{"id":"b97e1baa-482e-4d99-a229-80e246e19109","arxiv_id":"2411.08386","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"An alternating optimization algorithm that jointly tunes secure beamformers and fluid antenna positions can raise the secrecy rate of a downlink NOMA system whose edge user may eavesdrop.","lead":"This paper designs a secure beamforming scheme for a downlink NOMA system whose base station uses fluid antennas, and it optimizes both beamforming vectors and antenna positions to maximize secrecy rate. A reader interested in 6G physical layer security would find a numerical demonstration that combining fluid antennas with NOMA can improve secrecy rate over fixed-antenna baselines.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The position subproblem in Sec. III-B reverses the NOMA rate-constraint sign: μ_{i,j}=w'_{1,i}w_{1,j}−L_r w'_{2,i}w_{2,j} makes (25) enforce d_{k,1}−L_r d_{k,2} ≥ L_r σ² instead of d_{k,2}−L_r d_{k,1} ≥ L_r σ², so the optimized positions may violate (9b).","rationale":"The reader's weakest_assumption concerned the general validity and convergence of the convex approximations, which is a legitimate issue but is not the most decisive defect. A closer reading of the position-optimization subproblem reveals a concrete sign/order error in the transformation of the NOMA rate constraint: the μ matrix in Section III-B is defined as w'_{1,i}w_{1,j} − L_r w'_{2,i}w_{2,j}, which yields d_{k,1} − L_r d_{k,2} rather than the required d_{k,2} − L_r d_{k,1}. Consequently, constraint (25) imposes the opposite condition to (9b). This is not an external critique about missing proofs or convention disagreement; it is an internal algebraic inconsistency in the exact transformation of a constraint that the algorithm must preserve. If implemented as written, the optimized positions can violate the minimum rate requirement r, making the secrecy-rate comparisons in Figs. 1–2 unreliable. This single error is enough to invalidate the central numerical claim as presented. The verdict should move from CONDITIONAL to REJECT because the core algorithmic derivation contains a reversible-sign error that would require a corrected formulation and new simulations to restore confidence. I also note the abstract's statement that each user has a fluid antenna contradicts Section II's 'fixed-position antenna' model; while less central, it adds to the need for a careful revision.","tokens_in":8186,"tokens_out":23931,"duration_ms":214922,"concrete_test":"Re-derive the transformation of (9b) in Section III-B: with d_{k,q}(t) = w†_q h_k h†_k w_q, the constraint (9b) is d_{k,2}(t) − L_r d_{k,1}(t) ≥ L_r σ², so μ_{i,j} must equal w'_{2,i}w_{2,j} − L_r w'_{1,i}w_{1,j}. Substitute the paper's μ into the left side of (25) and verify it equals d_{k,1}(t) − L_r d_{k,2}(t); if so, run the published AO algorithm (or re-solve the position subproblem with the corrected μ) and evaluate the final R_{c,2} and R_{e,2}. If R_{k,2} < r for the published version, the algorithm violates (9b) and the reported secrecy rates are invalid.","verdict_should_be":"REJECT","load_bearing_attack":"In Section III-B, the paper transforms constraint (9b) R_{k,2} ≥ r into the position-dependent inequality Σ_{i,j} μ_{i,j} g†_k(t_i)V_k g_k(t_j) ≥ L_r σ², with μ_{i,j} = w'_{1,i}w_{1,j} − L_r w'_{2,i}w_{2,j}. Since d_{k,q}(t) = Σ_{i,j} w'_{q,i}w_{q,j} g†_i V_k g_j, this choice of μ makes the left side equal d_{k,1}(t) − L_r d_{k,2}(t). But (9b) is equivalent to d_{k,2}(t) − L_r d_{k,1}(t) ≥ L_r σ², as correctly used earlier in Eq. (17). Thus the position subproblem (25) enforces the opposite inequality: it requires the s1 signal power to dominate the s2 signal power, rather than the SIC-compatible condition that s2 be strong enough to decode. If the authors implemented (25) as written, the optimized FA positions will generally not satisfy the minimum rate requirement r; the secrecy rates reported in Figs. 1–2 may correspond to infeasible or constraint-violating solutions. This is a concrete algebraic error in the core derivation, not merely a missing convergence proof, and it directly undermines the central numerical claim. Even if the intent was a typo, the manuscript as written does not support the claimed result without a corrected derivation and rerun simulations.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The letter studies a downlink MISO NOMA system in which the base station has M fluid antennas and serves a cell-center user (CU) and a cell-edge user (CEU), with the CEU treated as a potential eavesdropper. The authors formulate a secrecy-rate maximization problem over transmit beamforming vectors and fluid-antenna positions, and propose an alternating optimization (AO) algorithm. The beamforming subproblem is handled by majorization-minimization, and the antenna-position subproblem is handled by Taylor-based approximations that are claimed to yield convex subproblems solvable by CVX. Numerical results compare the proposed scheme against fixed-position antenna, random-position antenna, and OMA baselines, and report secrecy-rate gains.","tokens_in":8564,"tokens_out":8541,"duration_ms":87088,"significance":"If the derivation were correct, the paper would provide a useful design for combining fluid-antenna position optimization with secure NOMA beamforming, with a concrete geometric channel model and a benchmark comparison including FPA, RPA, and OMA. The problem formulation is relevant to physical-layer security for FAS-assisted NOMA, and the AO/MM structure is a standard and plausible approach. However, the paper provides no proofs, no reproducibility artifacts, and, as detailed below, contains a sign error and a convexity error in the position subproblem that directly affect the validity of the reported results. The central claim is therefore not currently supported by the manuscript.","major_comments":[{"comment":"The definition μ_{i,j}=w'_{1,i}w_{1,j}−L_r w'_{2,i}w_{2,j} makes the left side of the transformed constraint equal d_{k,1}(t)−L_r d_{k,2}(t). Constraint (9b), as correctly rewritten in Eqs. (10) and (17), is equivalent to d_{k,2}(t)−L_r d_{k,1}(t) ≥ L_r σ². Thus Eq. (25) enforces the opposite inequality: it requires the s1 signal power to dominate the s2 signal power, rather than the SIC-compatible condition that s2 be decodable with the required rate r. As written, the position subproblem can return FA placements that violate (9b), so the secrecy rates in Figs. 1–2 cannot be attributed to feasible solutions of problem (9). The coefficient should be μ_{i,j}=w'_{2,i}w_{2,j}−L_r w'_{1,i}w_{1,j}, and if the simulations were run using the reversed expression, they need to be rerun.","section":"Section III-B, Eq. (25)"},{"comment":"The paper states that (23) is convex with respect to t_m and that problem (31) is convex. However, b^l_{k,q}(t_m) in (23) contains a negative quadratic term −(β_{k,m}/2)||t_m−t^{(l)}_m||², so it is concave, not convex. Meanwhile ψ(τ,ε_e;τ^{(l)},ε^{(l)}_e) defined before Eq. (16) is convex in (τ,ε_e). Constraint (24) therefore takes the form concave ≥ convex, which is not a convex constraint in general. Consequently, problem (31) is not obviously convex and cannot be solved by CVX as claimed. The authors need either a valid convex surrogate for (24) or a different algorithm with a proof that the feasible set is convex or that the iterative procedure retains feasibility.","section":"Section III-B, Eqs. (23)–(24) and (31)"},{"comment":"No convergence or feasibility guarantee is provided for the alternating procedure. The manuscript labels (18) and (31) as convex recasts, but they are inner approximations via Taylor bounds; the paper does not prove that the bounds are globally valid or that the AO sequence converges to a locally optimal or even feasible point of the original problem. In particular, the lower bound in (23) requires a Lipschitz constant β_{k,m} that is stated but not derived, and no monotonicity or fixed-point argument is given for the alternating updates. Even if the sign error and convexity issue are corrected, this missing analysis leaves the algorithm's output without a formal claim to optimality or feasibility.","section":"Section III, overall AO convergence"}],"minor_comments":[{"comment":"The abstract says the CU and CEU are each equipped with a fluid antenna, while Section II states that both users are equipped with fixed-position antennas and Section II-B takes the receive field response vectors f_c and f_e to be all-ones. Please correct this inconsistency and align the abstract with the actual system model.","section":"Abstract and Section II"},{"comment":"The notation w'_{q,i} is not defined; it presumably denotes the complex conjugate of w_{q,i}. Please define it explicitly, since the subsequent quadratic-form manipulations depend on it.","section":"Section III-B, Eq. (22) and surrounding text"},{"comment":"The derivation of d_{k,q}(t_m) as Σ_{n=1}^M ξ_{n,m} g†_k(t^{(l)}_n)V_k g_k(t_m) is written as an equality, but the actual quadratic form contains additional terms with indices i=m, j≠m; these are later accounted for through the 2Re and W^m_k,q terms in the lower bound. This is confusing and should be rewritten as a lower-bound construction rather than an equality.","section":"Section III-B"},{"comment":"There are several typos and notation issues: the duplicated 'h_c and h_c' in Section II-A, the subscript/superscript q appearing inconsistently in μ^q_{1,m}, and the misspelling 'assited' in the conclusion. The names 'Vandenberghe' and the indexing of L_r^{(k)} vs. r in (10)–(17) should also be checked for consistency.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The sign error in Eq. (25) and the false convexity claim in Eq. (31) are load-bearing. If the authors can correct the derivation and provide corrected numerical results, the paper may be salvageable. I would ask the authors to also share or describe the simulation implementation in enough detail to verify that the corrected constraint, not the printed one, was used in the reported figures."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague — quick take: the problem formulation is new and the beamforming half is competent, but the position subproblem flips the sign of the NOMA rate constraint, so the reported secrecy rates probably rest on infeasible antenna positions.\n\nWhat's new and good. This is the first combination I've seen of secure beamforming, fluid antenna systems, and NOMA in a downlink with the cell-edge user treated as an eavesdropper. The AO decomposition — optimize beamformers with positions fixed, then positions with beamformers fixed — is a natural fit, and the MM/Taylor machinery in the beamforming subproblem is standard and largely correct. The numerical comparisons against fixed-position, random-position, and OMA baselines are reasonable. The citation pattern looks normal for a systems letter; the relevant FAS-NOMA and physical-layer-security references are there.\n\nWhere it falls apart. In Sec III-B, the authors rewrite (9b) using μ_{i,j} = w'_{1,i}w_{1,j} − L_r w'_{2,i}w_{2,j}. But (9b), via (10), is equivalent to d_{k,2} − L_r d_{k,1} ≥ L_r σ². Their μ produces d_{k,1} − L_r d_{k,2} ≥ L_r σ² — the opposite inequality. So (25) does not enforce the SIC rate requirement; it enforces its negation. The stress-test note is correct. This is a concrete algebraic error in the core derivation, not a missing convergence proof. If the simulations used (25) as written, the secrecy rates in Figs. 1–2 correspond to positions that violate the minimum rate r. The results may be salvageable if this is a typo and the sign in μ is swapped, but the manuscript as written does not support the central claim.\n\nOther soft spots are secondary: (18) and (31) are called convex 'recasts' but are really inner approximations via Taylor bounds, with no proof of global validity or AO convergence. The abstract says users each have a FA while the system model gives them fixed-position antennas. Simulation details (FA region size, position sampling) are thin. These are normal for a letter, but they compound the main problem.\n\nRecommendation: I wouldn't cite this in its current state. The idea is worth pursuing, and the fix is straightforward, so I'd send it to peer review rather than desk reject — a competent referee will catch the sign flip and the authors can rerun the simulations. For now, treat the numerical claims as unverified.","headline":"The FAS+NOMA secure beamforming idea is new and the beamforming optimization is sound, but the position subproblem reverses the NOMA rate constraint, so the reported secrecy gains likely come from infeasible antenna positions.","tokens_in":9074,"tokens_out":3920,"would_cite":false,"duration_ms":35894,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Jointly optimizing secure beamforming and fluid antenna positions is claimed to maximize the achievable secrecy rate in a downlink NOMA system, outperforming fixed-antenna NOMA and OMA baselines in simulations.","keywords":["fluid antenna systems","NOMA","secure beamforming","physical layer security","secrecy rate","alternating optimization","majorization-minimization","MISO"],"falsifier":"Directly evaluate the returned $w_1^*, w_2^*, t^*$ by computing the original constraints (9b)-(9e) and the secrecy rate (8). If the point violates $R_{k,2} \\ge r$ or $\\|t_m - t_k\\|_2 \\ge D$, or if running the alternating algorithm from many random initializations yields materially different secrecy rates, the claim of a locally optimal secure design would be contradicted.","tokens_in":1386,"feed_emoji":"📡","tokens_out":2027,"duration_ms":56980,"temperature":0.7,"pith_summary":"This paper argues that letting the base station's antennas move, not just steering beams, can substantially improve physical-layer secrecy in a downlink non-orthogonal multiple access system. Treating the cell-edge user as a potential eavesdropper, it maximizes the achievable secrecy rate by jointly optimizing transmit beamforming vectors and the positions of M fluid antennas. The authors propose alternating optimization that iterates between a convex beamforming subproblem and a convex per-antenna position subproblem, using majorization-minimization and Taylor expansions to handle nonconvex terms. Numerical simulations show secrecy-rate gains over fixed-position antenna NOMA, random-position antenna NOMA, and fluid-antenna OMA baselines.","feed_headline":"Moving antennas lifts secrecy rate beyond fixed-antenna NOMA","feed_subtitle":"Jointly optimizing beamforming and fluid-antenna positions beats fixed-antenna NOMA and OMA in simulated downlinks.","key_machinery":"The central machinery is alternating optimization combined with majorization-minimization surrogates. Slack variables $\\tau$ and $\\epsilon_e$ split the fractional secrecy objective, the bilinear product $\\tau\\epsilon_e$ is bounded by a first-order Taylor surrogate, convex quadratic terms in $w_1$ are lower-bounded by tangent lines, and the quadratic term $g_e^\\dagger(t_m)V_e g_e(t_m)$ is upper-bounded using the maximum eigenvalue of $V_e$. Each subproblem is thereby recast as a convex program, solved with standard convex solvers, and the antenna positions are updated one at a time while the others stay fixed.","core_discovery":"The paper claims that moving the transmit antennas is itself a degree of freedom for secrecy: by reconfiguring antenna positions, the base station shapes the channel to the intended cell-center user more favorably than to the eavesdropping cell-edge user, and it does so jointly with secure beamforming. Specifically, the achievable secrecy rate $R_s = R_{c,1} - R_{e,1}$ is maximized under a cell-edge rate requirement, a total power constraint, and a minimum antenna separation $D$; the resulting alternating optimization yields a locally optimal secure beamforming pair and fluid-antenna position set, with computed secrecy rates above the tested benchmarks.","pith_inferences":["A natural testable extension is to replace the perfect-channel assumption with a robust design over channel uncertainty, since position optimization would likely need to hedge against estimation error.","Because the algorithm optimizes one fluid-antenna position at a time with the others fixed, a fully joint position update might reach comparable secrecy rates in fewer outer iterations.","The same MM and Taylor machinery transfers to multiple cell-edge users or multiple eavesdroppers by summing the eavesdropper rate terms, suggesting the formulation is not limited to one potential eavesdropper.","If the convergence and feasibility concerns are resolved, the paper's approach would establish antenna position as a spatial resource on par with beamforming power for physical-layer security."],"forward_implications":["Secrecy in a downlink NOMA system can be improved without extra transmit power, purely by re-positioning the base station's fluid antennas alongside the beamforming design.","The alternating algorithm is computationally tractable, since each subproblem is convex and can be solved with standard convex optimization tools.","Simulated secrecy rate increases with the number of fluid antennas $M$, and the proposed design consistently beats fixed-position NOMA, random-position NOMA, and fluid-antenna OMA across the tested power levels and antenna counts.","The approach combines NOMA's spectral efficiency with FAS's spatial flexibility, reducing the channel-similarity problem that normally weakens successive interference cancellation.","The secrecy-rate objective (8) and the rate-constraint reformulations extend directly to other secure multi-user settings, as long as the eavesdropper's channel model is available."],"supporting_citations":[{"why":"Defines fluid antenna systems, establishing the physical premise that antenna positions can be reconfigured within a region.","marker":"[3]"},{"why":"Shows FAS enhancing orthogonal and non-orthogonal multiple access, providing the baseline scenario that this paper extends to secure beamforming.","marker":"[8]"},{"why":"Motivates the security risk in NOMA systems where simultaneous transmissions can facilitate eavesdropping.","marker":"[13]"},{"why":"Supplies the secrecy rate formulation for NOMA, specifically the difference between the intended user's rate and the eavesdropper's rate.","marker":"[15]"},{"why":"Provides the convex optimization framework and solver used to compute both the beamforming and position subproblems.","marker":"[16]"},{"why":"Provides the quadratic upper-bounding technique used to handle the nonconvex constraint (14) in the position update.","marker":"[17]"}],"fun_headline_variants":["Movable antennas boost secrecy in NOMA downlinks","Fluid antennas meet NOMA: better secrecy rates","Reconfigurable antenna positions enhance NOMA secrecy","Joint beamforming and antenna placement lifts secrecy","Secrecy rate jumps with fluid antennas in NOMA"],"cache_read_input_tokens":11136,"weakest_assumption_plain":"The load-bearing premise is that the successive Taylor approximations and the alternating procedure preserve feasibility of the original constraints and converge to a meaningful local optimum; if the surrogate bounds are not globally valid, the final point may violate the rate, power, or antenna-separation constraints.","fun_headline_variants_meta":{"raw":{"variants":["Movable antennas boost secrecy in NOMA downlinks","Fluid antennas meet NOMA: better secrecy rates","Reconfigurable antenna positions enhance NOMA secrecy","Joint beamforming and antenna placement lifts secrecy","Secrecy rate jumps with fluid antennas in NOMA"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000202,"raw_usage":{"total_tokens":1332,"prompt_tokens":842,"completion_tokens":490,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":458,"completion_tokens_details":{"reasoning_tokens":416}},"tokens_in":458,"tokens_out":490,"duration_ms":5512,"temperature":1.0,"reasoning_tokens":416,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T21:38:06.727680+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Directly evaluate the returned $w_1^*, w_2^*, t^*$ by computing the original constraints (9b)-(9e) and the secrecy rate (8). If the point violates $R_{k,2} \\ge r$ or $\\|t_m - t_k\\|_2 \\ge D$, or if running the alternating algorithm from many random initializations yields materially different secrecy rates, the claim of a locally optimal secure design would be contradicted.","supporting_citations":[{"cited_title":"Flu id antenna systems,","cited_arxiv_id":null,"evidence_quote":"Defines fluid antenna systems, establishing the physical premise that antenna positions can be reconfigured within a region."},{"cited_title":"Fluid antenna system enhancing orthogonal and non- orthogonal multiple access,","cited_arxiv_id":null,"evidence_quote":"Shows FAS enhancing orthogonal and non-orthogonal multiple access, providing the baseline scenario that this paper extends to secure beamforming."},{"cited_title":"Physical layer security for NOMA systems: Re- quirements, issues, and recommendations,","cited_arxiv_id":null,"evidence_quote":"Motivates the security risk in NOMA systems where simultaneous transmissions can facilitate eavesdropping."},{"cited_title":"Secrecy Sum Rate Maximization in Non-orthogonal Multiple Access,","cited_arxiv_id":null,"evidence_quote":"Supplies the secrecy rate formulation for NOMA, specifically the difference between the intended user's rate and the eavesdropper's rate."},{"cited_title":"Optimization method s for design- ing sequences with low autocorrelation sidelobes,","cited_arxiv_id":null,"evidence_quote":"Provides the quadratic upper-bounding technique used to handle the nonconvex constraint (14) in the position update."}],"review_version":1}