Abstract
This paper investigates the problem of covert communications in a heterogeneous wireless network where multiple communication modalities are used simultaneously. In this setup, a legitimate transmitter sends confidential data to its receiver by selecting multiple modalities with the goal of maximizing communication covertness against a passive adversary (Willie) while satisfying a transmission rate requirement. We analyze two distinct scenarios for a given observation time by Willie: (i) Willie knows the modalities selected, and (ii) Willie is unaware of the selected modalities. We derive the optimal detector for Willie and provide a computationally efficient Detection Error Probability (DEP) approximation using two-moment matching. Building on this, we propose a novel low-complexity modality set selection technique ($O(M^2)$) designed to maximize the DEP subject to a rate constraint. Numerical simulations validate the derived analytical expressions and demonstrate that the proposed technique achieves near-optimal performance.
Introduction
In high-stakes tactical environments, traditional encryption is often insufficient. The mere detection of a transmission can reveal a unit's position or intent. Covert communications, or Low Probability of Detection (LPD), aim to hide the existence of the communication signal itself. While prior work has focused on single-modality switching, this research, conducted by researchers at the U.S. Army Combat Capabilities Development Command (DEVCOM), explores the simultaneous use of multiple frequency bands—ranging from low-VHF to millimeter-wave—to enhance stealth. By distributing a signal across multiple modalities, a transmitter can maintain high data rates while significantly complicating an adversary's detection task.
System Model and Detection Analysis
The system consists of a transmitter (Alice), a legitimate receiver (Bob), and a passive adversary (Willie). Alice selects a subset of $M$ available modalities to transmit her message. Willie monitors the spectrum to decide whether Alice is transmitting (hypothesis $H_1$) or not (hypothesis $H_0$).
The authors analyze two primary scenarios: the Informed Adversary, who knows exactly which frequency bands are being used, and the Uninformed Adversary, who knows only the potential set of bands and must monitor the entire spectrum. Willie's optimal strategy is shown to be a Likelihood Ratio Test (LRT). However, calculating the exact DEP is computationally intensive, involving complex numerical integration. To address this, the researchers propose a "two-moment matching" method, approximating the test statistics with a Gamma distribution. This provides a closed-form approximation that is both fast and accurate for system design.
Modality Selection Optimization
Selecting the optimal subset of modalities to maximize covertness while meeting a target data rate $U_{\text{target}}$ is a combinatorial problem with $2^M$ possibilities. To avoid the exponential complexity of an exhaustive search, the authors frame this as a variation of the Knapsack problem.
They introduce a greedy algorithm that ranks modalities based on a "rate-to-cost efficiency" metric, $\Psi_m$. This metric balances the contribution to the total data rate against the "cost" in terms of increased detectability. The algorithm operates with a complexity of $O(M^2)$, making it suitable for real-time applications where the true optimal exhaustive search ($O(2^M)$) would be prohibitive.
Results and Discussion
The proposed framework was evaluated using 10 modalities with center frequencies ranging from 300 MHz to 3 GHz. The results demonstrate that the Gamma-based DEP approximation is nearly identical to Monte Carlo simulations, validating its use for rapid system optimization.
Key findings from the analysis include:
- Uncertainty as a Force Multiplier: When Willie is uninformed about the active modalities, his detection performance drops significantly. He is forced to integrate noise from inactive bands, which "dilutes" the signal energy he is attempting to find.
- Frequency Trade-offs: Higher frequency bands (e.g., mmWave) offer better covertness due to higher path loss but provide lower rates at long distances. The optimal strategy typically involves a "strategic cocktail" of several high-frequency bands rather than a single low-frequency band.
- Near-Optimal Efficiency: The greedy selection algorithm consistently matches the performance of an exhaustive search while maintaining low computational overhead, even when only statistical knowledge of the adversary's channel is available.
Conclusion
This paper provides a robust mathematical foundation for multi-modal covert communications. By leveraging simultaneous transmissions across diverse frequency bands and employing efficient optimization algorithms, wireless systems can achieve high-throughput communication with a low probability of detection. Future research directions include expanding the model to multi-hop networks and considering scenarios with multiple coordinated adversaries.
Summary updated