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

As deep neural networks continue to be used in critical domains, concernsover their security have emerged. Deep learning models are vulnerable tobackdoor attacks due to the lack of transparency. A poisoned backdoor model mayperform normally in routine environments, but exhibit malicious behavior whenthe input contains a trigger. Current research on backdoor attacks focuses onimproving the stealthiness of triggers, and most approaches require strongattacker capabilities, such as knowledge of the model structure or control overthe training process. These attacks are impractical since in most cases theattacker's capabilities are limited. Additionally, the issue of modelrobustness has not received adequate attention. For instance, modeldistillation is commonly used to streamline model size as the number ofparameters grows exponentially, and most of previous backdoor attacks failedafter model distillation; the image augmentation operations can destroy thetrigger and thus disable the backdoor. This study explores the implementationof black-box backdoor attacks within capability constraints. An attacker cancarry out such attacks by acting as either an image annotator or an imageprovider, without involvement in the training process or knowledge of thetarget model's structure. Through the design of a backdoor trigger, our attackremains effective after model distillation and image augmentation, making itmore threatening and practical. Our experimental results demonstrate that ourmethod achieves a high attack success rate in black-box scenarios and evadesstate-of-the-art backdoor defenses.

Launching a Robust Backdoor Attack under Capability Constrained Scenarios