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

All existing backdoor attacks to deep learning (DL) models belong to thevertical class backdoor (VCB). That is, any sample from a class will activatethe implanted backdoor in the presence of the secret trigger, regardless ofsource-class-agnostic or source-class-specific backdoor. Current trends ofexisting defenses are overwhelmingly devised for VCB attacks especially thesource-class-agnostic backdoor, which essentially neglects other potentialsimple but general backdoor types, thus giving false security implications. Itis thus urgent to discover unknown backdoor types.  This work reveals a new, simple, and general horizontal class backdoor (HCB)attack. We show that the backdoor can be naturally bounded with innocuousnatural features that are common and pervasive in the real world. Note that aninnocuous feature (e.g., expression) is irrelevant to the main task of themodel (e.g., recognizing a person from one to another). The innocuous featurespans across classes horizontally but is exhibited by partial samples per class-- satisfying the horizontal class (HC) property. Only when the trigger isconcurrently presented with the HC innocuous feature, can the backdoor beeffectively activated. Extensive experiments on attacking performance in termsof high attack success rates with tasks of 1) MNIST, 2) facial recognition, 3)traffic sign recognition, and 4) object detection demonstrate that the HCB ishighly efficient and effective. We extensively evaluate the HCB evasivenessagainst a (chronologically) series of 9 influential countermeasures ofFine-Pruning (RAID 18'), STRIP (ACSAC 19'), Neural Cleanse (Oakland 19'), ABS(CCS 19'), Februus (ACSAC 20'), MNTD (Oakland 21'), SCAn (USENIX SEC 21'), MOTH(Oakland 22'), and Beatrix (NDSS 23'), where none of them can succeed even whena simplest trigger is used.

Horizontal Class Backdoor to Deep Learning