English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Graph neural networks (GNNs) have emerged as a state-of-the-art approach tomodel and draw inferences from large scale graph-structured data in variousapplication settings such as social networking. The primary goal of a GNN is tolearn an embedding for each graph node in a dataset that encodes both the nodefeatures and the local graph structure around the node. Embeddings generated bya GNN for a graph node are unique to that GNN. Prior work has shown that GNNsare prone to model extraction attacks. Model extraction attacks and defenseshave been explored extensively in other non-graph settings. While detecting orpreventing model extraction appears to be difficult, deterring them viaeffective ownership verification techniques offer a potential defense. Innon-graph settings, fingerprinting models, or the data used to build them, haveshown to be a promising approach toward ownership verification. We presentGrOVe, a state-of-the-art GNN model fingerprinting scheme that, given a targetmodel and a suspect model, can reliably determine if the suspect model wastrained independently of the target model or if it is a surrogate of the targetmodel obtained via model extraction. We show that GrOVe can distinguish betweensurrogate and independent models even when the independent model uses the sametraining dataset and architecture as the original target model. Using sixbenchmark datasets and three model architectures, we show that consistentlyachieves low false-positive and false-negative rates. We demonstrate that isrobust against known fingerprint evasion techniques while remainingcomputationally efficient.

GrOVe: Ownership Verification of Graph Neural Networks using Embeddings