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

Deep neural networks (DNNs) have proven to be powerful predictors and arewidely used for various tasks. Credible uncertainty estimation of theirpredictions, however, is crucial for their deployment in many risk-sensitiveapplications. In this paper we present a novel and simple attack, which unlikeadversarial attacks, does not cause incorrect predictions but instead cripplesthe network's capacity for uncertainty estimation. The result is that after theattack, the DNN is more confident of its incorrect predictions than about itscorrect ones without having its accuracy reduced. We present two versions ofthe attack. The first scenario focuses on a black-box regime (where theattacker has no knowledge of the target network) and the second scenarioattacks a white-box setting. The proposed attack is only required to be ofminuscule magnitude for its perturbations to cause severe uncertaintyestimation damage, with larger magnitudes resulting in completely unusableuncertainty estimations. We demonstrate successful attacks on three of the mostpopular uncertainty estimation methods: the vanilla softmax score, DeepEnsembles and MC-Dropout. Additionally, we show an attack on SelectiveNet, theselective classification architecture. We test the proposed attack on severalcontemporary architectures such as MobileNetV2 and EfficientNetB0, all trainedto classify ImageNet.

Disrupting Deep Uncertainty Estimation Without Harming Accuracy