Radio Signal Classification by Adversarially Robust Quantum Machine Learning

Radio signal classification plays a pivotal role in identifying the modulation scheme used in received radio signals, which is essential for demodulation and the proper interpretation of the transmitted information. Researchers have underscored the high susceptibility of machine learning (ML) algorithms for radio signal classification to adversarial attacks. Such vulnerability can result in severe consequences, including the misinterpretation of critical messages, the interception of classified information, or the disruption of communication channels. Recent advancements in quantum computing have revolutionized theories and implementations of computation, leading to the unprecedented development of quantum machine learning (QML). It has been shown that quantum variational circuits (QVCs) provide notably enhanced robustness against classical adversarial attacks in image classification. However, no research has yet explored whether QML can similarly mitigate adversarial threats in the context of radio signal classification. This work applies QVCs to radio signal classification and investigates their robustness to various adversarial attacks. We also propose the novel application of the approximate amplitude encoding (AAE) technique to efficiently encode radio signal data. Our extensive simulation results show that attacks generated on QVCs transfer well to convolutional neural network (CNN) models, indicating that these adversarial examples can fool neural networks that they are not explicitly designed to attack. However, the converse is not true: QVCs largely resist attacks generated on CNNs. Overall, through comprehensive simulations, our results shed new light on the growing field of QML by bridging knowledge gaps in QML for radio signal classification and revealing the advantages of applying QML methods in practical applications.