Finite Blocklength Analysis of Gaussian Random Coding in AWGN Channels Under Covert Constraint

Xinchun Yu, Shuangqin Wei, Yuan LuoIEEE Trans. IFS 2020 · DOI · arXiv

Abstract

It is well known that finite blocklength analysis plays an important role in evaluating performances of communication systems in practical settings. This paper considers the achievability and converse bounds on the maximal channel coding rate (throughput) at a given blocklength and error probability in covert communication over AWGN channels. The covert constraint is given in terms of an upper bound on total variation distance (TVD) between the distributions of eavesdropped signals at an adversary with and without presence of active and legitimate communication, respectively. For the achievability, Gaussian random coding scheme is adopted for convenience in the analysis of TVD. The classical results of finite blocklength regime are not applicable in this case. By exploiting and extending canonical approaches, we first present new and more general achievability bounds for random coding schemes under maximal or average probability of error requirements. The general bounds are then applied to covert communication in AWGN channels where codewords are generated from Gaussian distribution while meeting the maximal power constraint. We further show an interesting connection between attaining tight achievability and converse bounds and solving two total variation distance based minimax and maxmin problems. The TVD constraint is analyzed under the given random coding scheme, which induces bounds on the transmission power through divergence inequalities. Further comparison is made between the new achievability bounds and existing ones derived under deterministic codebooks. Our thorough analysis thus leads us to a comprehensive characterization of the attainable throughput in covert communication over AWGN channels.

TL;DR

This paper provides a rigorous theoretical framework for covert communication—sending messages without being detected—in practical, finite-length scenarios. By deriving new "achievability" and "converse" bounds for Gaussian random coding over noisy channels, the authors characterize exactly how much information can be hidden from an adversary while maintaining reliable communication with a legitimate receiver.

What Problem Does This Paper Address?

In traditional secure communication, the primary concern is preventing an eavesdropper from deciphering a message. In covert communication (or Low Probability of Detection), the goal is more fundamental: preventing the adversary (Willie) from even realizing a transmission is occurring.

System model for covert communication involving Alice, Bob, and the adversary Willie.

Information theory establishes that in the "asymptotic" limit (infinite message length), one can send a number of bits scaling with the square root of the blocklength ($O(\sqrt{n})$). However, real-world systems operate with finite blocklengths. Standard tools for finite blocklength analysis do not easily translate to covert settings because the transmitter (Alice) must keep her signal power extremely low to remain hidden. As the message length increases, this power must decrease, breaking the statistical assumptions used in classical communication theory.

Methods

The authors focus on the Additive White Gaussian Noise (AWGN) channel. Their approach involves several key technical components:

  1. Gaussian Random Coding: Codewords are drawn from a Gaussian distribution, which is mathematically convenient for analyzing detection risks and more practical than deterministic codebooks.
  2. Total Variation Distance (TVD): The authors use TVD as the primary metric for "covertness" instead of Kullback-Leibler (KL) divergence. TVD is bounded between 0 and 1 and directly relates to the adversary's probability of making a detection error, often visualized via a Receiver Operating Characteristic (ROC) curve.
Receiver Operating Characteristic (ROC) curve illustrating the trade-off between detection and false alarm probabilities.
  1. New General Bounds: They derive new "one-shot" achievability bounds that work for any random coding scheme under maximal power constraints.
  2. Minimax Optimization: Finding the tightest possible bounds is shown to be equivalent to solving specific minimax and maxmin optimization problems related to the TVD between "noise-only" and "signal+noise" distributions.
  3. Normal Approximations: Simplified formulas using the Gaussian $Q$-function are provided to estimate performance without heavy simulations.

Key Insights and Ideas

  • The Power Scaling Challenge: To stay covert, Alice's transmission power $P$ must scale roughly as $1/\sqrt{n}$. The authors demonstrate that even as power drops, the Central Limit Theorem and Berry-Esseen Theorem can still be applied to find tight bounds, provided the analysis accounts for the "vanishing" signal-to-noise ratio.
  • TVD as a Robust Metric: TVD provides a more "normalized" and intuitive measure for finite-length covertness than KL divergence. It represents the "gap" Willie has to exploit to distinguish Alice's signal from background noise.
  • Sphere Hardening: The paper exploits the "sphere hardening" effect—the fact that in high-dimensional space, high-probability Gaussian vectors tend to lie on a thin shell. This helps Alice control her detection risk by ensuring her codewords don't accidentally appear "too loud" to Willie.
  • Tightness of Bounds: The authors show that their upper (converse) and lower (achievability) bounds are remarkably close, providing an accurate "envelope" for the maximum possible throughput.

Key Results

  • Throughput Characterization: Explicit formulas are provided for the maximum number of bits ($M$) that can be sent given a blocklength ($n$), a required error probability ($\epsilon$), and a covertness constraint ($\delta$).
  • Numerical Validation: Simulations show that achievable throughput is significantly lower than the theoretical $\sqrt{n}$ limit when covertness requirements are strict (e.g., $\delta = 0.1$).
  • Error Probability Impact: The variation in allowed error probability ($\epsilon$) has a relatively small impact on throughput compared to the covertness constraint ($\delta$). In low-power regimes, the fundamental power constraint required to stay hidden dominates over second-order noise effects.

Limitations and Future Work

  • Adversary Knowledge: The model assumes Willie knows the exact noise level. In reality, noise uncertainty can help Alice, which is a suggested path for future exploration.
  • Non-Gaussian Signaling: For finite $n$, the absolute "best" distribution might not be Gaussian (e.g., APSK). The search for the globally optimal finite-length distribution remains open.
  • Complex Channel Models: Extending this analysis to fading channels or MIMO systems is the next logical step for practical wireless applications.

Summary updated

Abstract

It is well known that finite blocklength analysis plays an important role in evaluating performances of communication systems in practical settings. This paper considers the achievability and converse bounds on the maximal channel coding rate (throughput) at a given blocklength and error probability in covert communication over AWGN channels. The covert constraint is given in terms of an upper bound on total variation distance (TVD) between the distributions of eavesdropped signals at an adversary with and without presence of active and legitimate communication, respectively. For the achievability, Gaussian random coding scheme is adopted for convenience in the analysis of TVD. The classical results of finite blocklength regime are not applicable in this case. By exploiting and extending canonical approaches, we first present new and more general achievability bounds for random coding schemes under maximal or average probability of error requirements. The general bounds are then applied to covert communication in AWGN channels where codewords are generated from Gaussian distribution while meeting the maximal power constraint. We further show an interesting connection between attaining tight achievability and converse bounds and solving two total variation distance based minimax and maxmin problems. The TVD constraint is analyzed under the given random coding scheme, which induces bounds on the transmission power through divergence inequalities. Further comparison is made between the new achievability bounds and existing ones derived under deterministic codebooks. Our thorough analysis thus leads us to a comprehensive characterization of the attainable throughput in covert communication over AWGN channels.
BibTeX
@article{yu2020finite,
  author = {Xinchun Yu and Shuangqin Wei and Yuan Luo},
  title = {{Finite Blocklength Analysis of Gaussian Random Coding in AWGN Channels Under Covert Constraint}},
  journal = {IEEE Transactions on Information Forensics and Security},
  year = {2020},
  doi = {10.1109/tifs.2020.3032292},
}
Figure