TL;DR
This paper introduces a near-field Integrated Sensing and Communications (ISAC) framework designed to secure UAV networks against non-cooperative eavesdroppers. By exploiting the unique properties of spherical wave propagation in the near field, the system can simultaneously track an eavesdropper’s 3D velocity and position while optimizing the legitimate UAV's trajectory and base station beamforming to maximize the secrecy rate.
What Problem Does This Paper Address?
Unmanned Aerial Vehicles (UAVs) are becoming staples of modern wireless networks, but their high-altitude, Line-of-Sight (LoS) links make them sitting ducks for eavesdropping. Traditional physical layer security (PLS) relies on knowing the Eavesdropper's Channel State Information (CSI), which is nearly impossible since eavesdroppers don't exactly send pilot signals to help you block them.
While ISAC (using communication signals to "radar-sense" the environment) has been proposed to find eavesdroppers, most existing research assumes far-field propagation (planar waves). However, as base stations move toward extremely large antenna arrays (ELAA) and higher frequencies, the "near-field" region—where waves are spherical—can extend hundreds of meters. This paper tackles the challenge of performing secure ISAC within this complex near-field environment, where traditional far-field models fail but new opportunities for high-precision sensing emerge.
Methods
The authors propose a two-stage framework:
- Near-Field 3D Tracking: Unlike the far-field, where a moving target produces a single Doppler shift across the array, the near-field exhibits variant Doppler shifts across different antennas because each antenna "sees" the UAV at a slightly different angle. The authors use a Maximum Likelihood (ML) approach to estimate 3D velocities from these spatial variations. They then employ an Extended Kalman Filter (EKF) to fuse these measurements with location predictions, providing real-time tracking of a non-cooperative eavesdropping UAV (E-UAV).
- Joint Optimization: With the E-UAV's location estimated, the system optimizes the Ground Base Station (GBS) beamforming and the legitimate Communication UAV (C-UAV) trajectory. The goal is to maximize the instantaneous secrecy rate while ensuring a minimum sensing accuracy for the E-UAV.
- Algorithmic Approach: The resulting optimization problem is highly non-convex. The authors solve it using Alternating Optimization (AO), splitting the problem into beamforming and trajectory sub-problems. They use Successive Convex Approximation (SCA) and a penalty-based method to handle tricky rank-one constraints on the signal covariance matrices.
Datasets
The paper relies on numerical simulations rather than a physical dataset.
- Setup: A GBS with a $16 \times 16$ antenna array operating at 1.5 GHz (Rayleigh distance of 52m).
- Scenario: An E-UAV performing circular motion while a C-UAV moves within a defined 3D flight area.
- Parameters: Standardized values for radar cross-section (0.03 $m^2$), noise power (-80 dBm), and maximum transmit power (30 dBm).
Key Insights and Ideas
- The "Spatial Doppler" Advantage: In the near field, the Doppler shift is no longer a scalar but a vector across the antenna array. This extra "spatial dimension" of information is what allows the GBS to sense 3D velocity—something much harder to achieve in far-field models.
- Beamfocusing vs. Beamsteering: Near-field propagation allows for beamfocusing, where energy is concentrated on a specific 3D coordinate rather than just a direction. This allows the GBS to "laser-focus" jamming signals (Artificial Noise) on the E-UAV with surgical precision, minimizing leakage to the legitimate user.
- Dual-Purpose Artificial Noise (AN): The AN serves a double role: it jams the eavesdropper's reception and simultaneously acts as the "radar pulse" used to sense the eavesdropper's position via reflections.
Key Results
- High Tracking Accuracy: The proposed EKF-based scheme's estimated trajectory "perfectly aligns" with the ground truth in simulations, proving that near-field spherical wave properties are a powerful tool for localization.
- Rapid Convergence: The Alternating Optimization algorithm converges in approximately 5 iterations, making it potentially viable for real-time UAV applications where low latency is critical.
- Secrecy Performance: The framework achieves a secrecy rate that closely approaches the "perfect knowledge" baseline (where the eavesdropper's path is known in advance). It significantly outperforms "static" baselines where the C-UAV does not adjust its trajectory.
- Sensing-Security Trade-off: The results highlight a clear trade-off: as the requirement for sensing accuracy (tracking MSE) becomes stricter, the GBS must allocate more power to Artificial Noise, which can slightly reduce the communication rate for the legitimate UAV.
Limitations and Future Work
- Single Eavesdropper: The model currently considers only one E-UAV and one C-UAV. Real-world "swarm" scenarios would require more complex multi-target tracking and resource allocation.
- Perfect Legitimate CSI: The authors assume the CSI for the legitimate C-UAV is perfectly known. In practice, this would also need to be estimated, potentially introducing more errors into the optimization.
- Clutter Interference: While the paper assumes clutter (reflections from buildings/trees) is suppressed using standard techniques, the complexity of near-field clutter in urban environments remains a significant practical hurdle.
- Hardware Constraints: The assumption of a full-duplex GBS with perfect self-interference cancellation is a high bar for current hardware, though it represents an ideal target for 6G infrastructure.
Summary updated