under the assumption that removing edges is sufficient to prevent their recovery.
By Emna Bouguerra · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.