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

We present a Trojan (backdoor or trapdoor) attack that targets deep learningapplications in wireless communications. A deep learning classifier isconsidered to classify wireless signals using raw (I/Q) samples as features andmodulation types as labels. An adversary slightly manipulates training data byinserting Trojans (i.e., triggers) to only few training data samples bymodifying their phases and changing the labels of these samples to a targetlabel. This poisoned training data is used to train the deep learningclassifier. In test (inference) time, an adversary transmits signals with thesame phase shift that was added as a trigger during training. While thereceiver can accurately classify clean (unpoisoned) signals without triggers,it cannot reliably classify signals poisoned with triggers. This stealth attackremains hidden until activated by poisoned inputs (Trojans) to bypass a signalclassifier (e.g., for authentication). We show that this attack is successfulover different channel conditions and cannot be mitigated by simplypreprocessing the training and test data with random phase variations. Todetect this attack, activation based outlier detection is considered withstatistical as well as clustering techniques. We show that the latter one candetect Trojan attacks even if few samples are poisoned.

Trojan Attacks on Wireless Signal Classification with Adversarial Machine Learning