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

Deep neural networks (DNN)-based machine learning (ML) algorithms haverecently emerged as the leading ML paradigm particularly for the task ofclassification due to their superior capability of learning efficiently fromlarge datasets. The discovery of a number of well-known attacks such as datasetpoisoning, adversarial examples, and network manipulation (through the additionof malicious nodes) has, however, put the spotlight squarely on the lack ofsecurity in DNN-based ML systems. In particular, malicious actors can use thesewell-known attacks to cause random/targeted misclassification, or cause achange in the prediction confidence, by only slightly but systematicallymanipulating the environmental parameters, inference data, or the dataacquisition block. Most of the prior adversarial attacks have, however, notaccounted for the pre-processing noise filters commonly integrated with theML-inference module. Our contribution in this work is to show that this is amajor omission since these noise filters can render ineffective the majority ofthe existing attacks, which rely essentially on introducing adversarial noise.Apart from this, we also extend the state of the art by proposing a novelpre-processing noise Filter-aware Adversarial ML attack called FAdeML. Todemonstrate the effectiveness of the proposed methodology, we generate anadversarial attack image by exploiting the "VGGNet" DNN trained for the "GermanTraffic Sign Recognition Benchmarks (GTSRB" dataset, which despite having novisual noise, can cause a classifier to misclassify even in the presence ofpre-processing noise filters.

FAdeML: Understanding the Impact of Pre-Processing Noise Filtering on Adversarial Machine Learning