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/

Zendy Plus

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

Zendy Tools

Zendy Open

Backdoor attacks represent a serious threat to neural network models. Abackdoored model will misclassify the trigger-embedded inputs into anattacker-chosen target label while performing normally on other benign inputs.There are already numerous works on backdoor attacks on neural networks, butonly a few works consider graph neural networks (GNNs). As such, there is nointensive research on explaining the impact of trigger injecting position onthe performance of backdoor attacks on GNNs.  To bridge this gap, we conduct an experimental investigation on theperformance of backdoor attacks on GNNs. We apply two powerful GNNexplainability approaches to select the optimal trigger injecting position toachieve two attacker objectives -- high attack success rate and low cleanaccuracy drop. Our empirical results on benchmark datasets and state-of-the-artneural network models demonstrate the proposed method's effectiveness inselecting trigger injecting position for backdoor attacks on GNNs. Forinstance, on the node classification task, the backdoor attack with triggerinjecting position selected by GraphLIME reaches over $84 \%$ attack successrate with less than $2.5 \%$ accuracy drop

Explainability-based Backdoor Attacks Against Graph Neural Networks