Adaptive AI-Driven metamaterial cloaking against Trojan-horse probing in photonic quantum communication channels
Asgar Hosseinnezhad, H. Sabri
We present an adaptive cloaking and intrusion-detection framework that integrates anisotropic metamaterial shielding with machine-learning-based anomaly detection to enhance the security of photonic quantum communication channels. The threat model considered in this work corresponds to Trojan-horse probing attacks, in which an adversary injects high-power coherent light into the device-side interfaces of a fibe-based QKD system and analyzes the back-reflected spectra. The proposed cloak is modeled using the full vector Maxwell equation with an effective-medium anisotropic permittivity tensor, and its tunability is achieved through electro-optic modulation of an ITO-based metamaterial unit cell. Electromagnetic simulations are performed using 3D FEM with a 20-nm mesh resolution, and the resulting reflectivity and phase-shift profiles are mapped to a quantum channel via a CPTP model to evaluate entanglement fidelity and coherence preservation. A convolutional neural network trained on photon-arrival histograms, spectral-leakage vectors, and correlation-deviation features achieves high anomaly-detection performance and enables real-time reconfiguration of the cloak. Numerical results show that the adaptive cloak reduces scattering efficiency by more than 80%, suppresses probing success probability to below 5%, and maintains entanglement fidelity above 0.93 across the telecommunication band. The system also demonstrates nonlinear robustness under depolarizing noise and outperforms static cloaking in both security and coherence preservation. These findings highlight the potential of combining metamaterial cloaking with AI-driven monitoring as a scalable physical-layer defense for next-generation quantum networks.
Results in Optics•Mar 1, 2026•2 citations