aiqbench analysis
This research investigates the adversarial robustness of complex-valued neural networks (CVNNs), revealing their specific vulnerability to perturbations targeting phase information rather than magnitude. The study demonstrates that while CVNNs can outperform real-valued counterparts in certain scenarios, they remain highly susceptible to specialized "Phase Attacks" that degrade performance more severely than standard attacks. These findings are critical for practitioners deploying CVNNs in safety-critical applications where input signal integrity must be guaranteed against worst-case perturbations.
Perturbing the Phase: Analyzing Adversarial Robustness of Complex-Valued Neural Networks
arXiv:2602.06577v2 Announce Type: replace-cross Abstract: Complex-valued neural networks (CVNNs) are rising in popularity for all kinds of applications. To safely use CVNNs in practice, analyzing their robustness against outliers is crucial. One well known technique to understand the behavior of deep neural networks is to investigate their behavior under adversarial attacks, which can be seen as worst case minimal perturbations. We design Phase Attacks, a kind of attack specifically targeting the phase information of complex-valued inputs. Additionally, we derive complex-valued versions of commonly used adversarial attacks. We show that in some scenarios CVNNs are more robust than RVNNs and that both are very susceptible to phase changes with the Phase Attacks decreasing the model performance more, than equally strong regular attacks, which can attack both phase and magnitude.
Abstract:Complex-valued neural networks (CVNNs) are rising in popularity for all kinds of applications. To safely use CVNNs in practice, analyzing their robustness against outliers is crucial. One well known technique to understand the behavior of deep neural networks is to investigate their behavior under adversarial attacks, which can be seen as worst case minimal perturbations. We design Phase Attacks, a kind of attack specifically targeting the phase information of complex-valued inputs. Additionally, we derive complex-valued versions of commonly used adversarial attacks. We show that in some scenarios CVNNs are more robust than RVNNs and that both are very susceptible to phase changes with the Phase Attacks decreasing the model performance more, than equally strong regular attacks, which can attack both phase and magnitude.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2602.06577 [cs.LG]
(or arXiv:2602.06577v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2602.06577 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 218 words
Related coverage
Frontier Risk Report (February to March 2026)
METR · May 19, 2026
Frontier AI Risk Report (February–March 2026)
METR · May 19, 2026
Informe de riesgos de la IA de frontera (febrero–marzo de 2026)
METR · May 19, 2026
A Mike's-Eye View of ARC's Research
AI Alignment Forum · Jun 9, 2026