{"id":"7c2da328-11f8-4ce8-b580-478c6c85917a","arxiv_id":"1908.04648","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A two-ray multipath extension of FDA-based directional modulation is proposed, with joint design of excitation and artificial noise weights, and its secrecy rate is analyzed and simulated.","lead":"This paper extends frequency diverse array directional modulation to a two-ray ground-reflected channel, designing antenna weights and artificial noise so only the intended receiver gets a clean signal. Simulations show the secure region is narrower than in the single-path model, indicating more focused protection.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Bob-side normalization and AN cancellation in (15) hold only at one time instant because ε_n(t) in (8b) depends on t and Δf_n differ across elements; over a finite symbol duration Bob's signal and AN cancellation drift, so the 'no impact at Bob' claim is not yet established.","rationale":"The reader's weakest assumption is the flat-earth/perfect-ground idealization and far-field path-length approximations in (5)-(6). That is a legitimate model-validity concern. However, the most load-bearing issue is internal: the design constraints (15) are time-instantaneous because ε_n(t) in (8b) depends on t and Δf_n differ by element. Even with a perfect ground, the Bob-side cancellation and AN nulling in (16) only hold at a single time. The paper never specifies a symbol duration or a constraint that the frequency-offset phase rotations are negligible over a symbol; nor does the secrecy-rate analysis include this effect. This is a more fundamental threat to the central claim than the environmental approximation, because it affects the idealized model itself. The issue is addressable by adding a narrowband/short-symbol condition and analyzing the time-average SINR, so the conditional-acceptance verdict stands; no verdict change is needed. I therefore disagree with the reader's identification of the weakest assumption, but not with the resulting CONDITIONAL verdict.","tokens_in":7273,"tokens_out":8826,"duration_ms":103410,"concrete_test":"Take Table I and Section IV parameters and compute |κ_B(t)| and |η_B(t)| over one symbol interval T for T=10 μs and T=0.1 ms. If |κ_B(T/2)| deviates from 1 by more than 5%, or β2^2|η_B(T/2)|^2 is not negligible against σ_ξ^2, then the ideal-Bob SNR in (20) is not achieved for finite-duration symbols.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The design conditions (15) impose κ_B=Σ a_n μ_B ε_B ρ_B = 1 and η_B=0, but ε_B(t)=exp{j2πΔf_n(t-(r_B-d_n u_B)/c)} contains the time variable t. Since Δf_n=Δf ln(|n|+1) are distinct across elements, κ_B(t) and η_B(t) are time-varying. The paper chooses {a_n},{b_n} to satisfy (15) at one instant (implicitly t=0), but for t≠0 the useful-signal contributions at Bob are no longer phase-aligned and the AN term is not exactly canceled. No symbol duration T is specified, and no condition such as max_n Δf_n T ≪ 1 is stated. The secrecy-rate analysis in (17)-(20) silently uses these snapshot values. Consequently, the statement that the inserted AN 'imposes no impact on Bob' and that Bob receives exactly √Ps β1 s in (16) is established only instantaneously, not for a finite-duration symbol. This is an internal correctness risk that remains even under the idealized infinite, perfectly conducting ground: the time dependence is inherent to the FDA frequency offsets, not to the propagation geometry.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends frequency-diverse-array directional modulation (FDA-DM) to a two-ray multipath environment, modeled with a perfectly conducting ground plane and image theory. The authors jointly design the FDA excitation factors {a_n} and the artificial-noise weights {b_n} so that, at the intended receiver Bob, the AN cancels and the useful signal is normalized (constraints (15)). They then define SNR and SINR-based rates ((17)–(20)) and a corresponding 'secrecy rate' ((21)), and present BER and secrecy-rate simulations to argue that the two-ray FDA-DM scheme enables physical-layer secure transmission and yields a more focused secure area than the single-path model.","tokens_in":7495,"tokens_out":4490,"duration_ms":47435,"significance":"If the results are valid, this is a useful extension of FDA-DM from idealized free-space channels to a two-ray multipath scenario, with an explicit, non-iterative design procedure that gives Bob a clean signal while injecting AN into other locations. The paper includes a concrete numerical solution for the design constraints (Table I) and reproduces standard SNR/SINR formulas, which are self-consistent within the stated model. However, the central claims rest on two load-bearing points that need strengthening: the time-varying nature of the FDA phase terms is not handled, and the 'secrecy rate' is an SINR heuristic rather than a proven information-theoretic quantity. The claimed advantage over the single-path model is currently supported only by a few simulation examples.","major_comments":[{"comment":"The design constraints (15) are solved at a single time instant, but ε_B^n(t)=exp{j2πΔf_n(t−(r_B−d_n u_B)/c)} depends explicitly on t, and Δf_n differ across elements. Consequently κ_B(t) and η_B(t) are time-varying, so the equality y_Total(⃗r_B)=√Ps β1 s in (16) holds only at the instant at which (15) is imposed, not over a finite symbol duration. The paper does not specify a symbol duration T or a condition such as max_n |Δf_n| T ≪ 1 that would justify treating the coefficients as quasi-static. Because the SNR/SINR analysis in Section III and the simulations use these snapshot values, the claim that the inserted AN has no impact on Bob is not established for practical finite-duration symbols.","section":"Section II, Eqs. (8b), (15), (16)"},{"comment":"The quantity called 'secrecy rate' in (21) is defined as a difference of pointwise rates ζ(⃗r)=log2(1+SINR(⃗r)), where the AN is treated as Gaussian interference at the eavesdroppers and is assumed perfectly canceled at Bob. This is an SINR-based heuristic; the paper does not show that this ζ_sec is an achievable secrecy rate, a lower bound on secrecy capacity, or even the rate of a specific wiretap code. Furthermore, the minimization in (21) is only over the V eavesdroppers present in the simulation, not over all possible Eve locations, so the claim that the scheme 'is capable of wireless PLS transmission' is not supported as a worst-case guarantee. The paper should either present a proper information-theoretic secrecy formulation or explicitly state that ζ_sec is a DM-system performance metric rather than a proven secrecy rate.","section":"Section III, Eqs. (17)–(21)"},{"comment":"The central comparative claim that the two-ray multipath model achieves a 'more focused secure area' than the single-path model is based only on a few simulation examples with fixed parameters (f0=10 GHz, Δf=2 kHz, h0=4.25λ0, and one Bob location). No analytical expression for the beamwidth or secure area is derived, no metric for 'secure area' is defined, and no sensitivity analysis is provided. As this comparative claim appears in the abstract and conclusion, it needs stronger support—for example, an analytical comparison or a systematic simulation sweep showing that the effect is not an artifact of the particular parameter choices.","section":"Section IV, Figs. 2 and 3"}],"minor_comments":[{"comment":"The abbreviation 'LP' for low-pass filtering is not defined; the authors should state that a low-pass filter removes the carrier term exp{j2πf0t}.","section":"Section II, Eq. (11c)"},{"comment":"The approximation r_LoS ≈ r − d_n u − h0 v and r_NLoS ≈ r − d_n u + h0 v is stated without a derivation of its validity regime; citing [17] is fine, but a brief comment on when the far-field approximation is accurate (e.g., r ≫ array aperture) would improve clarity.","section":"Section II, Eq. (5) and (6)"},{"comment":"The authors say that 'multiple solutions' for {a_n} and {b_n} exist and Table I gives one with N=3, but no numerical method or heuristic for finding these solutions is described. A short explanation of how Table I was obtained would help reproducibility.","section":"Section II, after Eq. (15)"},{"comment":"The text states that secrecy rate increases with smaller β1, but Bob's signal-to-noise ratio in (20) decreases when β1 is smaller. The reason the net secrecy rate increases (presumably because Eve's interference grows faster) should be explained to avoid an apparent contradiction.","section":"Section IV, Fig. 3(a)"},{"comment":"The artificial noise z is a single scalar complex Gaussian injected onto every array element with different weights b_n. This is a valid architecture, but the paper should clarify that this is not a multi-dimensional AN vector, as in some other AN-aided schemes.","section":"Section II, Eq. (3)"},{"comment":"Several equations, such as (1), (5), and (6), use the notation '⃗r_Ante^n' which is a vector position; the dot product in (5) is clear but could be written more explicitly as ⃗e_r · ⃗r_Ante^n = d_n u + h0 v for readers unfamiliar with the notation.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope for a signal-processing journal and addresses a timely topic. The main concern is the unaddressed time-dependence of the FDA phases, which directly affects the validity of Eq. (16) and the subsequent rate analysis. This is fixable by adding a quasi-static condition (e.g., bounding the symbol duration relative to the inverse of the maximum frequency offset) and recalculating the residual interference terms. If the authors also qualify the 'secrecy rate' as a DM-system metric rather than an information-theoretic secrecy capacity, the paper would be closer to acceptance. I do not see grounds for rejection, but the current version needs substantive revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe short version: this is a legitimate, clearly written extension of FDA-based directional modulation to a two-ray multipath channel, and it probably deserves referee time. The catch is that the design constraint that cancels artificial noise at Bob only holds at one time instant. The authors do not state a condition on symbol duration, so the 'no impact at Bob' claim is stronger than what the equations justify.\n\nWhat's new: first paper, as far as the authors' review shows, to combine FDA-DM with a multipath model. The two-ray derivation in Section II is self-consistent: the image-theory approximation is stated, the signal expressions in (7)–(13) check out, and the design problem in (15) follows directly from the model. The simulations showing a narrower BER lobe for the multipath case than the single-path case support the 'more focused secure area' claim, at least for the parameters chosen. No curve fitting; the test is genuinely from the model.\n\nSoft spots, in order. (1) The time dependence in ε_n(t) (8b) is the biggest one. Because Δf_n differs across elements, κ_B and η_B in (15) are functions of t, not constants. The paper solves (15) at a single instant and then treats Bob's received signal as exactly √Psβ1 s in (16). Over a finite symbol interval, the useful signal drifts and the AN does not fully cancel. A condition such as max_n Δf_n T << 1 is needed, or a design that accounts for the time evolution. The authors should add this. (2) The secrecy rate in (17)–(21) is an SINR-based heuristic—treating AN as Gaussian noise and taking log2(1+SINR) as the achievable rate. This is common in DM papers, but it is not a proven secrecy capacity or even explicitly identified as an achievable lower bound. (3) The flat-earth, infinite-conductivity ground with a -1 reflection coefficient is load-bearing, and at Bob's range of 150 km with antenna height 4.25 wavelengths, earth curvature is not negligible. The model is a useful idealization, but the range of validity should be discussed.\n\nWho is this for? Researchers in physical layer security who work on FDA-based DM. They will find a useful first step and a few open problems. A general reader can skip it.\n\nRecommendation: send it to peer review, but with the time-variance condition and the secrecy-rate labeling as mandatory revisions. The core idea is sound; the paper just needs to qualify its own claims.\n\nBest,\n[Name]","headline":"A mostly sound FDA-DM extension to two-ray multipath, but the Bob-side cancellation is only instantaneous—needs a symbol-duration condition.","tokens_in":8009,"tokens_out":7478,"would_cite":false,"duration_ms":77102,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A ground-reflected path can be exploited, not fought, to create a focused secure zone in frequency-diverse-array directional modulation.","keywords":["directional modulation","physical layer security","frequency diverse array","two-ray multipath","artificial noise","secrecy rate","MISOME wiretap channel","multipath propagation"],"falsifier":"Run the same FDA-DM design at $f_0 = 10$ GHz, $h_0 = 4.25\\lambda_0$, with Bob at the paper's location, over ground with measured roughness or a finite dielectric; if the artificial noise cancellation at Bob degrades by more than the model's predicted margin, or if an eavesdropper at an angle where $\\sin(2\\pi f_n h_0 v/c) \\approx 0$ still recovers the symbol, the two-ray assumption is falsified. A simpler check is to numerically simulate the full-wave pattern over a lossy half-space and compare the BER lobe width with the two-ray prediction.","tokens_in":7051,"feed_emoji":"📡","tokens_out":7320,"duration_ms":67892,"temperature":0.7,"pith_summary":"Frequency diverse array (FDA) directional modulation has previously been designed for free-space line-of-sight links; this paper extends it to a two-ray multipath environment with a ground reflection, in a multi-input single-output multi-eavesdropper wiretap channel. The central claim is that the two-ray model works and even improves security: the added reflection produces a more tightly focused secure area around the intended receiver than the single-path model. The paper jointly designs the array excitation factors and the artificial noise weights so that at Bob the useful signal adds coherently and the artificial noise cancels, while at any other location the useful signal is distorted and the noise leaks through. Secrecy rate is derived for the model and verified through bit-error-rate and secrecy-rate simulations. If correct, this offers a practical route to keyless physical-layer security without assuming free-space propagation.","feed_headline":"Ground reflection sharpens the secure zone in FDA-DM","feed_subtitle":"Adding a reflected path lets Alice cancel artificial noise at Bob while distorting every eavesdropper.","key_machinery":"The central object is the two-ray multipath FDA-DM model: a symmetric $(2N+1)$-element frequency diverse array (an array whose elements radiate at slightly different frequencies, creating a range-and-angle-dependent pattern) placed at height $h_0$ above a flat perfect ground, with each receiver seeing a direct line-of-sight path plus a ground-reflected path. The reflected path is treated through image theory, so its geometry is the same as radiation from a mirror array below ground, and the sum of the two paths collapses each element's contribution to $a_n x_n \\mu_n \\varepsilon_n \\rho_n$ with $\\rho_n = j2\\sin(2\\pi f_n h_0 v/c)$. The load-bearing identity is the pair of constraints at Bob, $\\sum_n a_n \\mu_B^n \\varepsilon_B^n \\rho_B^n = 1$ and $\\sum_n a_n b_n \\mu_B^n \\varepsilon_B^n \\rho_B^n = 0$, which jointly enforce a unit useful-signal response and exact artificial-noise cancellation at Bob. This is the mechanism that converts the multipath geometry into a secure transmission design.","core_discovery":"The paper's central claim is that a two-ray multipath environment, modeled by replacing the ground with an image antenna array, is not an obstacle to FDA-based directional modulation but a resource. Modeling the reflected path as an image FDA with a $-1$ reflection coefficient, the total received signal at any point becomes the single-path signal with each element's contribution multiplied by $\\rho_n = j2\\sin(2\\pi f_n h_0 v/c)$, where $h_0$ is the array height and $v = \\sin\\theta$ depends on elevation. This makes the effective channel elevation-dependent. Choosing excitation factors $\\{a_n\\}$ and artificial noise weights $\\{b_n\\}$ to satisfy $\\kappa_B = 1$ and $\\eta_B = 0$ at Bob's location leaves Bob with exactly $\\sqrt{P_s}\\beta_1 s$, free of artificial noise, while eavesdroppers elsewhere fail these constraints and see both a distorted signal and interference. The paper verifies by BER and secrecy-rate simulations that this yields a more tightly focused secure region than the single-path FDA-DM model.","pith_inferences":["If the two-ray approximation transfers to other flat-reflector settings such as walls, water surfaces, or building facades, the same sine-factor design could be re-derived with a complex reflection coefficient and a different image height, giving a testable family of secure-zone shapes.","The elevation-dependent $\\rho_n$ suggests that height $h_0$ and elevation angle could be tuned to place artificial-noise suppression at Bob while creating a natural noise floor at many Eve locations, a degree of freedom the single-path model lacks.","The paper does not study robustness to imperfect Bob position knowledge; a natural extension is to quantify how much the constraints (15) degrade under small errors in range or angle, which would tell whether the narrow secure lobe survives in practice.","Extending to a moving Bob or to correlated Eve positions near Bob would require re-solving the constraints per block or adding randomization, which the proposed dynamic random generation of $\\{a_n\\}$ and $\\{b_n\\}$ already makes plausible."],"forward_implications":["Bob's receiver sees no artificial noise and full useful-signal power, so the demodulator works exactly as in a clean channel.","Eavesdroppers at any location other than Bob's must contend with both a distorted useful signal and residual artificial noise, so their achievable rate is lower.","The secure region shrinks as the number of array elements $N$ grows or the signal power split $\\beta_1$ decreases, giving a concrete design trade-off between array size and security focus.","Secrecy rate rises with SNR and falls with the number of eavesdroppers, matching standard physical-layer-security intuition."],"supporting_citations":[{"why":"Supplies the two-ray ground reflection geometry and image-theory path lengths used in equations (5) and (6).","marker":"[17]"},{"why":"Establishes the FDA-based directional modulation synthesis that this paper extends into multipath.","marker":"[10]"},{"why":"Provides the FDA-DM weighting design approach for secure transmission that is adapted to the two-ray model.","marker":"[11]"},{"why":"Introduces artificial-noise-aided secure transmission with directional modulation, the AN scheme this paper reuses.","marker":"[12]"},{"why":"Supplies the BER metric used to assess the directional modulation system's security.","marker":"[18]"},{"why":"Addresses FDA beamforming for proximal legitimate user and eavesdropper, the comparison baseline for secure-area behavior.","marker":"[16]"}],"fun_headline_variants":["Ground reflection tightens FDA-DM secure zone","Two-ray multipath sharpens directional modulation","Reflected path boosts FDA-DM secrecy","Elevation-dependent FDA-DM improves secure zone"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The design collapses if the environment is not well described by a flat, perfectly conducting ground plane: the path-length approximations and the $-1$ reflection coefficient are what make equations (5), (6), and the design constraints (15) hold exactly.","fun_headline_variants_meta":{"raw":{"variants":["Ground reflection tightens FDA-DM secure zone","Two-ray multipath sharpens directional modulation","Reflected path boosts FDA-DM secrecy","Elevation-dependent FDA-DM improves secure zone"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000316,"raw_usage":{"total_tokens":1754,"prompt_tokens":876,"completion_tokens":878,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":492,"completion_tokens_details":{"reasoning_tokens":821}},"tokens_in":492,"tokens_out":878,"duration_ms":9250,"temperature":1.0,"reasoning_tokens":821,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:36:09.948144+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same FDA-DM design at $f_0 = 10$ GHz, $h_0 = 4.25\\lambda_0$, with Bob at the paper's location, over ground with measured roughness or a finite dielectric; if the artificial noise cancellation at Bob degrades by more than the model's predicted margin, or if an eavesdropper at an angle where $\\sin(2\\pi f_n h_0 v/c) \\approx 0$ still recovers the symbol, the two-ray assumption is falsified. A simpler check is to numerically simulate the full-wave pattern over a lossy half-space and compare the BER lobe width with the two-ray prediction.","supporting_citations":[{"cited_title":"Multipath characteristics of frequency diverse arrays over a ground plane,","cited_arxiv_id":null,"evidence_quote":"Supplies the two-ray ground reflection geometry and image-theory path lengths used in equations (5) and (6)."},{"cited_title":"Directional modulation using frequency diverse array for secure communications,","cited_arxiv_id":null,"evidence_quote":"Establishes the FDA-based directional modulation synthesis that this paper extends into multipath."},{"cited_title":"DM using FDA antenna for secure transmission,","cited_arxiv_id":null,"evidence_quote":"Provides the FDA-DM weighting design approach for secure transmission that is adapted to the two-ray model."},{"cited_title":"Artificial-noise- aided secure transmission with directional modulation based on random frequency diverse arrays,","cited_arxiv_id":null,"evidence_quote":"Introduces artificial-noise-aided secure transmission with directional modulation, the AN scheme this paper reuses."},{"cited_title":"Establishing metrics for assessing the perfor- mance of directional modulation systems,","cited_arxiv_id":null,"evidence_quote":"Supplies the BER metric used to assess the directional modulation system's security."},{"cited_title":"Physical-layer security for proximal legitimate user and eavesdropper: a frequency diverse array beamforming approach,","cited_arxiv_id":null,"evidence_quote":"Addresses FDA beamforming for proximal legitimate user and eavesdropper, the comparison baseline for secure-area behavior."}],"review_version":1}