Rotatable Antenna Enabled Covert Communication

Abstract: Unlike conventional fixed-antenna architectures, rotatable antenna (RA) has shown great potential in enhancing wireless communication performance by exploiting additional spatial degrees of freedom (DoFs) in a cost-effective manner. In this letter, we propose a novel RA-enabled covert communication system, where an RA array-based transmitter (Alice) sends covert information to a legitimate user (Bob) in the presence of multiple wardens (Willies). To maximize the covert rate, we optimize the transmit beamforming vector and the rotational angles of individual RAs, subject to the constraints on covertness, transmit power, and antenna rotational range. To address the non-convex formulated problem, we decompose it into two subproblems and propose an efficient alternating optimization (AO) algorithm to solve the two subproblems iteratively, where the second-order cone programming (SOCP) method and successive convex approximation (SCA) approach are applied separately. Simulation results demonstrate that the proposed RA-enabled covert communication system can provide significantly superior covertness performance to other benchmark schemes.

• The paper introduces a rotatable-antenna system that jointly optimizes beamforming and each antenna’s orientation to direct energy toward Bob while suppressing leakage toward multiple wardens.
• The proposed alternating-optimization method combines an exact SOCP beamforming update with successive convex approximation for antenna rotations, producing monotonically improving covert rates and typically converging within about eight iterations.
• Simulations show consistent gains over fixed-orientation, random-orientation, and isotropic antennas as the array grows and across transmission distances, while performance remains limited by line-of-sight modeling, full CSI assumptions, and unmodeled hardware latency.

Motivation and contribution

Covert communication, or low-probability-of-detection (LPD) communication, aims to conceal the existence of a transmission rather than merely its content, distinguishing it from physical layer security and encryption. Existing covert schemes—channel uncertainty exploitation and artificial noise jamming—are built on fixed-antenna architectures, which leave the spatial degrees of freedom (DoFs) associated with antenna orientation unexploited. This letter proposes an RA-enabled covert communication system in which a uniform planar array (UPA) of $N$ directional rotatable antennas at Alice serves Bob under the scrutiny of $K$ wardens (Willies), each performing energy detection with noise uncertainty. The design jointly optimizes the transmit beamforming vector $\mathbf{w}$ and the rotational angles $\boldsymbol{\Theta}$ of each RA to maximize Bob's covert rate subject to covertness, transmit power, and rotational-range constraints.

The key mechanism is that each RA's boresight can be steered independently, so the aggregate array gain pattern is reconfigurable: radiation can be concentrated toward Bob while deliberately suppressing leakage toward every Willie, directly tightening the covertness constraint without sacrificing legitimate-link gain.

System model and covertness formulation

Each RA is characterized by a 3D pointing vector parameterized by zenith angle $\theta_{z,n}$ (bounded by $\theta_{\max} \in [0,\pi/2]$) and azimuth angle $\theta_{a,n}$. The element gain follows the standard directional pattern $G_0\cos^{2p}(\epsilon)$, where $p$ is the directivity factor and $G_0 = 2(2p+1)$ enforces power conservation. Channels are modeled as free-space line-of-sight (LoS); the authors note this is for tractability and that extension to multipath requires only replacing channel vectors and re-deriving gradients, leaving the algorithm unchanged.

Willie $K$0 performs optimal binary hypothesis testing via energy detection under noise uncertainty $K$1, yielding a minimum detection error probability of

$K$2

with $K$3. Enforcing $K$4 reduces to a per-Willie received-power cap $K$5, where $K$6 depends on the nominal noise power, $K$7, and the tolerance level $K$8. This converts the statistical covertness requirement into a deterministic power-leakage constraint, which is what makes the subsequent optimization tractable.

Two assumptions deserve emphasis: full CSI—including Willie channels—is assumed available at Alice, justified as a performance-limit baseline; the authors argue Willie CSI acquisition is technically plausible via superheterodyne receiver leakage detection ("Ghostbuster"), but robust designs under imperfect CSI remain open.

Alternating optimization algorithm

The non-convex problem couples $K$9 and $\mathbf{w}$0, so it is split into two subproblems solved iteratively:

Beamforming subproblem: with fixed $\mathbf{w}$1, the objective $\mathbf{w}$2 is non-concave, but global phase invariance allows assuming $\mathbf{w}$3 real, after which introducing a slack variable yields a second-order cone program (SOCP) solvable exactly.

Rotation subproblem: with fixed $\mathbf{w}$4, both the objective and the covertness constraints are non-convex in $\mathbf{w}$5. The authors expand $\mathbf{w}$6 via a second-order Taylor expansion; since its Hessian satisfies $\mathbf{w}$7, substituting $\mathbf{w}$8 produces a concave lower bound for the objective (and $\mathbf{w}$9 a convex upper bound for the leakage constraints). Cross-coupling terms between antennas are dropped from the quadratic coefficient matrix to reduce complexity, retaining only diagonal blocks $\boldsymbol{\Theta}$0, which guarantees concavity of the surrogate. The resulting convex problem is solved by CVX within a successive convex approximation (SCA) loop.

Because each AO iteration solves either an exact convex problem or a conservative SCA step, the covert rate is monotonically non-decreasing, and convergence follows from boundedness of the objective. Total complexity is $\boldsymbol{\Theta}$1 for $\boldsymbol{\Theta}$2 iterations. A caveat inherent to SCA: convergence is to a Karush–Kuhn–Tucker point, not the global optimum, so the reported rates should be read as achievable lower bounds on what joint optimization could deliver.

Numerical results

Simulations use $\boldsymbol{\Theta}$3 m, noise power $\boldsymbol{\Theta}$4 dBm, $\boldsymbol{\Theta}$5, $\boldsymbol{\Theta}$6, $\boldsymbol{\Theta}$7, noise uncertainty $\boldsymbol{\Theta}$8 dB, and $\boldsymbol{\Theta}$9 Willies at 30 m, averaged over 100 random realizations. Three findings stand out:

• Convergence: the covert rate stabilizes within approximately 8 iterations across all directivity factors $\theta_{z,n}$0, confirming practicality of the AO scheme.
• Antenna scaling: with $\theta_{z,n}$1 dBm and $\theta_{z,n}$2 m, the RA system consistently outperforms fixed-antenna, random-orientation, and isotropic benchmarks as $\theta_{z,n}$3 grows. Higher $\theta_{z,n}$4 further improves the rate by narrowing beamwidths.
• Distance robustness: at $\theta_{z,n}$5, the RA system retains the highest covert rate over all Alice–Bob distances considered, because orientation optimization compensates path loss more effectively than fixed or random orientations.

The comparison against the isotropic benchmark ($\theta_{z,n}$6) isolates the benefit of directionality itself, while the fixed-orientation benchmark isolates the benefit of optimized rotation; the consistent gap to both indicates the gains stem specifically from boresight adaptivity rather than from directional elements alone.

Limitations and open questions

Several assumptions bound the applicability of the results. The LoS-only channel model, while extendable in principle, means the reported gains have not been demonstrated under multipath fading, where orientation-dependent gains interact with random small-scale effects. Full CSI at Alice, particularly of passive Willies, is a strong assumption whose violation would degrade the guarantee that leakage stays below $\theta_{z,n}$7. The SCA-based rotation update offers no global optimality certificate, and the neglect of inter-antenna cross-coupling in $\theta_{z,n}$8 trades approximation tightness for complexity. Finally, mechanical/electronic reconfiguration latency and hardware feasibility of dense UPA rotation are not modeled, so the effective covert rate in time-varying scenarios remains unquantified.

Conclusion

This letter establishes that rotatable antennas provide a cost-effective lever for covert communication: by jointly optimizing beamforming and per-element boresight angles through an SOCP–SCA alternating optimization, the system maximizes the legitimate rate while provably bounding detectable leakage at multiple wardens. Simulations show consistent covert-rate superiority over fixed, random, and isotropic benchmarks with fast convergence. The results position RA as a candidate architecture for undetectable transmission, contingent on future validation under imperfect CSI, multipath propagation, and realistic actuation constraints.