Total Variation Distance Based Performance Analysis of Covert Communication over AWGN Channels in Non-asymptotic Regime

Xinchun Yu, Shuangqin Wei, Yuan LuoarXiv 2020 · DOI · arXiv

Abstract

This paper investigates covert communication over an additive white Gaussian noise (AWGN) channel in finite block length regime on the assumption of Gaussian codebooks. We first review some achievability and converse bounds on the throughput under maximal power constraint. From these bounds and the analysis of TVD at the adversary, the first and second asymptotics of covert communication are investigated by the help of some divergences inequalities. Furthermore, the analytic solution of TVD, and approximation expansions which can be easily evaluated with given snr (signal noise ratio) are presented. In this way, the proper power level for covert communication can be approximated with given covert constraint of TVD, which leads to more accurate estimation of the power compared with preceding bounds. Further elaboration on the effect of such asymptotic characteristics on the primary channels's throughput in finite block regime is also provided.

Introduction

Covert communication, or Low Probability of Detection (LPD) communication, aims to hide the very existence of a transmission from an adversary. Traditionally, this is governed by the Square Root Law (SRL), which posits that $O(\sqrt{n})$ bits can be sent covertly over $n$ channel uses. However, practical systems operate with finite block lengths, where asymptotic assumptions fail to provide accurate performance bounds. This paper addresses this gap by focusing on the non-asymptotic regime using Total Variation Distance (TVD) as the primary metric for covertness.

System Model and Problem Formulation

The communication scenario involves three parties: Alice (the transmitter), Bob (the intended receiver), and Willie (the adversary). Alice attempts to send a message to Bob over an Additive White Gaussian Noise (AWGN) channel while remaining undetected by Willie.

The standard covert communication model featuring Alice, Bob, and the adversary Willie.

Willie's goal is to perform a hypothesis test to determine if the received signal consists only of noise ($H_0$) or signal plus noise ($H_1$). The effectiveness of Willie's detection is measured by the sum of false alarm and missed detection probabilities, which is directly related to the TVD between the two distributions. TVD is considered the "gold standard" for covertness because it directly maps to the adversary's minimum total probability of error.

Methodology

The authors employ several sophisticated mathematical techniques to bridge the gap between theory and practice:

  • Analytic TVD Derivation: They derive a closed-form analytic solution for the TVD between two multivariate Gaussian distributions using incomplete gamma functions.
  • Non-asymptotic Analysis: Instead of just looking at limits, they investigate the first-order ($O(\sqrt{n})$) and second-order ($O(n^{1/4})$) asymptotics of throughput.
  • Approximation Expansions: Recognizing that exact gamma functions are computationally heavy, they provide series expansions that allow engineers to quickly estimate TVD for a given Signal-to-Noise Ratio (SNR).

Key Insights and Geometric Analysis

A critical aspect of the analysis is the behavior of Gaussian codebooks in finite dimensions. The paper explores "sphere hardening," where the power of the transmitted sequence concentrates near the surface of a high-dimensional sphere.

Geometric interpretation of power constraints and sphere hardening in the non-asymptotic regime.

This geometric insight justifies the effectiveness of Gaussian codebooks for block lengths as small as $n \ge 500$. The authors further refine the SRL by identifying a second-order term of $O(n^{1/4})$ for throughput, providing a granular view of the trade-offs between power, block length, and covertness. They also identify a critical threshold for the power scaling law $SNR = n^{-\tau}$: if $\tau < 0.5$, the signal is eventually detected with 100% certainty; if $\tau > 0.5$, it remains hidden.

Results and Conclusion

The study yields an exact formula for TVD at the adversary and demonstrates that using Hellinger distance bounds provides a tighter estimation of the maximum "safe" power level compared to standard K-L divergence bounds. By moving beyond asymptotic limits, this work provides practical tools for designing covert wireless systems. Future research will look to extend these TVD-based analyses to more complex scenarios, such as MIMO (Multiple-Input Multiple-Output) systems and channels with uninformed adversaries.

Summary updated

Abstract

This paper investigates covert communication over an additive white Gaussian noise (AWGN) channel in finite block length regime on the assumption of Gaussian codebooks. We first review some achievability and converse bounds on the throughput under maximal power constraint. From these bounds and the analysis of TVD at the adversary, the first and second asymptotics of covert communication are investigated by the help of some divergences inequalities. Furthermore, the analytic solution of TVD, and approximation expansions which can be easily evaluated with given snr (signal noise ratio) are presented. In this way, the proper power level for covert communication can be approximated with given covert constraint of TVD, which leads to more accurate estimation of the power compared with preceding bounds. Further elaboration on the effect of such asymptotic characteristics on the primary channels's throughput in finite block regime is also provided. The results will be very helpful for understanding the behavior of the total variation distance and practical covert communication.
BibTeX
@misc{yu2020total,
  author = {Xinchun Yu and Shuangqin Wei and Yuan Luo},
  title = {{Total Variation Distance Based Performance Analysis of Covert Communication over AWGN Channels in Non-asymptotic Regime}},
  year = {2020},
  eprint = {2007.03456},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2007.03456},
}
Figure