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

Adversarial attacks can mislead automatic speech recognition (ASR) systemsinto predicting an arbitrary target text, thus posing a clear security threat.To prevent such attacks, we propose DistriBlock, an efficient detectionstrategy applicable to any ASR system that predicts a probability distributionover output tokens in each time step. We measure a set of characteristics ofthis distribution: the median, maximum, and minimum over the outputprobabilities, the entropy of the distribution, as well as the Kullback-Leiblerand the Jensen-Shannon divergence with respect to the distributions of thesubsequent time step. Then, by leveraging the characteristics observed for bothbenign and adversarial data, we apply binary classifiers, including simplethreshold-based classification, ensembles of such classifiers, and neuralnetworks. Through extensive analysis across different state-of-the-art ASRsystems and language data sets, we demonstrate the supreme performance of thisapproach, with a mean area under the receiver operating characteristic curvefor distinguishing target adversarial examples against clean and noisy data of99% and 97%, respectively. To assess the robustness of our method, we show thatadaptive adversarial examples that can circumvent DistriBlock are much noisier,which makes them easier to detect through filtering and creates another avenuefor preserving the system's robustness.

DistriBlock: Identifying adversarial audio samples by leveraging characteristics of the output distribution