Secure Dual-Functional Radar-Comm Design

Updated 27 February 2026

The paper demonstrates secure DFRC by jointly designing radar and communication beamformers and integrating artificial noise to mitigate eavesdropping risks.
It leverages advanced hardware like RIS, CRA, and movable antennas to enhance spatial control and physical-layer security.
Robust optimization techniques address uncertainties in CSI and target angles, ensuring reliable radar resolution and secure data transmission.

A secure dual-functional radar-communication (DFRC) system is an integrated platform that simultaneously provides high-performance radar sensing and wireless communication over a shared spectral/hardware resource, while actively mitigating the risk of eavesdropping by adversarial targets. Such systems are structurally foundational for next-generation architectures, including 6G wireless and spectrum-shared sensing–communication networks. The evolution of secure DFRC design encompasses multidimensional trade-offs, encompassing joint waveform/beamforming, robust optimization under parametric uncertainties, the deployment of new reconfigurable hardware (e.g., RIS, MAs, CRA arrays), and strong algorithms for physical-layer security (PLS).

1. System Models and Security Threat Landscape

In canonical DFRC, a multi-antenna transmitter (often a uniform linear array) emits a composite waveform serving both communication users (“Bobs”) and probing radar targets (often potential eavesdroppers, “Eves”) (Su et al., 2019, Dong et al., 2023, Wen et al., 2022). The baseband transmit signal typically takes the form

$\mathbf{x} = W\,\mathbf{s} + \mathbf{n}~,$

with user data, beamforming, and artificial noise (AN) components. For MIMO configurations and advanced setups, these signals can be further shaped in polarization and radiation pattern through EM-domain reconfigurability (Liu et al., 22 Aug 2025). Sensing relies on the direct probing as well as, in cooperative scenarios, the assistance of surfaces (RIS, IRS) or intelligent surfaces (CRA, active RIS) (Zheng et al., 2024, Zhang et al., 2024, Li et al., 2023, Li et al., 2024).

The principal security threat arises from the fact that the radar waveform inherently conveys communication information, allowing targets—by virtue of being in the main beam—to intercept and attempt to decode confidential messages. This adversarial role is especially salient in DFRC/ISAC where the targets are not passive reflectors but intelligent, potentially colluding eavesdroppers with access to system knowledge and favorable channel conditions (Yang et al., 21 Jan 2026, Su et al., 2019, Su et al., 2021).

2. Physical-Layer Security (PLS) Metrics and Constraints

Secrecy rate is the fundamental PLS metric, given by

$R_s = [R_u - R_e]^+~,$

where $R_u$ and $R_e$ are the achievable rates for the legitimate user and the eavesdropper, respectively. These rates are contingent on the received SINRs, which are functions of transmit beamforming, artificial noise shaping, and, where relevant, reconfigurable/reflective surface parameters: $\mathrm{SINR}_u = \frac{|c_u^T w|^2}{\|c_u^T W_n\|^2+\sigma_u^2}~\qquad \mathrm{SINR}_e = \frac{|c_e^T w|^2}{\|c_e^T W_n\|^2+\sigma_e^2}~.$ (Li et al., 2023, Su et al., 2019, Zhang et al., 2024, Li et al., 2024)

Radar performance is simultaneously enforced by maintaining minimum SNR/SINR thresholds on echoes or minimum estimation rates for target parameter extraction: $\gamma_R\;\equiv\; \mathrm{tr}\{C_T(w w^H+W_n W_n^H)C_T^H\}/\sigma_R^2 \;\ge\;\gamma_{\rm req}~,$

$R_e\;\geq\;\zeta~,$

with $\gamma_{\rm req}$ or $\zeta$ tuned based on detection and estimation needs (Wen et al., 2022, Zheng et al., 2024).

Power constraints apply both globally over the transmitter and, in RIS/IRS/active RIS designs, locally per reflecting element or surface (Zheng et al., 2024, Zhang et al., 2024, Li et al., 2023).

3. Secure DFRC Design Methodologies

Joint Beamforming and Artificial Noise

A central theme is the joint design of information and radar beamformers, optionally with AN, to guarantee legitimate SINR, suppress eavesdropper SINR, and achieve desired radar beampatterns:

For MIMO DFRC,

$\min_{W_i,\,R_N}~\max_{\,\theta\in\Phi} \frac{|a(\theta)^H(\sum_i W_i)a(\theta)|} {a(\theta)^H R_N a(\theta) + \sigma^2}$

subject to $\mathrm{SINR}_i \geq \gamma_0, ~\mathrm{tr}(\sum_i W_i + R_N) \leq P$ and rank constraints (Su et al., 2019, Su et al., 2019, Dong et al., 2023).

Artificial noise is often aligned in the nullspace of user channels or steered directly toward eavesdropper directions to maximize confusion (Su et al., 2019, Li et al., 2023, Wen et al., 2022).
For multi-objective trade-off, SDR techniques are commonly used to obtain optimal (tight) solutions, with fractional programming (FP) and Dinkelbach-type transformations handling nonlinear secrecy metrics (Su et al., 2019, Li et al., 2024, Zheng et al., 2024).

Symbol-Level and Directional Modulation

Directional modulation and symbol-level constructive/destructive interference techniques exploit the fine-grained structure of constellations to scramble signals in undesired directions, providing security even when eavesdroppers reside in the main beam. Key constraints include guaranteeing that CUs receive their intended symbol in the decision region (constructive interference, CI) and pushing the interferer's symbol out of the correct region (destructive interference, DI) (Su et al., 2021, Su et al., 2021): $|{\rm Im}\{\tilde{h}_k^H x\}| \leq [ {\rm Re}\{\tilde{h}_k^H x\}-\sqrt{\sigma^2 \Gamma_k} ] \tan\phi~,$ with additional DI constraints at the target.

RIS/IRS and EM-Domain Reconfiguration

Reconfigurable intelligent surfaces (RISs) introduce a new design dimension, enabling synthesis of joint active (transmit beamformer) and passive (surface phase shift) strategies to focus energy toward users and null toward eavesdroppers (Li et al., 2023, Zheng et al., 2024, Li et al., 2024): $\max_{w,\,\Phi}~R_s(w,\,\Phi)\;;\quad |\Phi_{n,n}|=1~;\; \gamma_R \geq \gamma_{\rm req}~.$ Compound reconfigurable antenna (CRA) arrays push this further by enabling joint pattern and polarization state selection, vastly increasing spatial and EM-domain DoF for joint security and sensing (Liu et al., 22 Aug 2025). The resulting design is a mixed-integer nonlinear program over EM mode selection, BB-domain precoding, and combining, tackled via alternating optimization, fractional programming, and SOCP relaxations.

Advanced Hardware: Active RIS and Movable Antenna Arrays

Active RISs can amplify incident signals, providing resilience against multiplicative fading and additional power for both secure communication and radar sensing (Zhang et al., 2024). Movable antenna arrays introduce additional channel DoF by physically varying antenna positions, which can be jointly designed for covert communication under detection error probability (DEP) constraints against multiple colluding Willies (Yang et al., 21 Jan 2026).

4. Robust Optimization under Uncertainty

Robust secure DFRC design must account for:

Uncertainty in eavesdropper (target) angles (modeled as intervals or finite sets) (Su et al., 2019, Dong et al., 2023).
Channel state information (CSI) errors for legitimate users and targets, often handled by S-procedure–induced LMIs, or moment-based random phase-error models (Wen et al., 2022, Zheng et al., 2024).
In RIS-aided systems, uncertainties in the cascaded reflection channels are encompassed by bounded-error models, with worst-case secrecy rates enforced via LMI constraints and S-procedure relaxations (Zheng et al., 2024, Li et al., 2023, Li et al., 2024).
For covert DFRC, the KL-divergence is constrained to ensure a high minimum detection error probability at Willies, leading to covertness constraints of the form

$D( \mathbb{P}_0 \|\mathbb{P}_1 ) \leq 2\epsilon^2~.$

(Yang et al., 21 Jan 2026)

In most architectures, robust designs incur additional computational complexity but ensure secrecy and radar performance guarantees under worst-case channel conditions.

5. Algorithmic Frameworks and Solution Strategies

The high-dimensional, non-convex nature of secure DFRC design problems requires sophisticated optimization algorithms:

Semidefinite relaxation (SDR): Relaxation of rank constraints to yield convex SDPs, with tightness proven for canonical beamforming problems (Dong et al., 2023, Zheng et al., 2024).
Fractional programming & Dinkelbach: Transformation of secrecy/max-ratio objectives into difference or parameterized forms amenable to convex optimization (Su et al., 2019, Li et al., 2024, Zheng et al., 2024).
Majorization-minimization (MM): Quadratic surrogates constructed for non-convex radar SNR or IRS-induced objective terms, leading to iterative tractable optimization (Zhang et al., 2024, Li et al., 2024).
Manifold and Riemannian optimization: Efficiently handle unit-modulus constraints in IRS/RIS phase optimization, outperforming SDR for large IRS arrays (Zheng et al., 2024, Li et al., 2024).
Alternating optimization (AO): Decomposition of the global problem into beamforming (active), surface phase (passive), and power allocation sub-blocks for parallel or sequential update (Li et al., 2023, Li et al., 2024, Zhang et al., 2024, Zheng et al., 2024).

Table: Summary of Core Algorithms Adopted in Secure DFRC

Method	Problem Context	Typical Complexity
SDR + Dinkelbach	Beamforming/AN/IRS phase	$O(N^6)$ (SDR), $O(N^3)$ (manifold, per iter)
Alternating Optimization	Joint {beam, surface} design	$O(N_{out}(N^{6}+M^6))$
SCA / MM	Non-convex radar SNR	$O(N^3\sqrt{K+1})$ per iteration
Block Coordinated Descent	MA, BB/EM, RIS systems	Hierarchical, per-block complexity (see respective refs)

6. Hardware-Aided and Physical-Layer-Aware Secure DFRC

Emergent hardware—RIS, IRS, CRA, active RIS, and movable antennas—provides additional controllable DoFs, leading to significant improvements in secrecy rate and radar sensitivity:

RIS/IRS can realize 3–5 dB secrecy rate improvement over conventional DFRC, with nearly linear gains as the number of elements increases (Zheng et al., 2024, Li et al., 2023, Zhang et al., 2024).
CRA (pattern+polarization reconfigurable) gains up to 12 dB in radar SCNR, and supports robust secrecy even under clustered, correlated-channels (Liu et al., 22 Aug 2025).
Active RIS counters multiplicative fading and outperforms passive RIS under equivalent power budgets (Zhang et al., 2024).
Movable antennas facilitate dynamic spatial “nulling” and DoF-matched shaping of both the legitimate and eavesdropper channels, supporting higher covert communication rates even in the presence of colluding adversaries (Yang et al., 21 Jan 2026).

Security is further enhanced via physical-layer scrambling schemes—frequency-hopping permutation, directional modulation, symbol level interference, and radar-based masking—each directly rendering the intercepted signal by the target either undecodable or effectively randomized (Wu et al., 2020, Su et al., 2021, Shi et al., 2024).

7. Performance Trade-offs and Practical Design Guidelines

The joint optimization of secrecy and radar metrics inherently commands system-level trade-offs:

Increasing legitimate user SINR (QoS) often diminishes available DoFs for jamming or scrambling the information at the eavesdropper and may degrade the radar beampattern (Dong et al., 2023, Su et al., 2019).
Enforcing tighter secrecy-requirements (lower allowable $R_e$ or SINR at Eves) can result in a broader main-lobe/beamwidth for radar, thereby reducing resolution (Su et al., 2019, Su et al., 2019, Li et al., 2024).
Artificial noise, while boosting secrecy, subtracts from the SNR for legitimate detection, necessitating careful power-splitting (Li et al., 2023, Wen et al., 2022).
RIS/CRA dimension increases deliver secrecy and radar gains but with diminishing returns, especially in presence of channel or hardware quantization errors (Zheng et al., 2024, Liu et al., 22 Aug 2025).
Robust design machinery, although computationally heavier, is essential in practical deployments owing to the prevalence of angular/CSI uncertainties.

Typical practical recommendations include:

Use SDR-based beamforming only with robustification if CSI/angle estimates are imperfect.
Power-allocate 70–90 % to information beams when LoS is favorable, and to AN/jamming beams as target channel strength increases.
For RIS design, deploy N≥64 if closed-form MM/bisection methods are used (to ensure O(N) complexity scalability) (Li et al., 2024).
In physical-layer scrambling scenarios, frequency-hopping and random sign reversal can force eavesdropper SER to 1.0, providing resilient security at modest complexity (Wu et al., 2020).