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

This paper highlights vulnerabilities of deep learning-driven semanticcommunications to backdoor (Trojan) attacks. Semantic communications aims toconvey a desired meaning while transferring information from a transmitter toits receiver. An encoder-decoder pair that is represented by two deep neuralnetworks (DNNs) as part of an autoencoder is trained to reconstruct signalssuch as images at the receiver by transmitting latent features of small sizeover a limited number of channel uses. In the meantime, another DNN of asemantic task classifier at the receiver is jointly trained with theautoencoder to check the meaning conveyed to the receiver. The complex decisionspace of the DNNs makes semantic communications susceptible to adversarialmanipulations. In a backdoor (Trojan) attack, the adversary adds triggers to asmall portion of training samples and changes the label to a target label. Whenthe transfer of images is considered, the triggers can be added to the imagesor equivalently to the corresponding transmitted or received signals. In testtime, the adversary activates these triggers by providing poisoned samples asinput to the encoder (or decoder) of semantic communications. The backdoorattack can effectively change the semantic information transferred for thepoisoned input samples to a target meaning. As the performance of semanticcommunications improves with the signal-to-noise ratio and the number ofchannel uses, the success of the backdoor attack increases as well. Also,increasing the Trojan ratio in training data makes the attack more successful.In the meantime, the effect of this attack on the unpoisoned input samplesremains limited. Overall, this paper shows that the backdoor attack poses aserious threat to semantic communications and presents novel design guidelinesto preserve the meaning of transferred information in the presence of backdoorattacks.

Vulnerabilities of Deep Learning-Driven Semantic Communications to Backdoor (Trojan) Attacks