A Dataset with Adversarial Attacks on Deep Learning in Wireless Modulation Classification
This dataset contains adversarial attacks on Deep Learning (DL) when it is employed for the classification of wireless modulated communication signals. The attack is executed with an obfuscating waveform that is embedded in the transmitted signal in such a way that prevents the extraction of clean data for training from a wireless eavesdropper. At the same time it allows a legitimate receiver (LRx) to demodulate the data. The scheme works for both single carrier and multi-carrier orthogonal frequency division multiplexing (OFDM) waveforms and can be implemented as part of frame-based wireless protocols.
The related paper that we ask to be cited if you use this dataset is by D. Varkatzas and A. Argyriou that appears in IEEE MILCOM 2023: Limitations of Deep Learning for Modulation Classification of Obfuscated Wireless Signals.
Please see in the dataset itself. There are detailed matlab files that can be used for reading and data processing.