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

Classifiers in supervised learning have various security and privacy issues,e.g., 1) data poisoning attacks, backdoor attacks, and adversarial examples onthe security side as well as 2) inference attacks and the right to be forgottenfor the training data on the privacy side. Various secure andprivacy-preserving supervised learning algorithms with formal guarantees havebeen proposed to address these issues. However, they suffer from variouslimitations such as accuracy loss, small certified security guarantees, and/orinefficiency. Self-supervised learning is an emerging technique to pre-trainencoders using unlabeled data. Given a pre-trained encoder as a featureextractor, supervised learning can train a simple yet accurate classifier usinga small amount of labeled training data. In this work, we perform the firstsystematic, principled measurement study to understand whether and when apre-trained encoder can address the limitations of secure or privacy-preservingsupervised learning algorithms. Our key findings are that a pre-trained encodersubstantially improves 1) both accuracy under no attacks and certified securityguarantees against data poisoning and backdoor attacks of state-of-the-artsecure learning algorithms (i.e., bagging and KNN), 2) certified securityguarantees of randomized smoothing against adversarial examples withoutsacrificing its accuracy under no attacks, 3) accuracy of differentiallyprivate classifiers, and 4) accuracy and/or efficiency of exact machineunlearning.

Pre-trained Encoders in Self-Supervised Learning Improve Secure and Privacy-preserving Supervised Learning