{"id":"a0b7d4c9-83be-4855-9927-58c9328ad00c","arxiv_id":"1908.04633","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A WFRFT-based multi-beam directional modulation scheme replaces artificial noise with a keyed transform, saving about 1 dB of transmit power while keeping signals secure from eavesdroppers near or at the users' locations.","lead":"This paper combines two existing wireless security techniques, WFRFT and frequency-diverse-array directional modulation, to build multi-beam transmitters that need no artificial noise. This saves transmit power because the transform itself scrambles signals for eavesdroppers while legitimate receivers can undo it.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The printed WFRFT definitions are inconsistent: Eq. (4) is not the M_V=N_V=0 case of Eq. (3), and if Eq. (4) is used F^{-α}F^α ≠ I, so Bob recovery in Eq. (30b) is unsupported.","rationale":"The central claim needs each Bob to recover √Ps s_k plus Gaussian noise, which in turn needs the WFRFT used in Eqs. (30) and (44) to be invertible and to preserve Gaussian noise statistics. The printed equations do not provide a single transform with the required properties: Eq. (3) is a valid unitary multi-parameter WFRFT, but Eq. (4) is not its specialization and is itself non-unitary, making F^{-α}F^α≠I. This is a verification and reproducibility gap rather than evidence of fraud: if the simulations used Eq. (3), the main power-efficiency and BER conclusions may survive; if they used Eq. (4), the single-parameter results are invalid. That is exactly the conditionality the reader expressed. I retain the CONDITIONAL verdict because the issue is concrete, testable, and fixable, and because the reader's weakest-assumption selection matches this concern. Other issues, such as the optimal-Eve model and the same-location caveat, are secondary: they would weaken the security claims but would not by themselves overturn the Bob-side recovery if Eq. (3) is the implemented transform.","tokens_in":23776,"tokens_out":21830,"duration_ms":205031,"concrete_test":"Set M_V=N_V=0 and α=0.5; compute coefficients from Eq. (3) and from Eq. (4). Then apply the Eq. (4)-based F^{0.5} followed by F^{-0.5} to a length-4 vector proportional to [1,-1,1,-1]^T, an eigenvector of D with eigenvalue -1. If the result is zero rather than the input, Eq. (4) fails F^{-α}F^α=I. Independently check whether Eq. (3) with M_V=N_V=0 has unit-modulus frequency coefficients (it does), so the correct fix is to replace Eq. (4) with the complex formula and re-run Fig. 12(b) with the corrected single-parameter transform.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section II defines the multi-parameter WFRFT by Eq. (3), but the claimed single-parameter reduction in Eq. (4) is not the M_V=N_V=0 specialization of Eq. (3). Setting M_V=N_V=0 in Eq. (3) yields ω_i=(1/4)Σ_{k=0}^3 exp{-jπk(α-i)/2}, which is generally complex; for α=0.5, i=0 this is 0.25 - j(1+√2)/4, whereas Eq. (4) gives the real product cos(π/8)cos(π/4)cos(3π/8)=0.25. The conflict matters because Eq. (4)'s product coefficients are not unitary: for α=0.5 all four coefficients equal 0.25, so the forward transform is 0.25(I+D+D^2+D^3) and the inverse applied after it gives 0.25(I+D+D^2+D^3), not I; the DFT eigenspace with eigenvalue -1 is annihilated. Eq. (30b) and Eq. (44b) rely on F^{-αA}F^{αA}=I and on Gaussian-noise preservation, so the central Bob-recovery claim is not a consequence of the printed equations. If the simulations used the complex-coefficient standard WFRFT of Eq. (3), the derivations can be repaired, but the manuscript never states which formula was implemented.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes two power-efficient multi-beam directional modulation schemes for FDA LoS channels, replacing artificial noise with a secret-parameter weighted fractional Fourier transform. In the cooperative scheme, Alice applies one WFRFT to the K-symbol vector and precodes it to K Bobs; Bobs share their received signals and apply the inverse WFRFT. In the independent scheme, each data stream is WFRFT-transformed with its own length and parameters, padded to a common block length, and transmitted row-by-row. The paper derives BER, secrecy-rate, and robustness expressions and compares the schemes with conventional AN-DM through simulations, claiming an approximately 1 dB SNR saving at BER = 10^-3, positive secrecy rates even for co-located eavesdroppers, and support for independent modulations per Bob.","tokens_in":24127,"tokens_out":10147,"duration_ms":99426,"significance":"If the derivation were fully supported, the proposed approach would be a useful contribution to physical-layer security for FDA-based multi-beam directional modulation: it removes the power split needed for artificial noise, preserves a clean AWGN channel at the legitimate receivers, and provides neighbor security through secret WFRFT parameters. The paper is clearly organized at the system level, gives closed-form BER and secrecy-rate expressions, and addresses parameter leakage, robustness to location errors, and implementation complexity. However, the mathematical foundation of the WFRFT is internally inconsistent as printed, and the independent-scheme derivation contains a dimension mismatch. These issues are load-bearing for the central recovery and security claims, so the current version cannot be accepted without substantial revision.","major_comments":[{"comment":"Equation (4) is not the M_V = N_V = 0 specialization of Eq. (3). Setting M_V = N_V = 0 in Eq. (3) gives omega_i = (1/4) sum_{k=0}^3 exp{-j pi/2 [k alpha - k i]}, which is complex for non-integer alpha; for alpha = 0.5 and i = 0 this equals 0.25 - j(1+sqrt(2))/4, not the real product in Eq. (4). The product-form coefficients in Eq. (4) are not unitary: for alpha = 0.5 all four coefficients equal 0.25, so F^{-alpha}F^alpha is not I and the DFT eigenspace with eigenvalue -1 is annihilated. Consequently, the Bob-recovery steps in Eqs. (30b) and (44b), as well as the claim that WFRFT preserves complex Gaussian noise statistics, do not follow from the printed definitions. The authors must state which definition was implemented; if the standard complex-coefficient WFRFT of Eq. (3) was used, Eq. (4) and the robustness discussion based on M_V = N_V = 0 need to be corrected and the simulations re-examined.","section":"Section II, Eqs. (3) and (4)"},{"comment":"The derivation of the independent-case Eve signal conflates the WFRFT sequence length Q_k with the padded length Q. In Eq. (35), \\tilde{u}_k = F^{alpha_k}(s_k) has length Q_k, while u_k in Eq. (37) is length Q after padding. Equation (49b) then replaces u_{k'} with F^{alpha_{k'}}(s_{k'}) and Eq. (49c) writes eta_{k'} = omega_{1,k'} dot{s}_{k'} + omega_{2,k'} ddot{s}_{k'} + omega_{3,k'} ddot{dot{s}}_{k'} as a length-Q vector, although each term has length Q_{k'}. Because of this dimension mismatch, the equivalent-AN power used in the Eve SINR of Eq. (63) is not established for the padded block. The authors should reformulate the padding/overlap structure and rederive Eqs. (49)-(63) consistently.","section":"Section IV.C, Eqs. (49) and (63)"},{"comment":"The secrecy-rate analysis treats the 'equivalent AN' eta as zero-mean Gaussian noise independent of the message, with variance 1 - |omega_0|^2. However, eta is a deterministic linear combination of the same data symbols s (and their DFTs/permutations), so it is neither independent of the message nor Gaussian in general. Without an explicit threat model in which the WFRFT parameter is a secret key, or a proof that the equivalent-AN term is statistically indistinguishable from Gaussian noise for an uninformed Eve, the expressions log2(1 + SINR) in Eqs. (58)-(65) are not justified as information-theoretic secrecy rates. The authors should state the security model and either justify the SINR-based metric under that model or provide a proper achievable secrecy analysis.","section":"Section V.B, Eqs. (58) and (63)"}],"minor_comments":[{"comment":"The first contribution contains a typo: 'WFRFTT technology' should be 'WFRFT technology'.","section":"Section I, contribution list"},{"comment":"The underbraces in Eq. (32) use 'Ps' both for the transmit power and for the product P s of the precoding matrix and the symbol vector, which is confusing and should be disambiguated with different notation.","section":"Eqs. (32) and (33)"},{"comment":"The caption of Fig. 8 and the related text state the SNR saving 'approximately 1 dB' at BER = 10^-3; it would be helpful to state the simulation marker density or confidence intervals, since the curves in Fig. 8(a) appear to cross within a fraction of a dB.","section":"Section VI, Table II"}],"recommendation":"major_revision","confidential_remarks":"The core issue is a definitional inconsistency between Eq. (3) and Eq. (4) that affects the central recovery derivation. It appears fixable by adopting the standard complex-coefficient WFRFT and correcting the corresponding robustness simulations, but the authors must also clarify the security model behind the secrecy-rate computations. I see no indication of citation or novelty problems; the paper's scope is appropriate for the journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere's my read of the WFRFT-DM paper. The core idea is new and, on its own terms, sensible: instead of spending transmit power on injected artificial noise to hide DM symbols from eavesdroppers, Alice applies a WFRFT with secret parameter α to the symbol vector and precodes the transformed vector. Bobs, knowing α, apply the inverse transform and recover clean symbols; Eves without α see a distorted signal plus what the authors call \"equivalent AN\" (the non-ω0 components of the WFRFT output). The result is a roughly 1 dB saving at BER=10^-3 compared with AN-DM at β1=0.9, and better resilience to neighbor eavesdroppers. The two architectures (cooperative and independent receivers) are genuinely new combinations, and the BER/secrecy-rate derivations for the Bob side are straightforward and mostly clean. The comparison table and robustness simulations are useful.\n\nThe problems are in the foundations. Eq. (3) as printed does not reduce to Eq. (4) when M_V=N_V=0: for α=0.5 it gives complex coefficients, not the real product shown. This matters because the whole Bob recovery in Eqs. (30b) and (44b) leans on F^{-α}F^α=I and on WFRFT preserving Gaussian noise, both of which depend on which form of the coefficients is actually used. The manuscript doesn't state which one was implemented, so the central claim is not verifiable from the printed equations. That's not a nit; it's load-bearing. The stress-test note is right to flag it.\n\nSecond, the secrecy analysis treats the \"equivalent AN\" as Gaussian noise with variance 1-|ω0|^2 and computes Eve's SINR accordingly. That's only a lower bound on Eve's capability in the worst case where she knows α, and the paper admits in V.E.1 that if parameters leak and Eve sits exactly at Bob's location, security fails. The abstract's unconditional \"even if the eavesdroppers are located ... the same as the desired receivers\" is therefore overstated. The independent-scheme Eve derivation also has a dimension mismatch (lengths Q versus Q_k) that should be cleaned up.\n\nThird, the actual security gain over AN-DM is more modest than the \"neighbor security\" rhetoric suggests: without the keyed transform there is no security at identical locations. The power gain is real because you skip the AN power split, but the secrecy-rate advantage depends on a secrecy-by-key argument that should be stated as such.\n\nOverall: a worthwhile engineering paper, with an honest set of simulations and a discussion of parameter leakage, but the error in the WFRFT definition and the unqualified same-location claim mean it needs revision before I'd rely on the results. For a journal with serious refereeing, I'd send it out — it's important enough and the fix is likely straightforward. I'd cite the architecture if I worked in DM power efficiency, and I'd bring it to the reading group mainly to dissect the WFRFT identity.","headline":"A plausible power-efficiency trick (replace AN with a keyed WFRFT) with a clean Bob-side story, but the printed WFRFT definition doesn't add up and the 'same location' security claim is sold too hard.","tokens_in":24715,"tokens_out":4486,"would_cite":true,"duration_ms":43258,"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":"Weighted Fourier transform saves power and blocks eavesdroppers","keywords":["physical layer security","directional modulation","frequency diverse array","weighted fractional Fourier transform","artificial noise","power efficiency","multi-beam transmission","secrecy rate"],"falsifier":"Evaluate the coefficients of Eq. (3) at $M_V=N_V=[0\\,0\\,0\\,0]^T$ and compare them with Eq. (4), then compute $F^{-0.5}F^{0.5}$ on a random QPSK vector using the same coefficient formula the simulations use. If the inverse transform does not return the original vector to machine precision, the receiver equations (30b) and (44b) are not valid, and the claimed SNR saving and neighbor-security results would need to be recomputed with a transform whose inverse property actually holds.","tokens_in":23460,"feed_emoji":"📡","tokens_out":10729,"duration_ms":98133,"temperature":0.7,"pith_summary":"The paper's proposal is to remove the artificial-noise branch from multi-beam directional modulation and let a secret-parameter weighted fractional Fourier transform (WFRFT) do the scrambling instead. In the proposed frequency-diverse-array transmitter, each legitimate Bob applies the inverse transform and recovers $\\sqrt{P_s}s_k+\\xi_k$ (Eqs. (31) and (45)), while an eavesdropper at any other location sees a distorted signal plus an 'equivalent AN' term produced by the transform. The central quantitative claim is a roughly 1 dB SNR saving at BER $=10^{-3}$ relative to AN-aided DM with power split $\\beta_1=0.9$, together with a positive secrecy rate in the exact situation where the AN baseline gives zero---an Eve placed at a Bob's own coordinates. If true, the scheme would make FDA-based physical-layer security essentially free of the power penalty that has motivated the artificial-noise design.","feed_headline":"Weighted Fourier transform saves power and blocks eavesdroppers","feed_subtitle":"Bobs decode cleanly; eavesdroppers, even at the same location, see only distortion.","key_machinery":"The load-bearing object is the 4-order weighted fractional Fourier transform $F^\\alpha(s)=\\omega_0 s+\\omega_1\\dot{s}+\\omega_2\\ddot{s}+\\omega_3\\dddot{s}$, in which $\\dot{s}$ is the normalized discrete Fourier transform of the symbol vector and the four weights $\\omega_i$ are functions of one real parameter $\\alpha$ and two integer vectors $(M_V,N_V)$ via Eq. (3). What carries the argument is the claimed additive property $F^{-\\alpha}F^{\\alpha}=I$, which makes Bob's receiver a simple inverse transform, and the claimed preservation of complex Gaussian noise statistics, which keeps the noise in Eqs. (31) and (45) white with the same variance. The secret parameters serve as the key; for anyone without them the same transform splits the signal into a distorted part and an equivalent-noise part with variance $1-|\\omega_0|^2$, exactly mimicking the role of injected AN but at zero power cost.","core_discovery":"The paper's central claim is that WFRFT can replace artificial noise without sacrificing the directional-modulation guarantee. After Alice computes $u=F^{\\alpha_A}(s)$ and precodes with the Moore-Penrose pseudo-inverse of the Bob steering matrix, a Bob who knows $\\alpha_A$ applies $F^{-\\alpha_A}$ and obtains exactly $\\sqrt{P_s}s_k+\\xi_k$ for his own symbol, with Gaussian noise of unchanged variance. An Eve who does not know the WFRFT key receives $\\sqrt{P_s}\\omega_0 h_E^H P s + \\sqrt{P_s}h_E^H P\\eta + \\xi_E$, where the second term is the equivalent AN; even an Eve at the same range and angle as a Bob is left with the $\\omega_1\\dot{s}+\\omega_2\\ddot{s}+\\omega_3\\dddot{s}$ component and cannot demodulate. This yields the paper's two headline results: power efficiency (about 1 dB SNR gain at BER $10^{-3}$ for $\\beta_1=0.9$, with the gain tied to the AN power split) and neighbor security (positive secrecy rate when an AN-based scheme's secrecy rate collapses to zero because an Eve shares a Bob's location).","pith_inferences":["Inference: The same 'substitute a keyed transform for injected noise' trick need not be tied to FDA; any null-space precoder (phased array, MIMO) could host it, so the power-efficiency result likely generalizes to other multi-beam DM architectures.","Inference: Neighbor security is keyed security, not unconditional information-theoretic security: a collocated Eve who obtains the WFRFT parameters would decode as well as Bob unless parameters are refreshed, so the scheme's guarantee in practice rests on the key-establishment layer and on refreshing the parameters regularly.","Inference: The 1 dB figure is quoted at one operating point ($\\beta_1=0.9$, BER $10^{-3}$); at smaller $\\beta_1$ the AN overhead is larger, so the WFRFT scheme's advantage should widen, and a parametric curve of the saving versus $\\beta_1$ would make the comparison complete.","Inference: Because the WFRFT is computed via FFT, the added complexity is $O(Q\\log Q)$ per data path, which suggests the scheme scales to long blocks and large arrays; extending it to multi-path channels would require the Bobs to equalize the channel before the inverse transform, which is a natural next test."],"forward_implications":["A transmitter can remove the artificial-noise branch entirely and spend all transmit power on the information-bearing precoded signal, which is the source of the roughly 1 dB SNR gain at BER $10^{-3}$ with $\\beta_1=0.9$.","An Eve located at exactly a Bob's position no longer receives an unscrambled copy of that Bob's symbols, because the WFRFT's equivalent-AN term remains in her observation unless the WFRFT parameters have been leaked.","Independent receivers can be served in the same frame with different modulations (BPSK, QPSK, 8PSK in the paper's simulations) and with different WFRFT block lengths, so the scheme supports heterogeneous multi-user traffic without per-user AN allocation.","The secrecy rate of the proposed scheme stays positive when the secrecy rate of the AN baseline is zero (an Eve sharing a Bob's location), and it degrades more slowly than the AN scheme's as the number of cooperating Eves grows.","Practical robustness is claimed within a window: about 0.5 dB SNR loss for a $2^\\circ$ angle error or a 1 km range error, about 1 dB for both together, and about 0.5 dB for a WFRFT parameter mismatch of $\\Delta\\alpha=0.05$."],"supporting_citations":[{"why":"Defines the 4-order WFRFT and states the boundary, periodicity, additivity, and linearity properties that Bob's inverse recovery relies on.","marker":"[28]"},{"why":"Supplies the equivalent-AN interpretation and the variance $1-|\\omega_0|^2$ that the eavesdropper SINR and secrecy-rate expressions use.","marker":"[34]"},{"why":"Provides the zero-forcing AN-aided multi-beam DM synthesis whose precoding matrix and power-split baseline the proposed schemes replace.","marker":"[17]"},{"why":"Gives the secrecy-rate formulation and the AN-aided FDA communication model used for comparison and for defining $\\gamma$ and $\\beta_1$.","marker":"[14]"},{"why":"Defines the symmetrical multi-carrier FDA steering vector on which the angle-range-dependent precoding geometry is built.","marker":"[42]"},{"why":"Introduces the orthogonal AN-based baseband DM synthesis approach that motivates the power-efficiency question and serves as a conceptual baseline.","marker":"[11]"},{"why":"Introduces the alterable-parameter 4-WFRFT whose nine-parameter form Eq. (3) is the keying material for the scheme.","marker":"[35]"}],"fun_headline_variants":["WFRFT beams cut power, keep signals secret","No noise power needed: WFRFT multi-beam security","Even co-located eavesdroppers can't decode WFRFT beams","WFRFT: save power, keep directional modulation secure","Ditch the artificial noise: WFRFT keeps security"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire recovery and equivalent-noise analysis depends on the multi-parameter WFRFT of Eq. (3) being a genuine invertible transform: $F^{-\\alpha}F^{\\alpha}$ must be the identity, and it must send complex Gaussian noise to complex Gaussian noise; as written, Eq. (3) does not reduce to the single-parameter coefficients in Eq. (4) when $M_V=N_V=[0\\,0\\,0\\,0]^T$, so the paper does not actually demonstrate that the transform used in its simulations has these properties.","fun_headline_variants_meta":{"raw":{"variants":["WFRFT beams cut power, keep signals secret","No noise power needed: WFRFT multi-beam security","Even co-located eavesdroppers can't decode WFRFT beams","WFRFT: save power, keep directional modulation secure","Ditch the artificial noise: WFRFT keeps security"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001866,"raw_usage":{"total_tokens":7358,"prompt_tokens":1010,"completion_tokens":6348,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":626,"completion_tokens_details":{"reasoning_tokens":6265}},"tokens_in":626,"tokens_out":6348,"duration_ms":45482,"temperature":1.0,"reasoning_tokens":6265,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:38:03.880422+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the coefficients of Eq. (3) at $M_V=N_V=[0\\,0\\,0\\,0]^T$ and compare them with Eq. (4), then compute $F^{-0.5}F^{0.5}$ on a random QPSK vector using the same coefficient formula the simulations use. If the inverse transform does not return the original vector to machine precision, the receiver equations (30b) and (44b) are not valid, and the claimed SNR saving and neighbor-security results would need to be recomputed with a transform whose inverse property actually holds.","supporting_citations":[{"cited_title":"Research on the application of 4-weighted fractional Fourier transform in communication system,","cited_arxiv_id":null,"evidence_quote":"Defines the 4-order WFRFT and states the boundary, periodicity, additivity, and linearity properties that Bob's inverse recovery relies on."},{"cited_title":"On physical layer security: weighted fractional Fourier transform based user cooperation,","cited_arxiv_id":null,"evidence_quote":"Supplies the equivalent-AN interpretation and the variance $1-|\\omega_0|^2$ that the eavesdropper SINR and secrecy-rate expressions use."},{"cited_title":"Artificial-noise-aided zero-forcing synthesis approach for secure multi-beam directional modulation,","cited_arxiv_id":null,"evidence_quote":"Provides the zero-forcing AN-aided multi-beam DM synthesis whose precoding matrix and power-split baseline the proposed schemes replace."},{"cited_title":"Secrecy capacity analysis of AN-aided FDA communication over Nakagami- m fading,","cited_arxiv_id":null,"evidence_quote":"Gives the secrecy-rate formulation and the AN-aided FDA communication model used for comparison and for defining $\\gamma$ and $\\beta_1$."},{"cited_title":"Dot-shaped range- angle beampattern synthesis for frequency diverse array,","cited_arxiv_id":null,"evidence_quote":"Defines the symmetrical multi-carrier FDA steering vector on which the angle-range-dependent precoding geometry is built."},{"cited_title":"A vector approach for the analysis and synthesis of directional modulation transmitters,","cited_arxiv_id":null,"evidence_quote":"Introduces the orthogonal AN-based baseband DM synthesis approach that motivates the power-efficiency question and serves as a conceptual baseline."},{"cited_title":"Secure communication system based on alterable-parameter 4-weighted fractional Fourier transform,","cited_arxiv_id":null,"evidence_quote":"Introduces the alterable-parameter 4-WFRFT whose nine-parameter form Eq. (3) is the keying material for the scheme."}],"review_version":1}