Dual Security for MIMO-OFDM ISAC Systems: Artificial Ghosts or Artificial Noise

Technical Summary: Dual Security for MIMO-OFDM ISAC Systems

Problem Statement
Integrated Sensing and Communication (ISAC) systems, recognized as a key enabler for 6G, unify spectrum and hardware to support simultaneous data transmission and environmental perception. However, this convergence introduces novel physical-layer security vulnerabilities. While communication security (protecting data from eavesdroppers) is well-studied, sensing security remains largely unexplored. Specifically, passive sensing eavesdroppers (Eves) can exploit the broadcast nature of electromagnetic waves to intercept direct-path signals as reference waveforms and reflected echoes from targets. By jointly processing these signals, a passive sensing Eve can infer sensitive target parameters (velocity, location, physiological states) without the legitimate users' knowledge. Existing defensive mechanisms often assume perfect Channel State Information (CSI) of the eavesdropper or rely on non-coherent detection, which is impractical when the Eve is close to the transmitter and possesses high-quality reference signals. Furthermore, current solutions typically address either communication or sensing security in isolation, failing to provide a joint defense strategy under unknown CSI conditions.

Methodology
This paper proposes a two-layer dual-secure framework for MIMO-OFDM ISAC systems that protects against both communication eavesdroppers (who intercept data) and sensing eavesdroppers (who infer target parameters), without requiring the CSI of either adversary. The system model involves a monostatic ISAC base station (BS) serving multiple communication users (CUs) and sensing far-field targets, while facing potential passive sensing and communication Eves.

The proposed defense mechanism operates on two distinct layers:

1. Artificial Noise (AN) Injection: The BS jointly designs transmit beamformers to inject AN. This AN serves a dual purpose: it creates strong interference for communication Eves to reduce information leakage (secrecy rate), and it deliberately distorts the reference signal received by sensing Eves. By corrupting the reference signal, the AN impairs the sensing Eve's ability to perform matched filtering and parameter estimation.
2. Artificial Ghosts (AGs): To address scenarios where sensing Eves are located close to the BS and may still recover a high-quality reference signal despite AN, a second layer of defense is introduced. The system deliberately embeds "Artificial Ghosts"—fake targets with fabricated angle-delay-Doppler profiles—into the sensing Eve's estimation results. This is achieved by optimizing the Ambiguity Function (AF) to generate artificial peaks in the angle-delay-Doppler domain. These AGs are observable only by the sensing Eve, misleading them toward false targets and significantly reducing the probability of correctly detecting the true targets.

The design is formulated as a non-convex optimization problem aiming to maximize the sensing Signal-to-Noise Ratio (SNR) at the legitimate BS. This maximization is subject to constraints on:

• Communication Performance: Minimum data rate for legitimate CUs.
• Communication Security: Minimum secrecy rate for CUs against communication Eves.
• Sensing Security: Constraints on the Integrated Sidelobe Level (ISL) and Peak Sidelobe Level (PSL) to control the magnitude of AGs, and a constraint on the maximum allowable data rate at the sensing Eve to degrade its reference signal quality.

To solve the resulting non-convex problem, the authors employ convex surrogate functions via difference-of-convex (DC) programming and first-order Taylor approximations. An iterative algorithm is used to update the beamforming matrices and AN covariance matrices, with Gaussian randomization applied to recover rank-one solutions.

Key Contributions

1. Dual-Secure Framework: The paper introduces a novel design that simultaneously secures communication and sensing against passive eavesdroppers without requiring their CSI. It uniquely combines AN injection (to degrade reference signals) with AG generation (to create false targets) to ensure sensing security even when the Eve is in a favorable position.
2. Performance Metrics and Formulation: The authors establish specific performance measures for communication performance (data rate), communication security (secrecy rate), sensing performance (post-matched filter SNR), and sensing security (ISL, PSL, and Eve's reference signal rate). They formulate a four-way trade-off optimization problem balancing these metrics under limited power resources.
3. Tractable Solution: The paper provides a tractable solution to the non-convex optimization problem by introducing convex surrogates and an iterative algorithm, verified through numerical simulations.

Results
Numerical results demonstrate the effectiveness of the proposed framework:

• Trade-offs: The study reveals inherent trade-offs among the four key metrics. Increasing communication security (secrecy rate) or sensing security (via higher ISL/PSL for AGs) inevitably degrades the sensing performance (SNR) at the legitimate BS and, to some extent, communication performance.
• Detection Probability: The proposed design significantly reduces the correct detection probability of sensing Eves. When AGs are employed, the sensing Eve's detection probability drops monotonically as the PSL threshold increases, as the AGs trigger false alarms.
• Robustness: The dual-layer approach remains effective even when the sensing Eve uses a Minimum Variance Unbiased (MVU) estimator to recover the reference signal. The combination of AN and AGs creates a high noise floor and obscures target signatures for the Eve, while the legitimate BS, possessing prior knowledge of true target angles, can distinguish and suppress the AGs.
• Visual Evidence: Delay-Doppler and Delay-Angle maps confirm that AGs are clearly observable by the sensing Eve but are negligible or suppressible for the legitimate BS, validating the "ghost" nature of the deception.

Significance and Claims
The paper claims that its primary significance lies in bridging the research gap where communication and sensing security must be jointly addressed in the absence of eavesdropper CSI. Unlike previous works that assume perfect CSI or focus solely on one domain, this framework offers a practical, two-layer defense mechanism. The authors assert that their design effectively enhances both communication and sensing security while preserving the performance of legitimate users and receivers, albeit with quantifiable trade-offs. The work highlights that relying solely on communication-centric waveforms for sensing is inherently vulnerable and that active deception (AGs) combined with interference (AN) is necessary to secure ISAC systems against sophisticated passive sensing threats.