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

In the last decade, deep learning algorithms have become very popular thanksto the achieved performance in many machine learning and computer vision tasks.However, most of the deep learning architectures are vulnerable to so calledadversarial examples. This questions the security of deep neural networks (DNN)for many security- and trust-sensitive domains. The majority of the proposedexisting adversarial attacks are based on the differentiability of the DNN costfunction.Defence strategies are mostly based on machine learning and signalprocessing principles that either try to detect-reject or filter out theadversarial perturbations and completely neglect the classical cryptographiccomponent in the defence. In this work, we propose a new defence mechanismbased on the second Kerckhoffs's cryptographic principle which states that thedefence and classification algorithm are supposed to be known, but not the key.To be compliant with the assumption that the attacker does not have access tothe secret key, we will primarily focus on a gray-box scenario and do notaddress a white-box one. More particularly, we assume that the attacker doesnot have direct access to the secret block, but (a) he completely knows thesystem architecture, (b) he has access to the data used for training andtesting and (c) he can observe the output of the classifier for each giveninput. We show empirically that our system is efficient against most famousstate-of-the-art attacks in black-box and gray-box scenarios.

Bridging machine learning and cryptography in defence against adversarial attacks