WiP: Quantum-Assisted Secure ISAC with Encrypted Inference

Abstract

Integrated sensing and communication (ISAC) reuses a wireless waveform for data delivery and environmental perception, but this coupling exposes communication payloads, sensing semantics, and downstream inference to a common attacker. We present QAS-ISAC, a cross-layer design that combines an entanglement-assisted security tag with postquantum encrypted inference. Sparse quantum-correlated tags enable online interception and tampering detection through quantum bit error rate (QBER), while Learning-With-Errors (LWE) ciphertexts protect task-relevant features and support a shallow polynomial inference model. Simulations of an OFDM/FMCW ISAC link show that QAS-ISAC achieves 0.042 sensing NMSE and 4.3 × 10-3 BER at 10 dB, raises composite quantum-AI eavesdropper error to 0.82, and retains 85.7% certified accuracy at perturbation radius 0.05. Encrypted inference takes 5.1 ms per sample, supporting practical post-quantum secure ISAC.