Abstract
This paper considers the problem of detecting a high dimensional signal (not necessarily sparse) based on compressed measurements with physical layer secrecy guarantees. First, we propose a collaborative compressive detection (CCD) framework to compensate for the performance loss due to compression with a single sensor. We characterize the trade-off between dimensionality reduction achieved by a universal compressive sensing (CS) based measurement scheme and the achievable performance of CCD analytically. Next, we consider a scenario where the network operates in the presence of an eavesdropper who wants to discover the state of the nature being monitored by the system. To keep the data secret from the eavesdropper, we propose to use cooperating trustworthy nodes that assist the fusion center (FC) by injecting artificial noise to deceive the eavesdropper. We seek the answers to the questions: Does CS help improve the security performance in such a framework? What are the optimal values of parameters which maximize the CS based collaborative detection performance at the FC while ensuring perfect secrecy at the eavesdropper?
Introduction
In the context of the Internet of Things (IoT) and remote sensing, devices often process high-dimensional data, such as high-resolution images or complex radio spectrums. Compressive Sensing (CS) is a widely adopted technique to reduce the volume of transmitted data by taking fewer, random measurements. However, CS introduces two significant challenges: performance loss, as compression makes it harder to distinguish signals from noise, and security risks, as wireless transmissions are susceptible to interception by eavesdroppers (Eve). This paper addresses these challenges by proposing a collaborative framework that utilizes multiple sensors to recover performance and introduces an artificial noise injection strategy to achieve physical layer secrecy.
Collaborative Compressive Detection (CCD) Framework
The authors develop a CCD framework based on a parallel network of $N$ sensors observing a phenomenon. As illustrated in the system model, each sensor $S_i$ captures a high-dimensional observation $u_i$ and compresses it into a low-dimensional vector $y_i$ using a random projection matrix $\phi$. These compressed reports are then transmitted to a central Fusion Center (FC).
The FC employs a Likelihood Ratio Test (LRT) to make a global decision. The performance is analyzed using the $\epsilon$-stable embedding concept, which describes how random projections preserve the distance between signals. A key finding is the $1/c$ recovery rule: the performance loss from a compression ratio $c$ can be fully recovered if the number of collaborating sensors $N$ is at least $1/c$, demonstrating that spatial diversity can effectively substitute for individual data resolution.
Secrecy via Artificial Noise Injection
To counter eavesdropping, the framework incorporates "trustworthy Byzantines"—nodes that intentionally inject artificial noise into their reports. This strategy aims to deceive the eavesdropper without sacrificing the accuracy of the FC's final decision. The paper utilizes the Modified Deflection Coefficient to optimize system parameters, including the compression ratio and the number of noise-injecting nodes, to maximize detection performance at the FC while ensuring perfect secrecy at the eavesdropper.
Key Insights and Results
The study, conducted using synthetic multivariate Gaussian models, reveals several critical insights:
- Exponential Convergence: The probability of error at the FC vanishes exponentially as either the number of nodes ($N$) or the compression ratio ($c$) increases.
- Secrecy Independence: Counter-intuitively, in a perfect secrecy regime, the system's security performance does not depend on the compression ratio.
- Deterministic Noise Advantage: Injecting a deterministic signal is proven to be more effective than random noise for deceiving the eavesdropper.
- Optimal Design: To maximize performance under perfect secrecy, designers should select the maximum allowable compression ratio ($c_{max}$) and the minimum fraction of noise-injecting nodes ($\alpha_{min}$).
Conclusion and Future Work
The paper establishes a theoretical foundation for secure collaborative sensing. While the current model assumes Gaussian distributions and ideal communication channels, it provides analytical bounds for detection performance. Future research will explore non-Gaussian signal models, noisy communication links, and scenarios where the perfect secrecy constraint is relaxed to potentially further boost performance.
Summary updated