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

Backdoor attacks are an important type of adversarial threat against deepneural network classifiers, wherein test samples from one or more sourceclasses will be (mis)classified to the attacker's target class when a backdoorpattern is embedded. In this paper, we focus on the post-training backdoordefense scenario commonly considered in the literature, where the defender aimsto detect whether a trained classifier was backdoor-attacked without any accessto the training set. Many post-training detectors are designed to detectattacks that use either one or a few specific backdoor embedding functions(e.g., patch-replacement or additive attacks). These detectors may fail whenthe backdoor embedding function used by the attacker (unknown to the defender)is different from the backdoor embedding function assumed by the defender. Incontrast, we propose a post-training defense that detects backdoor attacks witharbitrary types of backdoor embeddings, without making any assumptions aboutthe backdoor embedding type. Our detector leverages the influence of thebackdoor attack, independent of the backdoor embedding mechanism, on thelandscape of the classifier's outputs prior to the softmax layer. For eachclass, a maximum margin statistic is estimated. Detection inference is thenperformed by applying an unsupervised anomaly detector to these statistics.Thus, our detector does not need any legitimate clean samples, and canefficiently detect backdoor attacks with arbitrary numbers of source classes.These advantages over several state-of-the-art methods are demonstrated on fourdatasets, for three different types of backdoor patterns, and for a variety ofattack configurations. Finally, we propose a novel, general approach forbackdoor mitigation once a detection is made. The mitigation approach was therunner-up at the first IEEE Trojan Removal Competition. The code is onlineavailable.

MM-BD: Post-Training Detection of Backdoor Attacks with Arbitrary Backdoor Pattern Types Using a Maximum Margin Statistic