Analysis of key generation rate from wireless channel in in-band full-duplex communications

Alireza Sadeghi, Michele Zorzi, Farshad Lahouti

Abstract

In this paper, the impact of in-band full-duplex (IBFD) wireless communications on secret key generation via physical layer channel state information is investigated. A key generation strategy for IBFD wireless devices to increase the rate of generated secret keys over multipath fading channels is proposed. Conventionally, due to the half-duplex (HD) constraint on wireless transmissions, sensing simultaneous reciprocal channel measurements is not possible, which leads to a degraded key generation rate. However, with the advent of IBFD wireless devices, the legitimate nodes can sense the shared wireless link simultaneously at the possible cost of a self-interference (SI) channel estimation and some residual self-interference (RSI). As we demonstrate, with HD correlated observations the key rate is upper bounded by a constant, while with IBFD the key rate is only limited by the SI cancellation performance and is in general greater than that of its HD counterpart. Our analysis shows that with reasonable levels of SI cancellation, in the high SNR regime the key rate of IBFD is much higher, while in low SNRs, the HD system performs better. Finally, the key rate loss due to the overhead imposed by the SI channel estimation phase is discussed.

Introduction

Physical Layer Security (PLS) leverages the inherent randomness and reciprocity of wireless channels to establish secret keys between two legitimate nodes, Alice and Bob. Because the multipath characteristics between two points are identical in both directions, both nodes can derive matching keys by measuring channel state information (CSI). As shown in Figure 1, these measurements are highly correlated for legitimate users, while an eavesdropper (Eve) located just a few wavelengths away observes a completely different, uncorrelated channel.

RSSI measurements showing channel reciprocity between Alice and Bob, and decorrelation for an eavesdropper Eve.

The Half-Duplex Bottleneck

Traditional wireless hardware is constrained by a half-duplex (HD) limitation, meaning it cannot transmit and receive on the same frequency simultaneously. This forces Alice and Bob to measure the channel at different times ($T_1$ and $T_2$). In dynamic environments where the channel changes rapidly (high mobility), this time lag causes the measurements to decorrelate. This temporal decorrelation acts as a fundamental "hard ceiling" on the secret key generation rate, regardless of how much transmission power is used.

Proposed In-Band Full-Duplex (IBFD) Strategy

To overcome this bottleneck, the authors propose utilizing In-Band Full-Duplex (IBFD) technology. IBFD allows nodes to transmit and receive at the same time, enabling simultaneous sensing of the reciprocal channel.

Comparison of training frame structures for (a) traditional Half-Duplex systems and (b) the proposed In-Band Full-Duplex system.

As illustrated in Figure 4, the proposed IBFD strategy employs a specific frame structure. A small fraction of the time ($\alpha T$) is dedicated to estimating the node's own Self-Interference (SI) channel. Following this, Alice and Bob simultaneously exchange training sequences ($S_A$ and $S_B$). While this introduces Residual Self-Interference (RSI), it eliminates the time-lag error inherent in HD systems.

Key Results and Insights

Through mathematical modeling and numerical simulations, the paper establishes several critical findings:

  • High SNR Dominance: In high Signal-to-Noise Ratio (SNR) regimes, the IBFD key rate is significantly higher than the HD counterpart. The benefit of simultaneous sensing far outweighs the noise introduced by RSI.
  • Robustness to Mobility: When the channel changes very fast (low temporal correlation), the HD key rate crashes toward zero. In contrast, the IBFD rate remains stable or even increases, as the rapid changes provide more "fresh" randomness to be captured simultaneously.
  • Low SNR Trade-off: At low SNRs, HD systems perform better because the overhead of SI estimation and the impact of RSI are more detrimental than the temporal decorrelation.

Conclusion

IBFD technology offers a transformative path for wireless security. By enabling simultaneous channel sensing, it bypasses the fundamental temporal limitations of half-duplex hardware, allowing for significantly higher secret key rates in high-mobility and high-signal environments. Future research will focus on optimizing the time-split between SI estimation and channel sensing to further maximize these gains.

Summary updated

Abstract

In this paper, the impact of in-band full-duplex (IBFD) wireless communications on secret key generation via physical layer channel state information is investigated. A key generation strategy for IBFD wireless devices to increase the rate of generated secret keys over multipath fading channels is proposed. Conventionally, due to the half-duplex (HD) constraint on wireless transmissions, sensing simultaneous reciprocal channel measurements is not possible, which leads to a degraded key generation rate. However, with the advent of IBFD wireless devices, the legitimate nodes can sense the shared wireless link simultaneously at the possible cost of a self-interference (SI) channel estimation and some residual self-interference (RSI). As we demonstrate, with HD correlated observations the key rate is upper bounded by a constant, while with IBFD the key rate is only limited by the SI cancellation performance and is in general greater than that of its HD counterpart. Our analysis shows that with reasonable levels of SI cancellation, in the high SNR regime the key rate of IBFD is much higher, while in low SNRs, the HD system performs better. Finally, the key rate loss due to the overhead imposed by the SI channel estimation phase is discussed.
BibTeX
@misc{sadeghi2016analysis,
  author = {Alireza Sadeghi and Michele Zorzi and Farshad Lahouti},
  title = {{Analysis of key generation rate from wireless channel in in-band full-duplex communications}},
  year = {2016},
  eprint = {1605.09715},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1605.09715},
}
Figure