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

Link prediction, inferring the undiscovered or potential links of the graph,is widely applied in the real-world. By facilitating labeled links of the graphas the training data, numerous deep learning based link prediction methods havebeen studied, which have dominant prediction accuracy compared with non-deepmethods. However,the threats of maliciously crafted training graph will leave aspecific backdoor in the deep model, thus when some specific examples are fedinto the model, it will make wrong prediction, defined as backdoor attack. Itis an important aspect that has been overlooked in the current literature. Inthis paper, we prompt the concept of backdoor attack on link prediction, andpropose Link-Backdoor to reveal the training vulnerability of the existing linkprediction methods. Specifically, the Link-Backdoor combines the fake nodeswith the nodes of the target link to form a trigger. Moreover, it optimizes thetrigger by the gradient information from the target model. Consequently, thelink prediction model trained on the backdoored dataset will predict the linkwith trigger to the target state. Extensive experiments on five benchmarkdatasets and five well-performing link prediction models demonstrate that theLink-Backdoor achieves the state-of-the-art attack success rate under bothwhite-box (i.e., available of the target model parameter)and black-box (i.e.,unavailable of the target model parameter) scenarios. Additionally, we testifythe attack under defensive circumstance, and the results indicate that theLink-Backdoor still can construct successful attack on the well-performing linkprediction methods. The code and data are available athttps://github.com/Seaocn/Link-Backdoor.

Link-Backdoor: Backdoor Attack on Link Prediction via Node Injection