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

Adversarial attacks represent a security threat to machine learning basedautomatic speech recognition (ASR) systems. To prevent such attacks we proposean adversarial example detection strategy applicable to any ASR system thatpredicts a probability distribution over output tokens in each time step. Wemeasure a set of characteristics of this distribution: the median, maximum, andminimum over the output probabilities, the entropy, and the Jensen-Shannondivergence of the distributions of subsequent time steps. Then, we fit aGaussian distribution to the characteristics observed for benign data. Bycomputing the likelihood of incoming new audio we can distinguish maliciousinputs from samples from clean data with an area under the receiving operatorcharacteristic (AUROC) higher than 0.99, which drops to 0.98 for less-qualityaudio. To assess the robustness of our method we build adaptive attacks. Thisreduces the AUROC to 0.96 but results in more noisy adversarial clips.

Leveraging characteristics of the output probability distribution for identifying adversarial audio examples