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

The collection and availability of big data, combined with advances inpre-trained models (e.g., BERT, XLNET, etc), have revolutionized the predictiveperformance of modern natural language processing tasks, ranging from textclassification to text generation. This allows corporations to provide machinelearning as a service (MLaaS) by encapsulating fine-tuned BERT-based models asAPIs. However, BERT-based APIs have exhibited a series of security and privacyvulnerabilities. For example, prior work has exploited the security issues ofthe BERT-based APIs through the adversarial examples crafted by the extractedmodel. However, the privacy leakage problems of the BERT-based APIs through theextracted model have not been well studied. On the other hand, due to the highcapacity of BERT-based APIs, the fine-tuned model is easy to be overlearned,but what kind of information can be leaked from the extracted model remainsunknown. In this work, we bridge this gap by first presenting an effectivemodel extraction attack, where the adversary can practically steal a BERT-basedAPI (the target/victim model) by only querying a limited number of queries. Wefurther develop an effective attribute inference attack which can infer thesensitive attribute of the training data used by the BERT-based APIs. Ourextensive experiments on benchmark datasets under various realistic settingsvalidate the potential vulnerabilities of BERT-based APIs. Moreover, wedemonstrate that two promising defense methods become ineffective against ourattacks, which calls for more effective defense methods.

Killing One Bird with Two Stones: Model Extraction and Attribute Inference Attacks against BERT-based APIs