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

Textual backdoor attacks pose a practical threat to existing systems, as theycan compromise the model by inserting imperceptible triggers into inputs andmanipulating labels in the training dataset. With cutting-edge generativemodels such as GPT-4 pushing rewriting to extraordinary levels, such attacksare becoming even harder to detect. We conduct a comprehensive investigation ofthe role of black-box generative models as a backdoor attack tool, highlightingthe importance of researching relative defense strategies. In this paper, wereveal that the proposed generative model-based attack, BGMAttack, couldeffectively deceive textual classifiers. Compared with the traditional attackmethods, BGMAttack makes the backdoor trigger less conspicuous by leveragingstate-of-the-art generative models. Our extensive evaluation of attackeffectiveness across five datasets, complemented by three distinct humancognition assessments, reveals that Figure 4 achieves comparable attackperformance while maintaining superior stealthiness relative to baselinemethods.

ChatGPT as an Attack Tool: Stealthy Textual Backdoor Attack via Blackbox Generative Model Trigger