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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

To achieve reliable, robust, and safe AI systems, it is vital to implementfallback strategies when AI predictions cannot be trusted. Certifiers forneural networks are a reliable way to check the robustness of thesepredictions. They guarantee for some predictions that a certain class ofmanipulations or attacks could not have changed the outcome. For the remainingpredictions without guarantees, the method abstains from making a prediction,and a fallback strategy needs to be invoked, which typically incurs additionalcosts, can require a human operator, or even fail to provide any prediction.While this is a key concept towards safe and secure AI, we show for the firsttime that this approach comes with its own security risks, as such fallbackstrategies can be deliberately triggered by an adversary. In addition tonaturally occurring abstains for some inputs and perturbations, the adversarycan use training-time attacks to deliberately trigger the fallback with highprobability. This transfers the main system load onto the fallback, reducingthe overall system's integrity and/or availability. We design two novelavailability attacks, which show the practical relevance of these threats. Forexample, adding 1% poisoned data during training is sufficient to trigger thefallback and hence make the model unavailable for up to 100% of all inputs byinserting the trigger. Our extensive experiments across multiple datasets,model architectures, and certifiers demonstrate the broad applicability ofthese attacks. An initial investigation into potential defenses shows thatcurrent approaches are insufficient to mitigate the issue, highlighting theneed for new, specific solutions.

Certifiers Make Neural Networks Vulnerable to Availability Attacks