{"id":"7ad1ecb4-772f-412e-8b5f-86a3e21003ce","arxiv_id":"2506.22082","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Measured indoor tests at 3.55 GHz show a varactor-based smart reflector can raise the legitimate user's power above an eavesdropper's for tested positions, while the proposed codebook method is not experimentally validated.","lead":"This paper reports indoor measurements showing that a smart reflector can strengthen wireless privacy at 3.55 GHz by boosting the intended user's signal while weakening an eavesdropper's. It compares two optimization approaches and proposes a precomputed codebook of reflector settings for different user positions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The codebook is the proposed fix for the passive-eavesdropper problem, yet no experiment selects a codebook entry without ED measurements, so the measured secrecy gains may not transfer to the motivating scenario.","rationale":"The reader's weakest-assumption analysis and my stress-test converge on the same gap: the secrecy gains are measured with online ED feedback, while the codebook that is supposed to handle the passive-ED case is never validated. The central empirical claim — that varactor RIS configurations found with ED feedback can flip the LU/ED power ordering relative to a uniform surface — is supported by Table II and Fig. 3 in the tested indoor geometry, and the uniform-surfcase baseline showing ED power above LU power makes the comparison meaningful. However, that claim is restricted to a scenario where the ED's received power is observable during optimization, which contradicts the passive-ED assumption in Section II. The codebook is the proposed remedy, but the paper reports no experiment where a codebook entry is selected without ED measurements and then evaluated against an unseen ED position. The Section III text itself flags sidelobe risks and downgrades the codebook's role to excluding critical regions, further weakening the 'arbitrary LU independent of ED' claim. This is not an internal contradiction in the field measurements, but it is a failure to validate the paper's stated novel contribution. The existing evidence is honest incremental progress; requiring a codebook validation experiment fits a CONDITIONAL verdict. I therefore recommend no change to the reader's verdict, while emphasizing that the codebook experiment is the decisive missing piece.","tokens_in":11563,"tokens_out":2463,"duration_ms":31219,"concrete_test":"Run a codebook validation experiment: (i) generate the codebook from the sector-pair grid in Fig. 2 as described; (ii) for a held-out LU sector, select a codebook entry using only the LU position (nearest sector, averaged over ED sectors or chosen by a predefined rule), without any ED measurement; (iii) place the ED at a random sector not used in codebook generation and measure LU/ED received powers and SSE for that entry; (iv) compare against the uniform surface and against online Algorithm 1 on the same positions. If the codebook-selected entry yields LU power above ED power and positive SSE for most trials, the passive-ED claim holds; if ED power frequently exceeds LU power, the codebook methodology's load-bearing claim is refuted.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's motivating threat model is a passive eavesdropper whose CSI is unknown (Section II). Algorithm 1 and Algorithm 2 both begin with 'Capture data for both the LU and ED' and require the ED's received power at every column/row flip to compute the LU/ED power ratio. Table II and Fig. 3 therefore demonstrate that online optimizers with ED feedback outperform a uniform surface; they do not demonstrate secrecy against an unobserved ED. The Section III codebook is explicitly introduced to remove this requirement, but Section IV contains no codebook evaluation: no trial selects a configuration using only LU position, no ED is placed at an off-grid or held-out location, and no SSE is reported for codebook entries. The paper itself concedes in Section III that sidelobe effects cause power at non-target locations to vary significantly and even exceed the targeted LU, and that this can only be resolved by high-resolution angle measurement or higher phase resolution; it then weakens the claim to 'exclude potential critical regions.' Thus the central codebook claim — that the methodology provides good RIS configurations for arbitrary LU positions independent of ED location — is unsupported by the reported measurements. The core narrowband comparison is plausible and internally consistent, but it validates an active/cooperative-ED scenario, not the passive-ED scenario the codebook is meant to serve.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an indoor measurement campaign at 3.55 GHz in which a varactor-based RIS is configured by two online search algorithms (Algorithm 1, using the whole surface, and Algorithm 2, partitioning the surface) to maximize a secrecy spectral efficiency metric between a legitimate user and an eavesdropper. The authors compare these configurations against LU-only beamforming, ED-only nulling, and a uniform surface, and they report received powers and SSE values for nine LU/ED angular pairs. The paper also proposes a codebook-generation methodology intended to provide RIS configurations for arbitrary LU positions without depending on the ED location, and it documents frequency-selectivity effects observed with a 5G PRS wideband signal.","tokens_in":11716,"tokens_out":8351,"duration_ms":80869,"significance":"The measurement data are a useful addition to the still-scarce experimental literature on RIS-aided physical-layer security; the hardware description is concrete, the comparison across optimization objectives is informative, and the observed frequency selectivity of the 1-bit varactor surface is a practically relevant caution. However, the experimental validation covers only the active/cooperative-eavesdropper case in which the RIS controller can measure the ED's received power. The central codebook claim—that good configurations can be selected for arbitrary LU positions independent of the ED—is asserted but not tested, and the reported gains are not accompanied by any uncertainty quantification. These gaps prevent the paper from fully supporting its advertised contributions in its current form.","major_comments":[{"comment":"The claim that both Algorithms 1 and 2 achieve an LU received power greater than the ED's \"for all positions\" is directly contradicted by Table II. In the row for LU 0° and ED 30°, Algorithm 2 reports LU = −68.15 dB and ED = −62.42 dB, i.e., the ED receives about 5.7 dB more power than the LU. This is a concrete counterexample to the paper's main empirical claim about Algorithm 2, and it also means the corresponding SSE value under the paper's metric would be negative. The authors must correct either the claim or the data.","section":"Section IV, Table II"},{"comment":"The codebook methodology is never evaluated. Section III describes dividing the indoor area into 15° sectors and storing configurations obtained by running Algorithms 1 and 2 for each LU/ED pair, but Section IV contains no experiment in which a codebook entry is selected using only the LU position, no ED placed at an off-grid or held-out location, and no SSE is reported for codebook entries. Consequently, the statement \"for any relative LU position with respect to the RIS, there exists a good RIS configuration independently of the ED's location\" is unsupported by the presented measurements. The paper should either add such a validation experiment or substantially weaken the claim to the level of a design proposal.","section":"Sections III and IV"},{"comment":"The measured secrecy gains require ED feedback and therefore do not apply to the passive-eavesdropper threat model that motivates the paper. Algorithm 1 (lines 1, 7–8) and Algorithm 2 (lines 1–3, 7–13, 23–24) both require the RIS controller to capture data and measure received power at the ED, and Table II reports exactly those online-optimized configurations. For a passive ED whose CSI is unknown—the scenario stated in Section I as the main challenge—these algorithms cannot be executed. Section III's codebook is proposed to remove this dependence, but as noted above it is not validated. The paper should clearly separate what is demonstrated (optimization with an observable ED) from what is conjectured (codebook operation without ED observations).","section":"Section II, Algorithms 1 and 2"},{"comment":"Reported values lack repeated trials or uncertainty quantification. Every entry in Table II appears to be a single measurement, and several conclusions rest on margins of a few dB (e.g., LU 0°/ED 30°, Algorithm 1: −61.76 dB at the LU vs. −64.40 dB at the ED). Indoor multipath environments exhibit variability that could easily change such conclusions, so the claim that Algorithms 1 and 2 improve LU-over-ED power \"for all positions\" needs either repeated measurements with error bars or at least a clear statement that these are single-trial results with correspondingly limited statistical strength. The evaluation is also in-sample: configurations are optimized and evaluated at the same positions in the same static geometry, so no generalization to unseen positions or channel realizations is demonstrated.","section":"Table II and Fig. 3"}],"minor_comments":[{"comment":"The word \"proviging\" in the abstract should be \"providing\".","section":"Abstract"},{"comment":"The phrase \"15° epicenter angle each\" should be \"15° angular width per sector\" or \"15° apex angle\", since the sectors are defined at the RIS center.","section":"Section III"},{"comment":"The sentence containing \"only 12 μ = 12 2 =6\" is garbled; it should read something like \"only 12/μ = 6 resource blocks were occupied\".","section":"Section II-A.2"},{"comment":"There is a missing space in \"inCB\" in the sentence about measuring the power pattern of each computed RIS configuration; it should read \"in CB\".","section":"Section III"},{"comment":"The text says \"Table II includes the SSE results\", but Table II actually reports received powers in dB; the SSE values are displayed in Fig. 3. This discrepancy should be fixed to avoid confusing readers about what each artifact contains.","section":"Section IV and Table II"}],"recommendation":"major_revision","confidential_remarks":"The paper would be strengthened by releasing the measurement data and by tightening the distinction between the online-optimization demonstration (which requires ED feedback) and the codebook proposal (which is intended for passive EDs). The row-level contradiction in Table II undercuts a headline claim, and the codebook validation gap is substantial; both are fixable with additional experiments, but the paper as submitted is not yet ready for acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take: this is a solid incremental measurement paper, and the main thing you should know is that the advertised codebook contribution is not experimentally supported. What the paper actually delivers is a clean FR1 (3.55 GHz) indoor measurement campaign with a varactor-based RIS, comparing two greedy secrecy-optimization algorithms against rate-based baselines and a uniform surface. That is genuinely useful: Table II and Fig. 3 show that for all tested LU/ED angular pairs, both Algorithm 1 (whole-surface ratio maximization) and Algorithm 2 (partitioned, from prior work) give higher received power at the LU than at the ED, while the uniform surface often gives the ED more power. The frequency-selectivity observation from the 5G PRS test is also a worthwhile practical data point. The hardware description and measurement setup are transparent, and the comparison logic is internally coherent.\n\nThe soft spots are real but not fatal. First, there are no repeated trials or error bars, so we have no sense of measurement variability; a few dB jitter could matter for some conclusions. Second—and this is the bigger issue—Algorithms 1 and 2 both start by capturing received power at the ED. That means the reported secrecy gains are for an active or cooperating eavesdropper whose measurements are available to the optimizer, not for the passive eavesdropper named in Section II. The codebook is explicitly the proposed remedy, but Section IV contains no evaluation of it: no held-out ED positions, no selection using only LU position, no SSE numbers for codebook entries. The paper itself concedes in Section III that sidelobe effects can make power at non-target locations vary significantly and even exceed the targeted LU, and then weakens the claim to 'excluding potential critical regions.' The stress-test note has this right. Third, the configurations are selected from the same measurements used to compute the reported SSE, so the improvement is partly self-validating; that is common in this kind of greedy experimental work but should be acknowledged. I also note that the SSE is computed without the [x]+ operator; the authors explain why, and they report raw powers, so this is a minor transparency choice rather than a flaw.\n\nThe citation pattern looks fine, and the reliance on [35] as the algorithm source is properly attributed. The paper is honest, incremental progress, not a breakthrough.\n\nRecommendation: deserves serious peer review. A good referee should push for repeated trials and for either a real codebook validation (held-out LU/ED positions) or a clear statement that the codebook is a proposal only. With that, the measurement core is credible.","headline":"Useful FR1 RIS-PLS measurement campaign; the codebook contribution is the weak link because it is never experimentally validated.","tokens_in":12342,"tokens_out":2511,"would_cite":true,"duration_ms":26853,"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":"Measured at 3.55 GHz, a 1-bit varactor RIS configured by either of two secrecy-maximizing algorithms keeps the legitimate user's power above the eavesdropper's at every tested angle.","keywords":["Reconfigurable intelligent surface","physical-layer security","secrecy spectral efficiency","indoor measurements","codebook generation","frequency selectivity","OFDM","1-bit RIS"],"falsifier":"Run the codebook selection with no eavesdropper feedback: place the ED at one of the positions in Fig. 2, pick the codebook entry using only the LU's sector, and compare measured LU and ED powers; if any tested sector yields ED power above LU power under the selected entry, the claim that the codebook provides an ED-independent good configuration fails.","tokens_in":11302,"feed_emoji":"📡","tokens_out":12034,"duration_ms":108594,"temperature":0.7,"pith_summary":"The paper uses real indoor measurements at 3.55 GHz to ask whether a varactor-based, 1-bit reconfigurable intelligent surface can provide physical-layer security. It compares two secrecy-spectral-efficiency maximization algorithms—one that steers the whole surface to maximize the legitimate-to-eavesdropper received-power ratio, and one that splits the surface into a legitimate-user-boosting half and an eavesdropper-suppressing half—against simple beamforming and a uniform surface. The measured claim is that both secrecy algorithms deliver higher received power at the legitimate user than at the eavesdropper for every tested angular pair, while the uniform surface often delivers more power to the eavesdropper. The paper also proposes a codebook-generation methodology: precomputing RIS configurations for all pairs of 15-degree angular sectors so that a good configuration exists for any legitimate-user location independent of the eavesdropper location. A wideband 5G-style PRS test shows the RIS's frequency-selective response weakens the legitimate-to-eavesdropper power separation compared with narrowband transmission.","feed_headline":"Smart surface keeps user signal above eavesdropper in all tests","feed_subtitle":"Measured at 3.55 GHz: both secrecy-maximizing algorithm configurations keep the legitimate user ahead at every tested angle.","key_machinery":"The load-bearing objects are the two optimization algorithms plus the codebook they populate. Algorithm 1 flips each RIS column and row in turn and keeps the flip only if the ratio of the LU's received power to the ED's received power improves. Algorithm 2 does the same on two halves of the panel, maximizing LU power on one half and minimizing ED power on the other. Both operate on a 1-bit phase model: each varactor element switches between two states with approximately unit amplitude and a phase shift in $[0,\\pi]$. The codebook $\\mathcal{CB}$ is the resulting table of configurations indexed by the 15-degree angular sector of the LU and ED, intended to supply a good configuration for any LU position without needing the ED's channel state at run time. The objective they all serve is the sum secrecy spectral efficiency $R_{\\mathrm{sec}}=\\sum_v[\\log_2(1+\\mathrm{SNR}_l[v])-\\log_2(1+\\mathrm{SNR}_e[v])]^+$.","core_discovery":"On the paper's own terms, the central discovery is that a 1-bit varactor RIS, tuned by over-the-air measurements, can invert the power balance in an indoor wiretap setup: for all tested angular pairs, the two SSE-maximizing algorithms leave the legitimate user with higher received power than the eavesdropper, whereas a uniform RIS surface frequently gives the eavesdropper more power than the legitimate user. Algorithm 1, which sweeps column and row phase flips over the entire panel to maximize the LU/ED power ratio, performs slightly better than Algorithm 2, which dedicates half the panel to boosting the LU and half to suppressing the ED. The authors further claim that their codebook methodology, built by storing the optimized configurations for every 15-degree sector pair, yields a good RIS configuration for any LU position independently of the ED position, and that the frequency-selective behavior of the RIS is the main obstacle to wideband operation.","pith_inferences":["An extension the paper leaves open is that a truly passive eavesdropper cannot supply the received-power readings Algorithms 1 and 2 require, so the codebook needs a selection rule using no ED measurements and an experimental check that such a rule preserves the secrecy gains.","The codebook could be made robust to the side-lobe effects the paper describes by storing fine-grained angular power patterns for each entry and choosing the entry whose pattern has the deepest nulls in the critical regions.","Because the directive transmit antenna blocked the direct AP-to-user paths, the measured gains are for a purely RIS-reflected link; realistic deployments with a strong direct path would dilute the LU/ED power separation and need separate validation."],"forward_implications":["If the measured power separation holds, a 1-bit RIS with the whole-surface ratio-maximizing algorithm can provide positive secrecy spectral efficiency in indoor multipath environments without relying on higher-layer encryption.","The slight superiority of Algorithm 1 over Algorithm 2 indicates that treating the panel as one coupled aperture, rather than two independent halves, preserves the cross-terms that matter for secrecy.","A codebook precomputed over angular sectors means that, whenever the eavesdropper's location is observable (e.g., an active but untrusted user), the RIS can be reconfigured by a lookup instead of an online optimization.","The degraded wideband PRS results imply that codebook entries tuned at a single carrier frequency may need to be validated across the full 37.44 MHz bandwidth before deployment."],"supporting_citations":[{"why":"Supplies the earlier indoor RIS-aided PLS measurement campaign at 5.2 GHz and the partitioned-RIS Algorithm 2 approach used as the comparison baseline.","marker":"[35]"},{"why":"Defines the varactor-based dual-polarized metasurface hardware whose 1-bit, near-unit-amplitude phase response the experiments use.","marker":"[36]"},{"why":"Foundational wiretap-channel result that establishes the secrecy-capacity concept behind the SSE objective.","marker":"[8]"},{"why":"Characterizes the secrecy capacity of the Gaussian wiretap channel, grounding the rate-difference form of the SSE metric.","marker":"[10]"},{"why":"Provides the 5G PRS/OFDM waveform structure used for the wideband frequency-selectivity measurements.","marker":"[37]"},{"why":"Documents beam-sweeping field trials with the same RIS prototype, supporting the frequency-selective reflection and side-lobe explanations for weak wideband separation.","marker":"[38]"},{"why":"Characterizes real-hardware phase-resolution and near-field effects, supporting the discussion of side-lobe behavior of 1-bit configurations.","marker":"[39]"},{"why":"A prior measurement-based PLS characterization campaign that, together with [35], supports the equal-noise-power assumption for the indoor SSE calculation.","marker":"[40]"}],"fun_headline_variants":["RIS codebook flips indoor wiretap power balance","Measured RIS secrecy: user beats eavesdropper at all angles","Over-the-air RIS tuning secures indoor links","Inverting power balance: indoor RIS security measured","RIS phase codebook keeps user signal ahead of eavesdropper"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the eavesdropper's received power is observable during configuration: both algorithms start by capturing data for the LU and the ED, so the measured secrecy gains hold only when the ED's channel can be measured, while the passive-eavesdropper scenario the paper motivates cannot supply that measurement and the codebook path is not validated end-to-end without it.","fun_headline_variants_meta":{"raw":{"variants":["RIS codebook flips indoor wiretap power balance","Measured RIS secrecy: user beats eavesdropper at all angles","Over-the-air RIS tuning secures indoor links","Inverting power balance: indoor RIS security measured","RIS phase codebook keeps user signal ahead of eavesdropper"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000268,"raw_usage":{"total_tokens":1636,"prompt_tokens":979,"completion_tokens":657,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":595,"completion_tokens_details":{"reasoning_tokens":578}},"tokens_in":595,"tokens_out":657,"duration_ms":7135,"temperature":1.0,"reasoning_tokens":578,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:10:55.150570+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the codebook selection with no eavesdropper feedback: place the ED at one of the positions in Fig. 2, pick the codebook entry using only the LU's sector, and compare measured LU and ED powers; if any tested sector yields ED power above LU power under the selected entry, the claim that the codebook provides an ED-independent good configuration fails.","supporting_citations":[{"cited_title":"Indoor measurements for RIS-aided communication: Practical phase shift optimization, coverage enhancement, and physical layer security,","cited_arxiv_id":null,"evidence_quote":"Supplies the earlier indoor RIS-aided PLS measurement campaign at 5.2 GHz and the partitioned-RIS Algorithm 2 approach used as the comparison baseline."},{"cited_title":"Fully-addressable varactor-based reflecting metasurface with dual- linear polarisation for low power reconfigurable intelligent surfaces,","cited_arxiv_id":null,"evidence_quote":"Defines the varactor-based dual-polarized metasurface hardware whose 1-bit, near-unit-amplitude phase response the experiments use."},{"cited_title":"The wire-tap channel,","cited_arxiv_id":null,"evidence_quote":"Foundational wiretap-channel result that establishes the secrecy-capacity concept behind the SSE objective."},{"cited_title":"The Gaussian wire-tap chan- nel,","cited_arxiv_id":null,"evidence_quote":"Characterizes the secrecy capacity of the Gaussian wiretap channel, grounding the rate-difference form of the SSE metric."},{"cited_title":"Receiver design and time of arrival estimation for opportunistic localization with 5G signals,","cited_arxiv_id":null,"evidence_quote":"Provides the 5G PRS/OFDM waveform structure used for the wideband frequency-selectivity measurements."},{"cited_title":"Evaluating beam sweeping for AoA estimation with an RIS prototype: Indoor/outdoor field trials,","cited_arxiv_id":null,"evidence_quote":"Documents beam-sweeping field trials with the same RIS prototype, supporting the frequency-selective reflection and side-lobe explanations for weak wideband separation."},{"cited_title":"Performance of RIS-aided nearfield localization under beams approximation from real hardware characterization,","cited_arxiv_id":null,"evidence_quote":"Characterizes real-hardware phase-resolution and near-field effects, supporting the discussion of side-lobe behavior of 1-bit configurations."},{"cited_title":"Measurement-based characterization of physical layer security for RIS-assisted wireless systems,","cited_arxiv_id":null,"evidence_quote":"A prior measurement-based PLS characterization campaign that, together with [35], supports the equal-noise-power assumption for the indoor SSE calculation."}],"review_version":1}