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

As online systems based on machine learning are offered to public or paidsubscribers via application programming interfaces (APIs), they becomevulnerable to frequent exploits and attacks. This paper studies adversarialmachine learning in the practical case when there are rate limitations on APIcalls. The adversary launches an exploratory (inference) attack by querying theAPI of an online machine learning system (in particular, a classifier) withinput data samples, collecting returned labels to build up the training data,and training an adversarial classifier that is functionally equivalent andstatistically close to the target classifier. The exploratory attack withlimited training data is shown to fail to reliably infer the target classifierof a real text classifier API that is available online to the public. Inreturn, a generative adversarial network (GAN) based on deep learning is builtto generate synthetic training data from a limited number of real training datasamples, thereby extending the training data and improving the performance ofthe inferred classifier. The exploratory attack provides the basis to launchthe causative attack (that aims to poison the training process) and evasionattack (that aims to fool the classifier into making wrong decisions) byselecting training and test data samples, respectively, based on the confidencescores obtained from the inferred classifier. These stealth attacks with smallfootprint (using a small number of API calls) make adversarial machine learningpractical under the realistic case with limited training data available to theadversary.

Generative Adversarial Networks for Black-Box API Attacks with Limited Training Data