A Covert Physical-Layer Communication Channel in UAV Systems

This thesis investigates a novel communication paradigm for unmanned aerial vehicles (UAVs) that encodes information through controlled variations of the rotor PWM duty cycle, generating distinctive vibration patterns measurable in the frequency–displacement domain. By modifying the ArduPilot firmware and conducting extensive SITL simulations, we validate that the proposed encoding method can be applied simultaneously to all four motors without compromising flight stability, altitude control, or mission feasibility. From the experimentally derived micro-dynamics, we model the resulting channel using two symbol constellations: a binary constellation, offering higher robustness to displacement noise, and a quaternary constellation, providing double the information per symbol at the cost of reduced noise tolerance. These models reveal an order-of-magnitude difference in acceptable displacement perturbations and enable a rigorous evaluation of error-control techniques. We assess both classical and modern Forward Error Correction (FEC) schemes and perform Monte-Carlo simulations using convolutional and LDPC codes, demonstrating substantial BER reductions even under noise levels producing approximately 10% uncoded BER. Reliability is further enhanced by integrating a Stop-and-Wait ARQ mechanism, yielding near-perfect frame delivery with practical airtime overhead. The combined results show that this vibration-based channel can reliably transmit payloads of 128–254 bits within sub-minute flight durations, enabling a covert or auxiliary communication link from a UAV to a ground control station. Thanks to its independence from RF channels, resilience to jamming, and inherent physical coupling to UAV actuation, the proposed system represents a promising building block for strengthening UAV security and authenticity in adversarial environments.