RSMA-Enhanced Secure Transmit Scheme

• RSMA-enhanced secure transmit scheme is a physical-layer strategy that uses message splitting and superposition coding to maximize spectral efficiency while confounding both internal and external eavesdroppers.
• It leverages advanced beamforming and iterative optimization techniques, such as SCA and alternating methods, to maintain robust secrecy even under imperfect channel conditions.
• The framework supports diverse applications, including IRS-assisted, covert, and cognitive radio systems, achieving notable gains in secrecy rate, energy efficiency, and system fairness.

Rate-Splitting Multiple Access (RSMA)-Enhanced Secure Transmit Scheme refers to the class of physical-layer transmission strategies that jointly exploit message splitting, superposition coding, advanced beamforming, and novel optimization to maximize spectral efficiency and guarantee secrecy in multi-user MIMO/MISO networks. RSMA transcends classical SDMA, OMA, or NOMA by partially decoding interference and injecting structured common information, yielding strong resilience to eavesdropping—both internal (by legitimate system users) and external (by out-of-network eavesdroppers)—while remaining robust to imperfect channel state information (CSI). Modern RSMA security frameworks accommodate MISO/IRS-assisted, cognitive radio, ISAC (Integrated Sensing and Communication), and covert communication regimes.

1. Core System and Signal Model

In RSMA-enhanced secure transmission, a base station (BS) with $N_t$ antennas serves $K$ single-antenna users. Each user's confidential message $W_k$ is split into a common part $W_{c,k}$ and a private part $W_{p,k}$. All common parts are superposed into a single common stream $s_c$; each private part is encoded into private stream $s_k$. The transmit signal is

$$\mathbf{x} = \mathbf{p}_c s_c + \sum_{k=1}^K \mathbf{p}_k s_k,$$

with $\mathbf{p}_c, \mathbf{p}_k \in \mathbb{C}^{N_t}$ the beamforming vectors, under total power constraint $$\mathrm{tr}([\mathbf{p}_c,\{\mathbf{p}_k\}][\cdot]^H) \leq P_t$$. The channel input-output model is $y_k = \mathbf{h}_k^H \mathbf{x} + n_k$, where each $\mathbf{h}_k \in \mathbb{C}^{N_t}$ and $n_k$ is AWGN.

Decoding proceeds via SIC: user $k$ first decodes the common stream (treating private streams as noise), then removes it and decodes its own private stream (treating other private streams as noise). Each user may attempt to decode the private streams of others (internal eavesdropping). External eavesdroppers are modeled via dedicated wiretap channels.

Key SINRs and rate expressions are: $$\Gamma_{c,k} = \frac{|\mathbf{h}_k^H \mathbf{p}_c|^2}{\sum_{i=1}^K |\mathbf{h}_k^H \mathbf{p}_i|^2 + \sigma_n^2}, \quad \Gamma_{p,k} = \frac{|\mathbf{h}_k^H \mathbf{p}_k|^2}{\sum_{i\neq k} |\mathbf{h}_k^H \mathbf{p}_i|^2 + \sigma_n^2}.$$ The achievable secrecy rate for user $k$ is $R_{s,k} = [R_{p,k} - \max_{j\neq k} R_{k,j}]^+$ (Xia et al., 2022), with $R_{p,k} = \log_2(1+\Gamma_{p,k})$ and $R_{k,j}$ defined as the eavesdropper's rate after subtracting decoded streams.

2. Optimization Frameworks and Algorithmic Design

The central design task is to maximize aggregated throughput or fairness, subject to secrecy, power, and feasibility constraints. The two main formulations are:

• Weighted Sum-Rate Maximization:

$$\max_{\mathbf{p}_c, \{\mathbf{p}_k\}, \{C_k\}} \sum_{k=1}^K u_k (C_k + R_{p,k})$$

subject to $R_{s,k} \geq R_{s,k}^{\text{th}}$ for all $k$, power constraints, and common-rate feasibility: $\sum_k C_k \leq R_c$, $C_k \geq 0$, with $R_c = \min_k R_{c,k}$ (Xia et al., 2022, Xia et al., 2022).

• Max-Min Secrecy Rate (especially with external Eve): maximize $\min_k \{ \text{user$k$'s secrecy rate} \}$ over beamformers, rate allocations, and possibly IRS phase-shifts or artificial noise (Gao et al., 2022).

Due to inherent non-convexity (involving quadratic-over-linear, exponential, and "max" constraints), tractable solutions rely on:

• Successive Convex Approximation (SCA): Linearize non-convex terms via Taylor expansion around the current solution; at each iteration, solve a convex QCQP for new beamformers and auxiliary variables (Xia et al., 2022, Xia et al., 2022).
• Alternating/Splitting Methods: In IRS-aided or hybrid analog/digital scenarios, decompose into subproblems (e.g., fixing IRS phase and updating precoders, then vice versa) and alternate updates (Gao et al., 2022, Zhou et al., 30 Nov 2025).
• Weighted MMSE/ADMM: Reformulate RSMA rate optimization via WMMSE duality, with alternating minimization over receiver equalizers and transmit precoders (Xia et al., 2022, Dizdar et al., 2021).
• Deep Reinforcement Learning: For dynamic or covert scenarios, power and rate allocation is cast as an RL problem, optimizing policies using modern DRL (PPO) to account for real-time observations (Hieu et al., 2022).

Convergence is typically guaranteed to a local stationary ($\mathrm{KKT}$) point, with overall complexity polynomial in system dimensions.

3. Physical-Layer Security Principles in RSMA

RSMA-enhanced secure transmission achieves physical-layer confidentiality using layered stream design and beamforming. The common stream, decoded by all users, is immune to internal eavesdroppers, while private streams are transmitted such that their rates at unintended users or Eves are suppressed.

Unique features include:

• Internal Eavesdropping Suppression: By adjusting the power allocation between the common and private layers, RSMA can inject constructive "friendly interference," which confounds internal eavesdroppers (other system users) more effectively than classical schemes (Xia et al., 2022, Xia et al., 2022).
• Robustness to Channel Conditions: As secrecy rate thresholds or CSI uncertainty increase, non-RSMA baselines (MULP/SDMA) exhibit sharply degraded performance, while RSMA’s flexible message splitting and power allocation provide stability.
• Integration with IRS and AN: Extensions to IRS-assisted systems incorporate IRS phase tuning, passive beamforming, and artificial noise injection, further increasing eavesdropper confusion and reducing the IRS size needed for given security targets (Gao et al., 2022).
• Dynamic Interleaving: Recent work proposes dynamically interleaving the common stream's bit sequence based on private bits, preventing eavesdropper decoding even with protocol knowledge but lacking the private-bit index (resulting in high eavesdropper BER) (Abidrabbu et al., 16 Apr 2025).

4. Extensions: IRS, ISAC, Cognitive, Covert, and Green Transmit Designs

RSMA-enhanced secure transmission has been generalized to advanced scenarios:

• IRS-Aided Secure Downlinks: Joint beamformer, AN, and IRS-phase optimization achieves max-min secrecy under non-convex constraints. The AO+SCA+penalty-SDP solution achieves considerably higher secrecy rates and can use fewer IRS elements than SDMA or NOMA (Gao et al., 2022).
• Near-Field ISAC with Hybrid Beamforming: RSMA common stream simultaneously supports downlink data, acts as artificial noise for secrecy, and serves as high-precision sensing waveform. A block coordinate descent (optimizing digital, analog beamformers, secrecy allocations) achieves near-full-digital performance with significantly reduced hardware complexity (Zhou et al., 30 Nov 2025).
• Covert RSMA: Covert transmission is modeled under Kullback–Leibler divergence constraints. DRL-based policy learning enables joint adaptation of rate, power and split, achieving robust positive finite-blocklength covert rates (unlike SDMA which saturates or collapses) (Hieu et al., 2022).
• Cognitive Radio and Jamming: In cognitive systems, RSMA supports simultaneous SU downlink and controlled jamming of adversaries while protecting primary user (PU) coexistence, solved via AO-ADMM and KKT-based threshold design (Dizdar et al., 2021).
• Green Secure ISAC: Joint beamforming over communication, sensing, and RSMA layers maximizes security energy efficiency (SEE); AO over echo beamformer, confidential beamformers, and RSMA beamformers is solved using Taylor expansion, majorization–minimization, and semidefinite programming (SDP), delivering gains in SEE, secrecy, and power usage (Li et al., 19 Feb 2025).

5. Security Metrics, Trade-offs, and Performance

The main quantitative metrics include:

• Secrecy Rate $R_{s,k}$: For internal eavesdroppers, $R_{s,k} = [R_{p,k} - \max_{j\neq k} R_{k,j}]^+$; with external Eve, $R_{s,k}$ generalizes to the difference between legitimate and Eve's achievable rates.
• Weighted Sum-Rate (WSR): $WSR = \sum_k u_k (C_k + R_{p,k})$, optimized subject to secrecy.
• Max-Min Secrecy: $\max_k R_{s,k}$ or $\min_k R_{s,k}$, especially in multi-user fairness critical regimes.
• Security Energy Efficiency (SEE): $\mathrm{SEE} = \frac{R_s}{P_{\rm tot}}$, capturing the trade-off between robust secrecy and energy consumption (Li et al., 19 Feb 2025).
• BER Differential: Particularly with dynamic interleaving, the eavesdropper BER remains close to $0.5$ even when partial interleaving side information is leaked; legitimate users retain low BER across SNR (Abidrabbu et al., 16 Apr 2025).

Trade-offs are controlled by the RSMA policy: as secrecy requirements increase, more power is assigned to private streams, reducing exposure of sensitive information in the common stream. RSMA enables a smooth reallocation across the secrecy-throughput region, outperforming SDMA, NOMA, and non-splitting baselines in both simulated and analytical studies (Xia et al., 2022, Xia et al., 2022, Gao et al., 2022).

6. Practical Implementation and Algorithmic Complexity

The iterative SCA, AO, and WMMSE/ADMM algorithms developed for RSMA-enhanced secure transmission converge to KKT points and have overall polynomial per-iteration complexity in the number of antennas, users, and (when relevant) IRS elements. For instance, the AO+SCA+penalty-SDP iteration in IRS-aided RSMA schemes has complexity $$O(\log(1/\epsilon)[\sqrt{M}(KM^3+K^2M^2+K^3)+\sqrt{N}(N^4+KN^3+K^2N^2+K^3)])$$ per iteration (Gao et al., 2022). In hybrid or ISAC scenarios, linear-algebraic (e.g., closed-form analog-beamfocuser update, SVD extraction) and standard convex optimization (CVX, SDP) dominate.

Convergence is robust and the overhead is feasible for practical deployment, especially given the strong performance gains in secrecy, rate, robustness to imperfect CSI, and energy efficiency. DRL-based strategies, by contrast, offload optimization into policy networks whose complexity is amortized at deployment (Hieu et al., 2022).

7. Summary Table: Key RSMA Secure Transmission Results

Design Scenario	Key Method	Notable Outcomes
Basic MISO RSMA secure beamforming	SCA-based QCQP	WSR and secrecy outperform SDMA/MULP
IRS-aided RSMA secure transmission	AO+SCA+penalty SDP	Max-min secrecy, IRS size reduction
Data-dependent interleaving RSMA	Private bit indexing	Eavesdropper BER near 0.5, robust against SID
Covert RSMA	DRL (PPO) policy	Positive FBL covert rate, adapts over time
ISAC/green RSMA	AO, SDP, Taylor+MM	High SEE, robust secrecy-sensing-rate trade-off
Near-field ISAC RSMA	Block coordinate WMMSE	Threefold use of common stream

The findings demonstrate that RSMA, when engineered for secure transmission, fundamentally enables joint optimization of communication, secrecy, and ancillary functions (e.g., jamming, sensing), providing clear rate and security advantages versus traditional approaches (Xia et al., 2022, Xia et al., 2022, Gao et al., 2022, Abidrabbu et al., 16 Apr 2025, Zhou et al., 30 Nov 2025, Li et al., 19 Feb 2025, Dizdar et al., 2021, Hieu et al., 2022).