Introduction
Physical Layer Security (PLS) leverages the inherent randomness of wireless channels to secure communications. By utilizing multi-antenna configurations (MIMO or SIMO), systems can achieve diversity gains that enhance secrecy. However, most theoretical models assume independent fading across all channels. In practice, physical constraints often lead to antenna correlation at the receiver. Furthermore, if an eavesdropper is located near the legitimate receiver, their respective channels may also be correlated. This paper investigates how these dual correlations impact the Secrecy Outage Probability (SOP)—the likelihood that the secrecy capacity falls below a required threshold.
System Model
The researchers analyze a wireless network consisting of a transmitter (A), a legitimate receiver (B), and an eavesdropper (E), as illustrated in Figure 1.
To model the complex correlations between the main channel and the eavesdropper channel, the authors move beyond traditional eigenvalue decomposition, which is insufficient when channels are cross-correlated. Instead, they employ a statistical model where correlated complex Gaussian variables are constructed from linear combinations of independent Gaussian variables. This approach allows for a precise mathematical link between the main and eavesdropper channels.
The study evaluates three primary diversity combining schemes at the receiver:
- Maximal Ratio Combining (MRC): Provides optimal performance but requires full channel state information (CSI).
- Selection Combining (SC): A simpler approach that selects the single best antenna.
- Equal Gain Combining (EGC): A compromise between complexity and performance.
Additionally, the analysis extends to MIMO systems using Transmit Antenna Selection (TAS), considering scenarios with and without knowledge of the eavesdropper’s CSI.
The Impact of Correlation
The paper reveals a nuanced relationship between channel correlation and security:
- Receiver Antenna Correlation (CMC): Correlation among the legitimate receiver's antennas is detrimental. It reduces the effective diversity gain by making the signal paths redundant, thereby increasing the SOP.
- Channel Correlation between Receiver and Eavesdropper (CMEC): Surprisingly, correlation between the main and eavesdropper channels can be beneficial. In high-SNR regimes where the legitimate receiver has a better average signal than the eavesdropper, CMEC ensures that both channels fade in tandem. This prevents the eavesdropper's signal from randomly peaking while the legitimate signal is weak, thus preserving the secrecy gap.
Despite these correlations, the fundamental secrecy diversity order remains unchanged: $M$ for SIMO systems and $M \times N_t$ for MIMO systems using TAS.
Performance Evaluation and Results
The mathematical derivations, validated by Monte Carlo simulations, yield several key results:
- Scheme Hierarchy: MRC consistently outperforms EGC, which in turn outperforms SC. However, under 100% antenna correlation, all schemes converge to the performance of a single-antenna system.
- Correlation Penalty: In high-SNR scenarios, the penalty for receiver antenna correlation is quantified by a factor of $1/\det(U)$, where $U$ is the correlation matrix.
- Eavesdropper CSI: Knowing the eavesdropper's CSI allows for antenna selection based on secrecy capacity rather than just main channel SNR. Lacking this information results in a specific array gain degradation factor of $(MN_t)! / (M!)^{N_t}$ for a single-antenna eavesdropper.
Conclusion
This research provides a rigorous framework for understanding PLS in realistic, correlated environments. While antenna correlation at the receiver degrades performance, the discovery that correlation with the eavesdropper can enhance secrecy offers a new perspective on physical layer security design. Future work may explore these models under non-Rayleigh fading distributions or in the presence of active jamming.
Summary updated