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

Despite considerable efforts on making them robust, real-world ML-basedsystems remain vulnerable to decision based attacks, as definitive proofs oftheir operational robustness have so far proven intractable. The canonicalapproach in robustness evaluation calls for adaptive attacks, that is withcomplete knowledge of the defense and tailored to bypass it. In this study, weintroduce a more expansive notion of being adaptive and show how attacks butalso defenses can benefit by it and by learning from each other throughinteraction. We propose and evaluate a framework for adaptively optimizingblack-box attacks and defenses against each other through the competitive gamethey form. To reliably measure robustness, it is important to evaluate againstrealistic and worst-case attacks. We thus augment both attacks and the evasivearsenal at their disposal through adaptive control, and observe that the samecan be done for defenses, before we evaluate them first apart and then jointlyunder a multi-agent perspective. We demonstrate that active defenses, whichcontrol how the system responds, are a necessary complement to model hardeningwhen facing decision-based attacks; then how these defenses can be circumventedby adaptive attacks, only to finally elicit active and adaptive defenses. Wevalidate our observations through a wide theoretical and empiricalinvestigation to confirm that AI-enabled adversaries pose a considerable threatto black-box ML-based systems, rekindling the proverbial arms race wheredefenses have to be AI-enabled too. Succinctly, we address the challenges posedby adaptive adversaries and develop adaptive defenses, thereby laying outeffective strategies in ensuring the robustness of ML-based systems deployed inthe real-world.

Adversarial Markov Games: On Adaptive Decision-Based Attacks and Defenses