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 (DNNs) have been found to be vulnerable to backdoorattacks, raising security concerns about their deployment in mission-criticalapplications. While existing defense methods have demonstrated promisingresults, it is still not clear how to effectively remove backdoor-associatedneurons in backdoored DNNs. In this paper, we propose a novel defense called\emph{Reconstructive Neuron Pruning} (RNP) to expose and prune backdoor neuronsvia an unlearning and then recovering process. Specifically, RNP first unlearnsthe neurons by maximizing the model's error on a small subset of clean samplesand then recovers the neurons by minimizing the model's error on the same data.In RNP, unlearning is operated at the neuron level while recovering is operatedat the filter level, forming an asymmetric reconstructive learning procedure.We show that such an asymmetric process on only a few clean samples caneffectively expose and prune the backdoor neurons implanted by a wide range ofattacks, achieving a new state-of-the-art defense performance. Moreover, theunlearned model at the intermediate step of our RNP can be directly used toimprove other backdoor defense tasks including backdoor removal, triggerrecovery, backdoor label detection, and backdoor sample detection. Code isavailable at \url{https://github.com/bboylyg/RNP}.

Reconstructive Neuron Pruning for Backdoor Defense