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

Together with impressive advances touching every aspect of our society, AItechnology based on Deep Neural Networks (DNN) is bringing increasing securityconcerns. While attacks operating at test time have monopolised the initialattention of researchers, backdoor attacks, exploiting the possibility ofcorrupting DNN models by interfering with the training process, represents afurther serious threat undermining the dependability of AI techniques. In abackdoor attack, the attacker corrupts the training data so to induce anerroneous behaviour at test time. Test time errors, however, are activated onlyin the presence of a triggering event corresponding to a properly crafted inputsample. In this way, the corrupted network continues to work as expected forregular inputs, and the malicious behaviour occurs only when the attackerdecides to activate the backdoor hidden within the network. In the last fewyears, backdoor attacks have been the subject of an intense research activityfocusing on both the development of new classes of attacks, and the proposal ofpossible countermeasures. The goal of this overview paper is to review theworks published until now, classifying the different types of attacks anddefences proposed so far. The classification guiding the analysis is based onthe amount of control that the attacker has on the training process, and thecapability of the defender to verify the integrity of the data used fortraining, and to monitor the operations of the DNN at training and test time.As such, the proposed analysis is particularly suited to highlight thestrengths and weaknesses of both attacks and defences with reference to theapplication scenarios they are operating in.

An Overview of Backdoor Attacks Against Deep Neural Networks and Possible Defences