High-Performance Estimation of Jamming Covariance Matrix for IRS-Aided Directional Modulation Network With a Malicious Attacker

Hangjia He, Ting Su, Hongjun Wang, Yin Teng, Weiping Shi, Feng Shu, Jiangzhou Wang

TL;DR

This paper tackles the challenge of robust communication in the presence of malicious jamming in IRS-assisted directional modulation systems. The authors propose three novel methods – EVD, PEM-GD, and PEM-AO – to accurately estimate the jamming covariance matrix (JCM), which is crucial for effective anti-jamming receive beamforming. Their methods significantly outperform existing techniques, especially in low to medium jamming-to-noise ratio scenarios.

What Problem Does This Paper Address?

In modern wireless communication, ensuring secure and reliable data transmission is paramount. Directional Modulation (DM) systems offer a way to send signals in specific directions, enhancing security. The integration of Intelligent Reflecting Surfaces (IRS) further boosts performance by intelligently shaping the wireless environment. However, these systems are vulnerable to malicious jamming, where an attacker (Mallory) intentionally disrupts communication. To counter this, receivers employ Receive Beamforming (RBF) techniques, which often rely on knowing the statistical properties of the jamming signal, specifically its covariance matrix (JCM). The challenge lies in accurately estimating this JCM at the receiver (Bob) when the attacker is non-cooperative and the system is complex, involving Alice (transmitter), Bob (receiver), and an IRS. Existing methods for JCM estimation, like the Sample Covariance Matrix (SCM), can be inaccurate, especially with limited data samples. This paper aims to bridge this gap by proposing more precise and efficient JCM estimation techniques.

Methods

The paper introduces three novel methods for estimating the Jamming Covariance Matrix (JCM):

  • Eigenvalue Decomposition (EVD) with Rank-2 Constraint: This method leverages the inherent property that the ideal JCM in this scenario has a rank of at most two. By imposing this rank constraint on the eigenvalue decomposition of the received signal's covariance matrix, the method aims for a more accurate estimation than standard EVD. It's noted that this method performs better than SCM but is slightly inferior to RCML, though it has the advantage of not requiring prior knowledge of the receiver noise variance.
  • Parametric Estimation Method by Gradient Descent (PEM-GD): This approach exploits the specific structural features of the JCM. The authors derive an expression for the JCM that depends on a set of unknown parameters. The estimation problem is then framed as an optimization problem to minimize the Euclidean distance between the estimated JCM and the sample covariance matrix. Gradient descent is used to iteratively update these parameters. A key advantage is that the estimated JCM is independent of phase changes in the IRS.
  • Parametric Estimation Method by Alternating Optimization (PEM-AO): Recognizing the computational complexity of PEM-GD, this method aims to reduce it by employing an alternating optimization strategy. It effectively reduces the number of optimization variables by decomposing the problem and iteratively optimizing subsets of parameters. This leads to a lower computational cost while achieving comparable performance to PEM-GD.

Datasets

Not discussed in the paper. The evaluation is based on numerical simulations with parameters set to represent a realistic scenario.

Key Insights and Ideas

  • Rank-2 Property is Key: The paper highlights that the jamming covariance matrix in this specific IRS-aided DM setup has a fundamental rank of at most two. Exploiting this property is crucial for improving estimation accuracy, as demonstrated by the proposed EVD method.
  • Structure Matters for Performance: Moving beyond simple statistical estimation (like SCM), the PEM-GD and PEM-AO methods leverage the underlying mathematical structure of the JCM. This structural insight allows for more precise parameter estimation, leading to better anti-jamming performance.
  • Complexity-Performance Trade-off: The development of PEM-AO from PEM-GD illustrates a common theme in signal processing: balancing estimation accuracy with computational efficiency. PEM-AO achieves near-identical performance to PEM-GD but with significantly reduced complexity, making it more practical for real-time applications.
  • Jamming Estimation is Critical for RBF: The paper strongly emphasizes that accurate JCM estimation is not just an academic exercise but a fundamental prerequisite for designing effective Receive Beamforming (RBF) strategies to combat malicious jamming.

Key Results

  • The proposed EVD, PEM-GD, and PEM-AO methods all outperform the standard Sample Covariance Matrix (SCM) method in terms of NMSE (Normalized Mean Squared Error), particularly in low and medium jamming-to-noise ratio (JNR) regions.
  • PEM-GD and PEM-AO demonstrate superior performance compared to the EVD method and the RCML (Rank-Constrained Maximum Likelihood) estimator.
  • In terms of Secrecy Rate (SR), using JCM estimated by PEM-AO or PEM-GD leads to higher SR compared to other estimation methods.
  • PEM-AO exhibits a faster convergence rate than PEM-GD and has lower computational complexity, making it a more practical choice.
  • As the dimension of the JCM matrix (NB) increases, the NMSE of all methods worsens, but the performance advantage of PEM-GD and PEM-AO becomes more pronounced.

Limitations and Future Work

  • Computational Complexity: While PEM-AO significantly reduces complexity compared to PEM-GD, the paper notes that the complexity of PEM-GD and PEM-AO is still higher than RCML and EVD. Further research could focus on even more efficient algorithms.
  • Ideal JCM Rank Assumption: The EVD method relies on the assumption of a rank-2 JCM. While this is derived from the system model, real-world scenarios might introduce deviations.
  • Mallory's Antenna Count: The PEM-GD method involves initializing parameters with a dimension greater than the surmised number of Mallory's antennas, suggesting that precise knowledge of the attacker's capabilities could further refine the estimation.
  • Future Work: The paper implicitly suggests that exploring more advanced optimization techniques for the parametric methods and investigating their performance under more complex jamming scenarios (e.g., non-Gaussian jamming) would be valuable future research directions.

Summary updated

Abstract

In this paper, we investigate the anti-jamming problem of a directional modulation (DM) system with the aid of intelligent reflecting surface (IRS). As an efficient tool to combat malicious jamming, receive beamforming (RBF) is usually designed by using the statistical properties of the jamming received by Bob. Thus, it is very necessary to estimate the receive jamming covariance matrix (JCM) at Bob. To achieve a precise JCM estimation, three JCM estimation methods, including eigenvalue decomposition (EVD), parametric estimation method by gradient descend (PEM-GD) and parametric estimation method by alternating optimization (PEM-AO), are proposed. Here, the proposed EVD is derived according to the rank-2 constraint of JCM. The PEM-GD method fully explores the structure features of JCM and the PEM-AO is proposed to decrease the computational complexity of the former via dimensionality reduction. The simulation results show that the proposed three methods perform better than directly using sample covariance matrix. Additionally, the proposed PEM-GD and PEM-AO outperform EVD method and the clutter and disturbance covariance estimator RCML.
BibTeX
@article{he2022high,
  author = {Hangjia He and Ting Su and Hongjun Wang and Yin Teng and Weiping Shi and Feng Shu and Jiangzhou Wang},
  title = {{High-Performance Estimation of Jamming Covariance Matrix for IRS-Aided Directional Modulation Network With a Malicious Attacker}},
  journal = {IEEE Trans. Vehicular Technology},
  year = {2022},
  url = {https://openreview.net/forum?id=MuDWrTL9eC},
}
Figure