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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Training deep neural networks (DNNs) requires large datasets and powerfulcomputing resources, which has led some owners to restrict redistributionwithout permission. Watermarking techniques that embed confidential data intoDNNs have been used to protect ownership, but these can degrade modelperformance and are vulnerable to watermark removal attacks. Recently,DeepJudge was introduced as an alternative approach to measuring the similaritybetween a suspect and a victim model. While DeepJudge shows promise inaddressing the shortcomings of watermarking, it primarily addresses situationswhere the suspect model copies the victim's architecture. In this study, weintroduce DeepTaster, a novel DNN fingerprinting technique, to addressscenarios where a victim's data is unlawfully used to build a suspect model.DeepTaster can effectively identify such DNN model theft attacks, even when thesuspect model's architecture deviates from the victim's. To accomplish this,DeepTaster generates adversarial images with perturbations, transforms theminto the Fourier frequency domain, and uses these transformed images toidentify the dataset used in a suspect model. The underlying premise is thatadversarial images can capture the unique characteristics of DNNs built with aspecific dataset. To demonstrate the effectiveness of DeepTaster, we evaluatedthe effectiveness of DeepTaster by assessing its detection accuracy on threedatasets (CIFAR10, MNIST, and Tiny-ImageNet) across three model architectures(ResNet18, VGG16, and DenseNet161). We conducted experiments under variousattack scenarios, including transfer learning, pruning, fine-tuning, and dataaugmentation. Specifically, in the Multi-Architecture Attack scenario,DeepTaster was able to identify all the stolen cases across all datasets, whileDeepJudge failed to detect any of the cases.

DeepTaster: Adversarial Perturbation-Based Fingerprinting to Identify Proprietary Dataset Use in Deep Neural Networks