{"id":"c8f6678f-ea1f-4bef-8804-722cdcea8337","arxiv_id":"2505.04249","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Underwater magnetic induction communication can be eavesdropped when a nearby coil is aligned with the transmitter, and small voltage changes at the legitimate receiver may hint at the eavesdropper's presence.","lead":"This paper tests whether the magnetic fields used in underwater magnetic induction communication can be secretly overheard by an eavesdropper with its own coil. Simulations and lab experiments show eavesdropping is possible, but it depends heavily on the eavesdropper coil's position and angle.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Quantitative underwater claims are not tested at seawater conductivity; air experiments validate only the low-conductivity model.","rationale":"The reader's weakest assumption identifies the same load-bearing concern: air experiments and 0.01 S/m simulations stand in for seawater. I looked for an even more direct internal gap, such as the unspecified noise power σ² in the secrecy-capacity expression, but the conductivity mismatch is the more central issue because it touches every voltage and secrecy-capacity value in the paper, not just the metric definition. The concern is explicit in the manuscript: Section III-C states that experiments were done in air because no underwater setup was available. That admission supports the critique rather than undermining it. The central qualitative finding, that an eavesdropper coil can couple to the MI link when aligned and nearby, is well supported by the cosθ null structure and by the consistency between FEM and the air experiments; I do not see a reason to reject it. The practical gap is quantitative: the reported curves are for low-conductivity water, and seawater attenuation at 100 kHz is strong enough over the tested distances to shift voltage levels and secrecy-capacity values substantially. Because the reader's conditional verdict already requires the authors to address exactly this before the quantitative underwater claims are accepted, no change to the verdict is needed.","tokens_in":12676,"tokens_out":5655,"duration_ms":59880,"concrete_test":"Re-run the FEM models of Section III-B and the secrecy-capacity setup with σ=4 S/m and ε_r≈81, keeping all other Table I parameters and the same geometries. Compare V_E/V_Rx and the secrecy capacity of Fig. 11 at each reported position and orientation. As an experimental cross-check, repeat the Configuration 1-2 position sweeps and Configuration 3-5 orientation sweeps in a water tank with 3.5% salinity; if the secrecy-capacity ordering or sign changes at any reported point, the quantitative conclusion must be restricted to fresh or low-conductivity water and shorter ranges, while the qualitative vulnerability claim can remain.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's own Section III-C admits that all real-world experiments were conducted in air, and Section III-B fixes water conductivity at 0.01 S/m. At the 100 kHz operating frequency, seawater has σ≈4 S/m, giving a skin depth of δ≈0.8 m in Eq. (2), whereas the simulated conductivity gives δ≈16 m. Over the distances used (Tx-Rx up to 4 ft; eavesdropper up to 7.5 ft from Tx), the eddy-current exponential factor in seawater is roughly 0.2 to 0.05, compared with about 0.9 in the simulated low-conductivity water. This materially changes the reported received voltages and therefore the secrecy-capacity values in Fig. 11. The angular nulls at cosθ=0 are geometric and should survive in seawater, so the qualitative statement that a nearby aligned coil can eavesdrop is credible. However, the paper's quantitative underwater vulnerability assessment is not validated for the intended medium: the air experiments agree with the FEM model because both are effectively non-conducting, not because either represents seawater. The limitation is acknowledged inside the paper but the conclusion states the underwater claim without this qualification.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes passive eavesdropping attacks on underwater magnetic induction (MI) communication. It considers a three-node scenario with a legitimate transmitter, a legitimate receiver, and an eavesdropping node, and develops finite-element-method (FEM) simulations for five configurations that vary the eavesdropper's position and orientation relative to the legitimate nodes. The performance metrics are induced voltage, magnetic flux density, and secrecy capacity. Laboratory experiments performed in air are used to replicate the simulation configurations. The authors conclude that underwater MI communication is vulnerable to eavesdropping, that the vulnerability depends strongly on the eavesdropper's position and orientation, and that legitimate nodes may detect malicious activity through changes in the received voltage.","tokens_in":12868,"tokens_out":5318,"duration_ms":54246,"significance":"If the claims hold, this work makes a useful contribution by challenging the common assumption that MI communication is inherently secure or covert underwater, and by providing a systematic map of the geometric conditions under which eavesdropping is most effective. The FEM simulations and laboratory experiments are mutually independent checks against standard coil-coupling physics, and the paper does not fit free parameters to its conclusions. The qualitative agreement between simulation and experiment across five configurations is a genuine strength. However, the quantitative transfer of the results to real seawater is not established: the FEM model uses a freshwater conductivity and the experiments are all in air. The qualitative conclusion that an aligned nearby coil can intercept the signal is credible, since the angular nulls are geometric in origin, but the voltage magnitudes and secrecy-capacity values in the paper should not be read as validated for seawater without further work.","major_comments":[{"comment":"The quantitative underwater claims, including the received-voltage magnitudes and the secrecy-capacity values in Fig. 11, are not supported for the intended seawater medium. The FEM model fixes water conductivity at 0.01 S/m (Table I), and the real-world experiments in Section III-C are conducted in air, as the authors acknowledge. At the 100 kHz operating frequency, seawater has a conductivity of about 4 S/m, giving a skin depth of roughly 0.8 m in Eq. (2), while the simulated conductivity gives a skin depth of about 16 m. Over the 4 ft to 7.5 ft distances used, the eddy-current exponential factor in Eq. (2) falls to roughly 0.2–0.05 in seawater, compared with about 0.9 in the low-conductivity simulation. This materially changes the reported voltage levels and hence the secrecy-capacity numbers. I request either a seawater-conductivity simulation (ideally a conductivity sweep from 0.01 to 5 S/m), or a saline-tank experimental validation, or an explicit and prominent statement that the quantitative results hold only for low-conductivity media. The conclusion should then be qualified accordingly. The qualitative dependence on position and orientation, including the angular null at orthogonal alignment, is likely to survive in seawater, but the quantitative vulnerability assessment is not yet validated.","section":"§III-C, Figs. 13–14"},{"comment":"The experimental section reports single measurements without error bars, repeated trials, or any quantitative agreement metric between the FEM and experimental results. The text states that the trends are 'almost similar' but does not quantify the deviations. Since these experiments are the only direct validation of the FEM model, the lack of uncertainty quantification makes it difficult to assess whether the observed discrepancies affect the paper's conclusions, especially the quantitative claims in Fig. 11. Please add repeated measurements with error bars for each configuration and a stated agreement criterion, such as normalized root-mean-square error or a tolerance band.","section":"§III-B6, Eq. (5), Fig. 11"},{"comment":"The secrecy-capacity evaluation does not specify the noise power sigma^2 used in Eq. (5), the load resistance R, or whether the same noise variance is assumed at the legitimate receiver and the eavesdropper. These parameters directly determine the secrecy-capacity values plotted in Fig. 11, which are a central quantitative output of the paper. The authors should state and justify these parameter choices, and preferably report sensitivity of the secrecy-capacity curves to the assumed noise power.","section":"§III-B6, Eq. (5), Fig. 11"}],"minor_comments":[{"comment":"Equation (4) contains a typographical error: the magnetic permeability constant should be 4π × 10^-7 H/m, not 4π × 10^7 H/m.","section":"Eq. (4)"},{"comment":"The phrase 'secret capacity' in Section III-A should be 'secrecy capacity' for consistency with the rest of the paper.","section":"§III-A"},{"comment":"There is a typo in Section III-B1: 'eavsdropper' should be 'eavesdropper'.","section":"§III-B1"},{"comment":"In Configuration 4, the description labels d_Tx-E = 6 ft, but the geometry rotates the eavesdropper around the legitimate receiver, so the transmitter-to-eavesdropper distance changes as a function of θ_Rx-E. The text and figures should clarify this angle-dependent distance.","section":"§III-B4"},{"comment":"The FEM implementation details, including the software used, mesh settings, and boundary conditions, are not reported. Adding these details would improve reproducibility, and for a security paper that proposes a new vulnerability assessment the lack of available code or data is a limitation worth acknowledging.","section":"§III-C"}],"recommendation":"major_revision","confidential_remarks":"The core qualitative finding — that MI links are not inherently secure against a nearby aligned eavesdropper — is plausible and consistent with standard coil-coupling physics, so a rejection is not warranted. The main obstacle to acceptance is the gap between the claimed underwater quantitative assessment and the actual low-conductivity/air validation. This gap is fixable through a seawater-conductivity simulation study or an explicit re-scoping of the claims. I would also gently note that the novelty is moderate: the paper applies known MI attenuation physics to an eavesdropping scenario, but this application is still valuable for the security community."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper's central qualitative finding—that a nearby aligned coil can eavesdrop on an underwater MI link and that position and orientation dominate the leakage—is credible and well supported by the FEM/experiment agreement. What's new is the quantitative parameter sweep: five configurations covering eavesdropper position and orientation, plus secrecy capacity curves. The physics is standard coil coupling, but no one had done this systematic sweep for underwater MI, and the paper does it honestly.\n\nCredit where due: the FEM and lab results agree qualitatively across all five setups, the 90-degree null is robust, and the secrecy capacity trend (further eavesdropper yields higher secrecy) is physically sensible. The authors are transparent about the experimental limitation: Section III-C states the experiments were run in air, and Table I fixes water conductivity at 0.01 S/m. No constants are fitted to conclusions; the model is standard.\n\nThe soft spot is the quantitative transfer to real seawater. At 100 kHz, seawater (σ≈4 S/m) has skin depth δ≈0.8 m, while the simulated conductivity gives δ≈16 m. Over the distances used—up to 7.5 ft between Tx and eavesdropper—the eddy-current exponential in Eq. (2) is roughly 0.2 to 0.05 in seawater, versus about 0.9 in the simulated water. That changes received voltage levels and the secrecy capacity numbers in Fig. 11. The angular nulls at cosθ=0 are geometric and should survive in seawater, so the qualitative claim holds; but the abstract and conclusion state that underwater MI is vulnerable without this qualification. The stress-test note is correct on this point.\n\nMinor issues: no error bars or repeated trials in the experiments; the noise model for secrecy capacity is unspecified (σ² is never defined); and the detection mechanism—legitimate Rx noticing voltage changes—is plausible but speculative, since the changes appear small and may be within measurement uncertainty. None of these undermine the core qualitative claim.\n\nThe citation pattern is normal; self-citations are background, not load-bearing. This is a solid empirical study for the underwater MI and physical-layer security community. It deserves peer review, but a referee should require either saltwater experiments or a careful scaling argument before the quantitative ocean claims are accepted.\n\nRecommendation: send to peer review; the paper can be accepted as a qualitative vulnerability study with major revisions on the seawater question.","headline":"Qualitative result holds—MI eavesdropping depends on geometry—but the seawater numbers are not validated because experiments were done in air and simulations used freshwater conductivity.","tokens_in":13387,"tokens_out":2544,"would_cite":true,"duration_ms":25083,"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":"The paper claims that underwater magnetic induction links, usually considered secure, can be eavesdropped, and that the risk depends on the eavesdropper's position and orientation.","keywords":["underwater magnetic induction","eavesdropping attack","secrecy capacity","coil orientation","finite element method","induced voltage","underwater wireless security","magnetic coupling"],"falsifier":"Place the same three coils in a seawater tank at $\\sigma \\approx 4$ S/m, repeat Configurations 1–5, and compare the eavesdropper and receiver voltages; if the eavesdropper voltage at the claimed working distances falls below the receiver noise floor, or if the receiver-voltage changes disappear, the paper's quantitative vulnerability assessment does not transfer to real oceans.","tokens_in":12486,"feed_emoji":"📡","tokens_out":5973,"duration_ms":57962,"temperature":0.7,"pith_summary":"Underwater magnetic induction (MI) communication is often treated as secure or covert because its magnetic fields are neither audible nor visible, unlike acoustic and optical signals. This paper argues that the assumption is unsafe: a nearby eavesdropper coil can pick up the same field the legitimate receiver uses, and the amount it learns is governed by where the eavesdropper sits and how its coil is angled. The claim is supported by finite-element simulations and by lab-bench experiments with resonant coils, which match in their main trends. The paper also reports that the legitimate receiver's induced voltage changes detectably in some eavesdropper configurations, suggesting that malicious activity could be sensed rather than merely endured. If the claim holds, secrecy in MI communication is a geometric property of coil placement, not an intrinsic property of the technology.","feed_headline":"Underwater magnetic links are not inherently secret","feed_subtitle":"Coil position and angle decide how much an eavesdropper receives, and the receiver may sense the intruder.","key_machinery":"The central object is the resonant coil pair and the angular dependence of magnetic flux through a coil. The paper uses the standard induced-voltage relation $V_{\\mathrm{ind}} = 2\\pi N A B \\cos\\theta$ with the flux-density expression $B = \\mu_0 N I r^2 \\cos\\theta / (2(r^2+d^2)^{3/2})$ and an exponential skin-depth attenuation term for conducting media, so every result is ultimately a function of coil offset angle $\\theta$ and distance $d$. Finite-element models of five eavesdropper configurations compute the magnetic flux density and the voltage induced in the legitimate receiver and the eavesdropper; those voltages feed the secrecy-capacity expression $\\mathrm{SC} = \\log_2(1+\\mathrm{SNR}_{\\mathrm{Rx}}) - \\log_2(1+\\mathrm{SNR}_E)$. The 90° null in $\\cos\\theta$ is what produces the paper's cleanest result: an orthogonally oriented eavesdropper receives no information, while an aligned one can.","core_discovery":"The paper's central claim is that underwater MI communication is genuinely vulnerable to passive eavesdropping, but the vulnerability is conditional, not absolute. An eavesdropper can extract information whenever its coil is close to the legitimate transmitter and oriented so that magnetic flux threads its coil; maximum received voltage occurs when the eavesdropper faces the transmitter, and reception drops to zero at a 90° orientation, where field lines are parallel to the coil plane. The finite-element simulations and the air-medium lab experiments both show these trends across five configurations (two position-based, three orientation-based). The paper further computes secrecy capacity from the received voltages and finds it falls as the eavesdropper approaches the transmitter or aligns with it, and rises when the eavesdropper is far or orthogonal. A second claimed finding is that the legitimate receiver's voltage changes in several of these configurations, which the paper presents as a possible basis for detecting malicious nodes.","pith_inferences":["In real seawater ($\\sigma \\approx 4$ S/m), the exponential attenuation term in the paper's Eq. (2) is far stronger than in the 0.01 S/m simulations, so the absolute voltages and secrecy-capacity values should be read as optimistic; the geometric null at 90° should persist, but the eavesdropper's usable range is likely shorter.","The observed perturbation of the legitimate receiver's voltage could be developed into an active authentication scheme: the transmitter can vary its load or current in a known pattern, and the receiver can check whether the received-voltage signature matches, flagging any coil that disturbs it.","A single-eavesdropper model may understate the threat: multiple malicious coils or a relay coil could detune or retune the field in ways that enlarge the eavesdropping region, and the paper does not model that case.","The experiment was run at 100 kHz in air; at the lower frequencies typical of long-range underwater MI, skin depth grows, so the quantitative gap between air and seawater may narrow, which is a testable extension."],"forward_implications":["A deployed MI link should treat the physical geometry of nearby coils as a security parameter: an aligned, close eavesdropper can read the traffic, so coil placement and orientation need active management.","Legitimate receivers that can measure small induced-voltage changes have a built-in intrusion signal in several configurations, which could trigger countermeasures such as jamming or link re-orientation.","Secrecy capacity predictions give operators a rule of thumb: keep the eavesdropper farther from the transmitter than the legitimate receiver, and avoid alignment between Tx and the eavesdropper's likely positions.","The angular null at 90° means there are orientations in which an eavesdropper receives nothing; directional coil designs or deliberate misalignment could be used as a physical-layer hardening measure."],"supporting_citations":[{"why":"Supplies the coil magnetic-field expression used for the legitimate link and the eavesdropper's received field.","marker":"[25]"},{"why":"Adds the skin-depth attenuation term for conducting media, the basis for extending air results to water.","marker":"[26]"},{"why":"Provides the induced-voltage relation V_ind = 2πNAB cosθ used to compute received voltage at Rx and eavesdropper.","marker":"[27]"},{"why":"Gives the mutual-inductance model and is cited to justify similar MI behavior across air and water.","marker":"[2]"},{"why":"Establishes underwater MI channel fundamentals and supports the claim of similar performance across media.","marker":"[29]"},{"why":"Supports the assumption that MI behaves similarly in different environments, used to justify air experiments.","marker":"[30]"}],"fun_headline_variants":["Underwater magnetic links leak when coils align","Eavesdropper's coil angle decides underwater secrecy","Coil orientation makes or breaks underwater MI secrecy","Receiver may detect nearby malicious coils in MI links","Underwater MI eavesdropping hinges on coil facing"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantitative results assume that air-medium lab tests and simulations at 0.01 S/m water conductivity faithfully represent real seawater, whose conductivity is roughly 4 S/m and whose eddy-current losses are much stronger.","fun_headline_variants_meta":{"raw":{"variants":["Underwater magnetic links leak when coils align","Eavesdropper's coil angle decides underwater secrecy","Coil orientation makes or breaks underwater MI secrecy","Receiver may detect nearby malicious coils in MI links","Underwater MI eavesdropping hinges on coil facing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000245,"raw_usage":{"total_tokens":1544,"prompt_tokens":962,"completion_tokens":582,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":578,"completion_tokens_details":{"reasoning_tokens":510}},"tokens_in":578,"tokens_out":582,"duration_ms":6038,"temperature":1.0,"reasoning_tokens":510,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:32:54.041414+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Place the same three coils in a seawater tank at $\\sigma \\approx 4$ S/m, repeat Configurations 1–5, and compare the eavesdropper and receiver voltages; if the eavesdropper voltage at the claimed working distances falls below the receiver noise floor, or if the receiver-voltage changes disappear, the paper's quantitative vulnerability assessment does not transfer to real oceans.","supporting_citations":[{"cited_title":"Characterization of a 3d underwater magneto-inductive transmitter coil array,","cited_arxiv_id":null,"evidence_quote":"Supplies the coil magnetic-field expression used for the legitimate link and the eavesdropper's received field."},{"cited_title":"Underwater communication employing high- sensitive magnetic field detectors,","cited_arxiv_id":null,"evidence_quote":"Adds the skin-depth attenuation term for conducting media, the basis for extending air results to water."},{"cited_title":"Design and evaluation of low-cost and energy-efficient magneto-inductive sensor nodes for wireless sensor networks,","cited_arxiv_id":null,"evidence_quote":"Provides the induced-voltage relation V_ind = 2πNAB cosθ used to compute received voltage at Rx and eavesdropper."},{"cited_title":"Fun- damentals and advancements of magnetic field communication for underwater wireless sensor networks,","cited_arxiv_id":null,"evidence_quote":"Gives the mutual-inductance model and is cited to justify similar MI behavior across air and water."},{"cited_title":"Magnetic induction for underwater wireless commu- nication networks,","cited_arxiv_id":null,"evidence_quote":"Establishes underwater MI channel fundamentals and supports the claim of similar performance across media."},{"cited_title":"Dy- namic magnetic induction wireless communications for autonomous- underwater-vehicle-assisted underwater IoT,","cited_arxiv_id":null,"evidence_quote":"Supports the assumption that MI behaves similarly in different environments, used to justify air experiments."}],"review_version":1}