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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

End-to-end training with global optimization have popularized graph neuralnetworks (GNNs) for node classification, yet inadvertently introducedvulnerabilities to adversarial edge-perturbing attacks. Adversaries can exploitthe inherent opened interfaces of GNNs' input and output, perturbing criticaledges and thus manipulating the classification results. Current defenses, dueto their persistent utilization of global-optimization-based end-to-endtraining schemes, inherently encapsulate the vulnerabilities of GNNs. This isspecifically evidenced in their inability to defend against targeted secondaryattacks. In this paper, we propose the Graph Agent Network (GAgN) to addressthe aforementioned vulnerabilities of GNNs. GAgN is a graph-structured agentnetwork in which each node is designed as an 1-hop-view agent. Through thedecentralized interactions between agents, they can learn to infer globalperceptions to perform tasks including inferring embeddings, degrees andneighbor relationships for given nodes. This empowers nodes to filteringadversarial edges while carrying out classification tasks. Furthermore, agents'limited view prevents malicious messages from propagating globally in GAgN,thereby resisting global-optimization-based secondary attacks. We prove thatsingle-hidden-layer multilayer perceptrons (MLPs) are theoretically sufficientto achieve these functionalities. Experimental results show that GAgNeffectively implements all its intended capabilities and, compared tostate-of-the-art defenses, achieves optimal classification accuracy on theperturbed datasets.

Graph Agent Network: Empowering Nodes with Decentralized Communications Capabilities for Adversarial Resilience