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

Speaker identification models are vulnerable to carefully designedadversarial perturbations of their input signals that induce misclassification.In this work, we propose a white-box steganography-inspired adversarial attackthat generates imperceptible adversarial perturbations against a speakeridentification model. Our approach, FoolHD, uses a Gated ConvolutionalAutoencoder that operates in the DCT domain and is trained with amulti-objective loss function, in order to generate and conceal the adversarialperturbation within the original audio files. In addition to hindering speakeridentification performance, this multi-objective loss accounts for humanperception through a frame-wise cosine similarity between MFCC feature vectorsextracted from the original and adversarial audio files. We validate theeffectiveness of FoolHD with a 250-speaker identification x-vector network,trained using VoxCeleb, in terms of accuracy, success rate, andimperceptibility. Our results show that FoolHD generates highly imperceptibleadversarial audio files (average PESQ scores above 4.30), while achieving asuccess rate of 99.6% and 99.2% in misleading the speaker identification model,for untargeted and targeted settings, respectively.

FoolHD: Fooling speaker identification by Highly imperceptible adversarial Disturbances