PaperlayerView exact revision on arXiv

Source-linked AI summary

Privacy-Aware ISAC for Full-Duplex Monostatic Systems Using Movable Antennas

Yasas Savinda, Mohammadali Mohammadi, Himal A. Suraweera, Henk Wymeersch

arXiv:2609.35534v1eess.SP

TL;DR

Sensing leakage from passive receivers remains relatively underexplored, and prior privacy approaches often use fixed arrays or do not target DD deception. The paper jointly optimizes beamforming and movable-antenna positions to create a dummy DD-bin response while preserving BS true-bin sensing quality. Simulations show reduced Eve true-bin selection and increased dummy-bin selection compared with benchmarks while maintaining BS sensing performance.

  • Problem

    Sensing leakage remains relatively underexplored in ISAC, while prior approaches often assume fixed arrays, target other privacy objectives, or rely on strong Eve assumptions.

    Source: §I · p. 1

  • Method

    The design jointly optimizes beamforming and pulse-indexed movable-antenna states to induce a dummy DD-bin response at Eve while meeting the BS true-bin sensing requirement.

    Source: §I · p. 1, B. Optimization Problem Solution · p. 4

  • Results

    The proposed design shifts Eve’s detector output toward the designed dummy DD bin and suppresses true-bin selection while maintaining BS sensing and downlink QoS requirements.

    Source: §V · p. 6, §VI · p. 6

  • Takeaways & Limitations

    In the simulated FD monostatic MA-ISAC setting, jointly controlling antenna locations and beamforming can mislead Eve toward a designed dummy bin while preserving BS sensing performance.

    Source: §V · p. 6, §VI · p. 6

Abstract

from arXiv · show

This work investigates sensing privacy in full-duplex (FD) monostatic integrated sensing and communication (ISAC) systems with movable antennas (MAs). The proposed approach jointly optimizes beamforming and antenna trajectories to create a deceptive dummy DD-bin response at a passive sensing eavesdropper (Eve), while satisfying a true-bin sensing-quality requirement at the base station (BS). The resulting problem is highly non-convex. {To address this, a stage-wise alternating local-search framework is developed to obtain suboptimal solutions. Within this framework, we maximize the worst-case margin between dummy and true delay-Doppler (DD)-bin detector-oriented SINR surrogates over a discretized uncertainty region for Eve, incorporating detector-aligned dummy-bin refinement and true-bin preservation.} Simulation results show that the proposed MA-enabled design suppresses Eve's true-target DD-bin selection and increases dummy-bin selection probability compared with benchmark schemes, while maintaining reliable BS sensing performance.

I. Introduction

Sensing leakage from passive receivers remains relatively underexplored in ISAC, while prior privacy approaches often use fixed arrays, target other privacy objectives, or assume strong Eve models. This work uses movable-antenna position control alongside beamforming to induce a dummy DD-bin response while preserving the BS true-bin sensing requirement.

I. Introduction

  • Sensing leakage remains relatively underexplored despite passive receivers’ ability to infer sensitive information from ISAC signals.Source: §I · p. 1
  • Prior privacy methods often assume fixed arrays, target communication or localization privacy rather than DD deception, or rely on strong Eve assumptions.MAs add spatial degrees of freedom through dynamic position reconfiguration.Source: §I · p. 1
  • The proposed framework jointly optimizes beamforming and pulse-to-pulse antenna positions to induce dummy DD-bin responses at Eve while satisfying the BS true-bin sensing-quality requirement.Source: §I · p. 1
  • A stage-wise AO heuristic addresses the non-convex ambiguity-function design, and simulations report privacy improvement while maintaining sensing performance.The heuristic uses a four-stage procedure to progressively refine the design.Source: §I · p. 1

II. System Model

The modeled system uses an FD BS with movable transmit and receive arrays to serve downlink and uplink users while sensing a target. Antennas move within planar regions under bounded pulse-to-pulse motion, and Eve’s location is represented by a finite set of candidates.

II. System Model

  • An FD BS with two MA-enabled UPAs simultaneously serves downlink and uplink single-antenna users while sensing one target in the presence of a single-antenna Eve.Source: §II · p. 1
  • Transmit and receive antennas move within separate two-dimensional regions, with pulse-indexed position matrices defining their states over the CPI.Source: §II · p. 2
  • Eve’s uncertainty region is discretized into candidate locations, while antenna motion is bounded pulse-to-pulse over the CPI.Minimum inter-antenna spacing and maximum displacement constraints also apply.Source: §II · p. 2

A. Channel Model

The channel model ties sensing and communication responses to movable-antenna positions, making the steering responses slow-time dependent. The system assumes LoS-dominated propagation and perfect synchronization within each CPI, with Eve’s exact location unknown.

A. Channel Model

  • Under LoS sensing, target elevation and azimuth determine steering responses, which become slow-time dependent as antenna positions vary.Source: A. Channel Model · p. 2
  • Communication channels are modeled as functions of the transmit and receive movable-antenna position vectors.Source: A. Channel Model · p. 2
  • The model assumes LoS-dominated geometric propagation and perfect synchronization within each CPI; the BS knows legitimate and modeled residual-SI CSI but only Eve’s uncertainty region.The assumptions are identified as scope boundaries, with relaxation left for future work.Source: A. Channel Model · p. 2

B. Signal Model

The signal model forms matched-filter DD statistics at Eve and the BS from the sensing reference sequence and their observations. It derives detector-oriented guarded SINR surrogates that distinguish true- and dummy-bin energy from local sidelobe and global off-bin interference, while enforcing BS true-bin sensing quality.

B. Signal Model

  • The BS transmits a combination of downlink communication signals and a sensing waveform, then receives target reflections, uplink transmissions, and residual self-interference.Source: B. Signal Model · p. 2
  • A fixed spatial sensing vector and known unit-modulus probing sequence form the scalar reference sequence used by the BS and Eve matched filters.Source: B. Signal Model · p. 2, §III · p. 3
  • Eve’s passive observation includes reflected ISAC signals, and its received phase depends on target–Eve geometry and movable-antenna positions.The model uses a nominal target DD hypothesis, neglects spatial variation in the bistatic DD offset, and assumes direct BS–Eve cancellation before DD processing.Source: B. Signal Model · p. 3
  • The detector-oriented Eve surrogate separates desired true- or dummy-bin energy from weighted local sidelobe leakage and global off-bin background.The guard region excludes the bin and nearest DD neighbors from interference computation, while the ring captures sidelobe leakage.Source: §III · p. 3
  • The guarded DD-bin SINR surrogate is detector-oriented rather than a conventional pre-detection physical SINR.The underlying interference weighting trades off local sidelobe leakage against global interference.Source: §III · p. 3
  • At the BS, the design imposes a true-bin sensing-quality requirement using true-bin energy and analogous weighted local and global interference terms.Source: §III · p. 3

A. Problem Formulation

The design jointly optimizes beamforming and pulse-indexed antenna positions to promote a dummy DD-bin response at Eve while meeting the BS true-bin sensing requirement. Its detector-aware objective accounts for Eve’s bin-selection thresholds and competing off-bin peaks, alongside BS true-bin quality.

Design objective

  • The design jointly optimizes pulse-indexed MA states and transmit beamforming to suppress Eve’s true-bin response, promote a predefined dummy bin, and satisfy the BS true-bin sensing-quality requirement.The formulation also imposes BS guarded DD-bin sensing-surrogate, DL QoS, and transmit-power constraints.Source: A. Problem Formulation · pp. 3–4

Guard regions and BS metrics

  • The dummy-detection off-bin set excludes guard regions around the true and dummy bins, while BS sensing quality also accounts for its strongest off-bin competitor.The BS threshold is based on off-bin background outside the true-bin guard region.Source: A. Problem Formulation · p. 4

Eve detector metrics

  • Eve selects the strongest threshold-exceeding DD bin, so dummy-bin detection is evaluated against off-bin competitors and a threshold based on off-bin mean and standard deviation.The threshold uses a positive weighting factor on the off-bin standard deviation.Source: A. Problem Formulation · p. 4

Detector-aware objective

  • Eve’s detector-aware objective uses the worst-case dummy-versus-true SINR margin, while BS-side terms promote true-bin dominance and threshold excess.The Eve-side terms encourage dummy-bin selection; the BS metrics support sensing-quality control.Source: A. Problem Formulation · p. 4

B. Optimization Problem Solution

The coupled beamforming and antenna-position design is non-convex, so the paper uses a four-stage alternating-optimization heuristic with derivative-free local search. Its stages progressively initialize and refine the design while retaining hard BS-sensing and DL-QoS feasibility in the constrained phases.

Stage-wise AO framework

Alternating search blocks

  • With antenna positions fixed, beamforming uses a reduced beamspace and derivative-free Powell search; with beamforming fixed, position updates use perturbation-based search over feasible MA trajectories.The trajectory search respects region, spacing, and pulse-to-pulse movement constraints.Source: B. Optimization Problem Solution · p. 4

MWS and HC

  • MWS searches multiple feasible initializations with soft BS-sensing rewards and DL-QoS penalties; HC then imposes BS-sensing and DL-QoS requirements as hard constraints.During HC and later hard-feasible stages, updates are accepted only when BS sensing and all DL QoS requirements remain satisfied.Source: 1) Stage-Wise AO Framework · pp. 4–5

EDD and BCP

  • EDD sharpens Eve’s dummy response while preserving HC feasibility, and BCP improves BS true-bin prominence while penalizing strong BS off-bin competitors.BCP retains reference levels for Eve privacy margin, BS dominance, threshold excess, and peak-window metrics.Source: 1) Stage-Wise AO Framework · p. 5

Computational complexity

  • The stated complexity is O(Itot[NPow+4NswL(Ntx+Nrx)] Ceval), with each MA-position sweep testing at most 4L(Ntx+Nrx) perturbations.For fixed antenna and user dimensions, direct DD-grid evaluation gives Ceval = O(NE|Ω|L).Source: 1) Stage-Wise AO Framework · p. 5

V. Numerical Results

The numerical results show that joint beamforming and MA optimization shifts Eve’s DD response toward the designed dummy bin and improves dummy-bin selection over benchmarks. The study evaluates this design across MA aperture ratios using a discretized Eve uncertainty region.

V. Numerical Results

  • The evaluated FD monostatic system uses six-element transmit and receive arrays, two downlink and two uplink users, and 16 pulses per CPI.The carrier is 30 GHz, and 0.2 of total transmit power is allocated to sensing.Source: §V · p. 5
  • The proof-of-concept robust design discretizes Eve’s uncertainty region into 16 locations; denser and off-grid validation is left for future work.The locations form a 4 × 4 grid over x ∈ [12, 16] m and y ∈ [−4, 4] m.Source: §V · p. 6
  • The aperture sweep keeps total transmit-plus-receive area at 0.08 m² while varying ρ, the receive-to-transmit area ratio.Source: §V · p. 6
  • Pulse-indexed MA positions represent switched or electronically reconfigurable states during the CPI, not continuous mechanical motion between pulses.Source: §V · p. 6
  • The comparison includes an unoptimized random initialization, BS-detection optimization without Eve-privacy terms, fixed-MA Eve-deception optimization, and joint privacy-aware optimization.The benchmarks use the same feasible random MA initialization and power-normalized channel-conjugate downlink beamformer.Source: §V · p. 6
  • At ρ = 0.5, the proposed design shifts Eve’s detector output toward the dummy DD bin, unlike the benchmark maps that retain a dominant true-bin response.Benchmark 3 only partially reduces the true-bin response; the proposed design also creates a more pronounced dummy peak.Source: Figure 1 · p. 6, §V · p. 6
  • Proposed design outperforms all benchmarks in Eve false-alarm probability, particularly for 0.3 < ρ < 0.6.False-alarm probability is defined as selecting the dummy DD bin.Source: Figure 2 · p. 6, §V · p. 6

VI. Conclusion

The paper jointly optimizes transmit beamforming and pulse-indexed MA states to create a deceptive dummy response at Eve while preserving BS true-bin sensing. Simulations report improved dummy-bin selection over benchmarks while satisfying BS sensing and downlink QoS requirements.

VI. Conclusion

  • Jointly optimized beamforming and pulse-indexed MA states induce a dummy response at Eve while preserving BS true-bin sensing.Source: §VI · p. 6
  • A four-stage AO heuristic yields feasible suboptimal solutions under sensing, communication, power, and MA constraints.The detector-aware formulation combines a worst-case dummy-to-true SINR margin, dummy-bin selection, and BS sensing control.Source: §VI · p. 6
  • Simulations show suppressed Eve true-bin selection and improved dummy-bin selection over benchmarks while satisfying BS sensing and downlink QoS requirements.Source: §VI · p. 6
Loading 2609.35534v1…