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

Graph neural network (GNN) with a powerful representation capability has beenwidely applied to various areas, such as biological gene prediction, socialrecommendation, etc. Recent works have exposed that GNN is vulnerable to thebackdoor attack, i.e., models trained with maliciously crafted training samplesare easily fooled by patched samples. Most of the proposed studies launch thebackdoor attack using a trigger that either is the randomly generated subgraph(e.g., erd\H{o}s-r\'enyi backdoor) for less computational burden, or thegradient-based generative subgraph (e.g., graph trojaning attack) to enable amore effective attack. However, the interpretation of how is the triggerstructure and the effect of the backdoor attack related has been overlooked inthe current literature. Motifs, recurrent and statistically significantsub-graphs in graphs, contain rich structure information. In this paper, we arerethinking the trigger from the perspective of motifs, and propose amotif-based backdoor attack, denoted as Motif-Backdoor. It contributes fromthree aspects. (i) Interpretation: it provides an in-depth explanation forbackdoor effectiveness by the validity of the trigger structure from motifs,leading to some novel insights, e.g., using subgraphs that appear lessfrequently in the graph as the trigger can achieve better attack performance.(ii) Effectiveness: Motif-Backdoor reaches the state-of-the-art (SOTA) attackperformance in both black-box and defensive scenarios. (iii) Efficiency: basedon the graph motif distribution, Motif-Backdoor can quickly obtain an effectivetrigger structure without target model feedback or subgraph model generation.Extensive experimental results show that Motif-Backdoor realizes the SOTAperformance on three popular models and four public datasets compared with fivebaselines.

Motif-Backdoor: Rethinking the Backdoor Attack on Graph Neural Networks via Motifs