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

We introduce a new framework of adversarial attacks, named calibrationattacks, in which the attacks are generated and organized to trap victim modelsto be miscalibrated without altering their original accuracy, hence seriouslyendangering the trustworthiness of the models and any decision-making based ontheir confidence scores. Specifically, we identify four novel forms ofcalibration attacks: underconfidence attacks, overconfidence attacks, maximummiscalibration attacks, and random confidence attacks, in both the black-boxand white-box setups. We then test these new attacks on typical victim modelswith comprehensive datasets, demonstrating that even with a relatively lownumber of queries, the attacks can create significant calibration mistakes. Wefurther provide detailed analyses to understand different aspects ofcalibration attacks. Building on that, we investigate the effectiveness ofwidely used adversarial defences and calibration methods against these types ofattacks, which then inspires us to devise two novel defences against suchcalibration attacks.

Calibration Attack: A Framework For Adversarial Attacks Targeting Calibration