ZERO-TRUST INTELLIGENT BEAM MANAGEMENT FOR B5G/6G MMWAVE

Abstract

Present a secure design beam-management framework for B5G/6G mmWave that couples convolutional neural networks (CNNs) for real-time beam selection with elliptic-curve cryptography (ECC) and AES-GCM, governed by a Zero-Trust control loop. The end-to-end pipeline is implemented in ns-3 with realistic channel and mobility models (static, pedestrian, vehicular) and evaluated using beam-alignment time, throughput, packet delivery ratio (PDR), and tail-sensitive latency metrics. Relative to exhaustive search, the CNN reduces alignment time by 60–70%. Enabling ECC adds 1.0–1.4 ms, while Zero-Trust enforcement yields 2.2–2.6 ms end-to-end overhead, predominantly confined to the upper tail (p90/p95/p99, ES₉₅); the distribution’s center (median/MAD) remains essentially unchanged. In vehicular tests, throughput progresses from 680 Mb/s (exhaustive) to 850 Mb/s (CNN), then 810 Mb/s (CNN+ECC) and 795–800 Mb/s with Zero-Trust; PDR stays within 1% of the CNN+ECC baseline. These results indicate that learning-based, crypto-hardened beam management can sustain low alignment delay and continuous trust guarantees under mobility, with operationally bounded overheads in realistic ns-3 settings.