PaperlayerView exact revision on arXiv

Source-linked AI summary

Secure Intelligent Reflecting Surface Aided Integrated Sensing and Communication

Meng Hua, Qingqing Wu, Wen Chen, Octavia A. Dobre, A. Lee Swindlehurst

arXiv:2207.09095v1cs.ITeess.SP

TL;DR

The paper addresses secure ISAC when an IRS-assisted sensing target may eavesdrop on multiuser downlink communication. It jointly designs BS communication and radar beamformers with IRS phases for ideal and uncertain channel/location information, using penalty-based and robust algorithms. Simulations demonstrate a trade-off between communication and sensing quality.

  • Problem

    The paper studies how to enhance physical-layer security and sensing in IRS-aided ISAC when the target may intercept users’ information.

  • Method

    It jointly optimizes communication beamformers, radar beamformers, and IRS phase shifts using a penalty-based method for perfect information and a robust S-procedure/sign-definiteness method for uncertainty.

  • Results

    Simulations demonstrate an effective trade-off between communication quality and target sensing quality.

  • Takeaways & Limitations

    IRS-assisted ISAC can support sensing and improve physical-layer security while dedicated sensing signals further improve system performance.

Abstract

from arXiv · show

In this paper, an intelligent reflecting surface (IRS) is leveraged to enhance the physical layer security of an integrated sensing and communication (ISAC) system in which the IRS is deployed to not only assist the downlink communication for multiple users, but also create a virtual line-of-sight (LoS) link for target sensing. In particular, we consider a challenging scenario where the target may be a suspicious eavesdropper that potentially intercepts the communication-user information transmitted by the base station (BS). We investigate the joint design of the phase shifts at the IRS and the communication as well as radar beamformers at the BS to maximize the sensing beampattern gain towards the target, subject to the maximum information leakage to the eavesdropping target and the minimum signal-to-interference-plus-noise ratio (SINR) required by users. Based on the availability of perfect channel state information (CSI) of all involved user links and the accurate target location at the BS, two scenarios are considered and two different optimization algorithms are proposed. For the ideal scenario where the CSI of the user links and the target location are perfectly known at the BS, a penalty-based algorithm is proposed to obtain a high-quality solution. In particular, the beamformers are obtained with a semi-closed-form solution using Lagrange duality and the IRS phase shifts are solved for in closed form by applying the majorization-minimization (MM) method. On the other hand, for the more practical scenario where the CSI is imperfect and the target location is uncertain, a robust algorithm based on the $\cal S$-procedure and sign-definiteness approaches is proposed. Simulation results demonstrate the effectiveness of the proposed scheme in achieving a trade-off between the communication quality and the sensing quality.

I. INTRODUCTION

This paper develops a secure IRS-aided ISAC system that jointly supports communication, sensing, and physical-layer security, including when the sensed target may eavesdrop. It formulates joint beamforming and IRS phase-shift designs for ideal and imperfect-information scenarios, with simulations showing a communication–sensing trade-off.

  • Motivation: IRS technology can create virtual LoS links for communication and sensing while offering low-power, low-cost control of wireless propagation.An IRS uses many reflecting elements whose phases and/or amplitudes can be independently controlled to provide beamforming gains.
  • Problem formulation: The proposed system assists downlink communication for multiple users and senses a target that may attempt to intercept information transmitted by the BS.The target is treated as a potential eavesdropper, making physical-layer security part of the ISAC design.
  • Problem formulation: The design jointly optimizes communication beamformers, radar beamformers, and IRS phase shifts to maximize sensing beampattern gain under leakage and user-SINR constraints.The formulation limits information leakage to the target while enforcing minimum SINR requirements for legitimate users.
  • Optimization methods: With perfect user-link CSI and known target location, a penalty-based algorithm addresses the coupled non-convex problem and unit-modulus IRS constraints.Lagrange duality yields beamformers semi-closed form, while MM provides a closed-form update for IRS phase shifts.
  • Optimization methods: With imperfect CSI and uncertain target location, a robust AO algorithm uses the S-procedure and sign-definiteness approaches to convert infinitely many inequalities into finite LMIs.The AO method alternates between transmit-beamformer and IRS-phase-shift optimization.
  • Results: Simulation results verify a trade-off between communication quality and target sensing quality and show that dedicated sensing signals further improve system performance.The results also indicate the potential of IRS technology to increase beampattern gains and improve ISAC security.

2) Radar Sensing and Interception Model:

The model uses the IRS to support target sensing while treating the target as a potential eavesdropper, with communication leakage constrained alongside user SINR. It formulates both perfect-information and robust uncertain-information designs around maximizing target beampattern gain.

  • The IRS creates a virtual LoS sensing link while the target may intercept information transmitted to legitimate users.
  • The design maximizes target beampattern gain subject to minimum user SINR and maximum tolerable information leakage.
  • Perfect CSI and known target location: The perfect-information formulation assumes known communication CSI and target location at the BS.
  • Imperfect CSI and uncertain target location: The robust formulation accounts for imperfect communication CSI, an uncertain target location region, and bounded channel-error models.
  • Both formulations are non-convex because IRS coefficients have unit-modulus constraints and optimization variables are coupled; the robust problem also contains infinitely many inequalities.

III. PROPOSED SOLUTION FOR PERFECT CSI AND KNOWN TARGET LOCATION

For perfect CSI and a known target location, the paper uses a penalty-based two-layer procedure to decouple coupled variables. Auxiliary variables, beamformers, and IRS phases are optimized through QCQP, duality, SCA, and MM-based updates.

  • The perfect-information case is treated as a performance upper bound, using auxiliary variables to decouple constraints between optimization-variable blocks.
  • Inner Layer Optimization: The inner layer alternates optimization over auxiliary variables, transmit beamformers, and IRS phase shifts.
  • Inner Layer Optimization: Auxiliary-variable subproblems are handled using QCQP duality, with optimal solutions obtained from Lagrange multipliers and complementary-slackness conditions.
  • Inner Layer Optimization: Transmit beamformers have a semi-closed-form optimal solution from Lagrange duality, while associated scalar searches use bisection because relevant functions are monotonic.
  • Inner Layer Optimization: The IRS phase-shift subproblem is non-convex under unit-modulus constraints, so MM constructs a convex surrogate and yields a closed-form phase update.
  • Overall Algorithm: The inner procedure updates auxiliary variables, beamformers, and phases until the objective increase falls below an inner threshold.

B. Outer Layer Optimization

The outer layer progressively updates the penalty parameter to reduce violations of the auxiliary equality constraints while controlling convergence through a scaling factor.

  • The penalty parameter is updated across outer iterations to penalize violations of the auxiliary equality constraints.
  • The scaling factor c controls the convergence behavior of the penalty-parameter update.

C. Overall Algorithm and Computational Complexity

The overall algorithm terminates when its penalty-based equality-constraint indicator is sufficiently small, with complexity expressed in terms of accuracy and inner and outer iteration counts.

  • The outer algorithm terminates when the indicator ξ falls below a predefined threshold, treating the penalty constraint as satisfied to the given accuracy.
  • The computational complexity is expressed using iteration accuracy ε and the inner and outer convergence counts Iin and Iout.

IV. PROPOSED SOLUTION FOR IMPERFECT CSI AND UNCERTAIN TARGET LOCATION

For imperfect CSI and uncertain target location, the paper replaces infinitely many robust constraints with finite LMIs and solves the resulting problem by alternating optimization of BS beamformers and IRS phase shifts.

  • Infinite robust constraints arise because CSI and target-location uncertainty must be handled in constraints (12b) and (12c).
  • The S-procedure converts infinitely many quadratic inequalities into equivalent finite sets of linear matrix inequalities with nonnegative auxiliary variables.The resulting LMIs can be handled using convex optimization techniques.
  • Schur’s complement transforms the robust SINR-related constraint into LMIs before uncertainty is eliminated through the lemma-based reformulation.
  • Nonconvex quadratic terms are replaced by lower bounds or first-order Taylor approximations, producing convex subproblems in the alternating procedure.The lower bound is linear and convex with respect to the IRS phase vector.
  • Algorithm 2 alternately updates BS beamformers and IRS phase shifts until the fractional objective increase falls below ε.The beamformer subproblem is an SDP, while the IRS phase-shift subproblem uses a square penalty and successive convex approximation.

1) Optimizing BS beamformers with fixed IRS phase shifts:

With IRS phase shifts fixed, the BS beamformer subproblem is formulated as a semidefinite program that standard convex optimization solvers can efficiently tackle.

  • With fixed IRS phase shifts, the BS beamformer subproblem is an SDP solvable by standard convex optimization solvers.

2) Optimizing IRS phase shifts with fixed BS beamformers:

With BS beamformers fixed, the IRS phase-shift subproblem handles the unit-modulus constraint through a square penalty and solves a convex approximation iteratively.

  • The IRS phase-shift subproblem is nonconvex mainly because of the unit-modulus constraint.
  • A sufficiently large square-penalty parameter enforces the unit-modulus constraint at the optimum without requiring gradual penalty adjustment.
  • Successive convex approximation replaces the nonconvex objective with a linear lower bound, yielding a convex problem solvable by convex optimization solvers.
  • Algorithm 2 alternates the two SDP subproblems until convergence, with complexity depending on iteration count L and the number of LMIs.

V. NUMERICAL RESULTS

The numerical setup evaluates secure transmission in a three-dimensional IRS-aided ISAC deployment with randomly distributed users, a specified target direction, and Rician or Rayleigh fading links.

  • The simulation places the BS at (0, 0, 2.5) m, the IRS at (20, 0, 2.5) m, and users randomly within a 2 m radius around (20, 5, 0) m.
  • The target uses azimuth θ = −30° and elevation ϕ = 40°, with a 10 m IRS-target distance and path loss exponent 2.
  • The BS-IRS and IRS-user links use Rician fading with factor 3 dB, while the BS-user link uses Rayleigh fading.The corresponding path loss exponents are 2.2 for the Rician links and 3.6 for the BS-user link.
  • The listed simulation parameters include noise power −90 dBm, ρ = 0.1, c = 0.85, and inner and outer tolerances of 10^-2 and 10^-4.

A. Perfect CSI and Known Target Location

The ideal case assumes known CSI and target location at the BS, with Algorithm 1 used for optimization. Its convergence is evaluated for multiple numbers of IRS reflecting elements.

  • Algorithm 1 is evaluated under the ideal case where CSI and target location are known at the BS.
  • Convergence behavior is examined for M = 50, M = 100, and M = 150 IRS reflecting elements.
  • The study uses Fig. 2 to assess convergence across different numbers of IRS reflecting elements.

1) Convergence Behavior of Algorithm 1:

The simulations compare the proposed joint design with random-phase, communication-only, separate-beamforming, and zero-forcing approaches. Results examine beampattern gain under varying transmit power and IRS size.

  • Beampattern gain versus Pmax: Beampattern gain increases monotonically with Pmax for all methods, and dedicated radar signals outperform communication signals only.The increase is attributed to suppressed co-channel interference and greater available power, while radar signals provide additional optimization DoFs.
  • Comparison approaches: The proposed approach is compared with random phase, communication signal only, separate beamforming, and communication- and sensing-based ZF methods.The comparison includes jointly designed beamformers and IRS phase shifts alongside alternative designs.
  • Beampattern gain versus M: The proposed approach increasingly outperforms random phase as M grows, while also exceeding the communication-signals-only design.More reflecting elements provide additional resource-allocation DoFs when IRS phases are properly adjusted.

2) Beampattern Gain Versus Transmit Power:

The proposed joint IRS and beamforming design improves sensing beampattern gain while balancing user SINR, information leakage, and imperfect channel or location knowledge. Dedicated radar signals, optimized IRS phases, and robust design together support secure ISAC performance.

  • Beampattern Gain Versus Transmit Power: Increasing transmit power improves beampattern gain, while the proposed approach outperforms communication-only and separate-beamforming baselines.The gain increase is attributed to suppressed co-channel interference and the additional degrees of freedom provided by dedicated radar signals.
  • Beampattern Gain Versus Number of IRS Reflecting Elements: Increasing the number of IRS elements improves beampattern gain, with larger gaps over random phases and baseline beamforming methods.More reflecting elements provide additional resource-allocation degrees of freedom, strengthening the benefit of jointly optimizing transmit beamformers and IRS phases.
  • Beampattern Gain Versus Minimum SINR Required by Communication Users: A stricter user SINR requirement lowers target beampattern gain because the BS and IRS must focus more energy on communication users.The communication-based zero-forcing approach degrades quickly as the SINR requirement increases because only radar signals can contribute to target illumination.
  • Beampattern Gain Versus Maximum Information Leakage SINR to Target: Relaxing the information-leakage constraint leaves the proposed beampattern gain nearly unchanged because radar-power and information-power contributions offset each other.As leakage tolerance increases, less power is allocated to radar signals and more to information signals.
  • Robust Design: Under imperfect CSI and uncertain target location, Algorithm 2 converges monotonically and improves beampattern gain as IRS elements or BS antennas increase.With M = 20 and N = 6, convergence occurs in about 20 iterations; performance remains improvable even at εerror = 0.05.
  • Beampattern Design: The imperfect-CSI beampattern uses a flatter, wider mainlobe and has lower peak gain than the perfect-CSI design while still focusing toward the target region.The wider mainlobe reflects coverage of possible target locations rather than a single known direction.
  • Overall Findings: The proposed secure ISAC scheme jointly optimizes communication beamformers, radar beamformers, and IRS phases to balance sensing quality, communication quality, and secrecy.The ideal and robust scenarios use penalty-based and S-procedure/sign-definiteness-based alternating optimization methods, respectively.
Loading 2207.09095v1…