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

Recent years have witnessed significant success in Self-Supervised Learning(SSL), which facilitates various downstream tasks. However, attackers may stealsuch SSL models and commercialize them for profit, making it crucial to protecttheir Intellectual Property (IP). Most existing IP protection solutions aredesigned for supervised learning models and cannot be used directly since theyrequire that the models' downstream tasks and target labels be known andavailable during watermark embedding, which is not always possible in thedomain of SSL. To address such a problem especially when downstream tasks arediverse and unknown during watermark embedding, we propose a novel black-boxwatermarking solution, named SSL-WM, for protecting the ownership of SSLmodels. SSL-WM maps watermarked inputs by the watermarked encoders into aninvariant representation space, which causes any downstream classifiers toproduce expected behavior, thus allowing the detection of embedded watermarks.We evaluate SSL-WM on numerous tasks, such as Computer Vision (CV) and NaturalLanguage Processing (NLP), using different SSL models, includingcontrastive-based and generative-based. Experimental results demonstrate thatSSL-WM can effectively verify the ownership of stolen SSL models in variousdownstream tasks. Furthermore, SSL-WM is robust against model fine-tuning andpruning attacks. Lastly, SSL-WM can also evade detection from evaluatedwatermark detection approaches, demonstrating its promising application inprotecting the IP of SSL models.

SSL-WM: A Black-Box Watermarking Approach for Encoders Pre-trained by Self-supervised Learning