Traditional wireless security relies almost entirely on cryptography. By encrypting a message, a transmitter ensures that even if an adversary intercepts the signal, the content remains unreadable. However, in high-stakes or adversarial environments, protecting the content of a message is often insufficient. The mere detection of a wireless signal can reveal a user's presence, trigger physical targeting, or expose a network to sophisticated traffic analysis.
To mitigate these risks, a growing body of research focuses on covert communication (or low probability of detection signaling). The goal shifts from making a message unreadable to making the transmission itself statistically invisible.
The Fundamental Tension: Hypothesis Testing and the Square-Root Law
To understand why hiding a signal is difficult, consider the perspective of an adversarial observer, often called a warden. The warden's objective is to solve a binary hypothesis testing problem based on the energy collected from the environment over $n$ time slots:
- $H_0$ (Null Hypothesis): The observed signal consists only of background noise.
- $H_1$ (Alternative Hypothesis): The observed signal consists of the covert transmission plus background noise.
If the warden knows the exact power of the background noise ($\sigma^2$), they can average their observations over $n$ samples. According to the Central Limit Theorem, the variance of the warden's noise estimation decreases at a rate of $1/\sqrt{n}$.
If the transmitter sends a signal with a constant power $P$, the total received power under $H_1$ becomes $\sigma^2 + P$. As $n$ grows large, the warden's measurement uncertainty shrinks below $P$, allowing them to distinguish between $H_0$ and $H_1$ with near-perfect accuracy.
To prevent detection, the transmitter must ensure that the signal power scales down as the number of channel uses increases—specifically, $P \propto 1/\sqrt{n}$. Because the total number of bits transmitted is proportional to $nP$, the maximum amount of information that can be sent covertly over $n$ channel uses scales as $O(\sqrt{n})$. This boundary is known as the Square-Root Law of covert communication. As demonstrated by Qiaosheng Zhang et al. in Computationally Efficient Deniable Communication, achieving this optimal scaling requires specialized, low-weight coding structures to ensure that transmissions remain mathematically tractable yet indistinguishable from pure noise.
Breaking the Square-Root Law via Noise Uncertainty
The $O(\sqrt{n})$ limit is a harsh constraint; as $n$ approaches infinity, the effective transmission rate ($O(\sqrt{n})/n$) drops to zero. To achieve a constant positive rate ($O(n)$ bits over $n$ channel uses), the transmitter must exploit or actively introduce noise uncertainty at the warden.
If the warden does not know the exact value of $\sigma^2$—perhaps due to environmental fluctuations, fading channels, or uncoordinated interference—they cannot set a clean threshold for their hypothesis test. If the transmitter keeps the signal power $P$ below the warden's uncertainty window, the warden cannot definitively prove whether a power spike is an active transmission or simply a natural fluctuation of the background noise.
Researchers have developed several mechanisms to artificially inject this vital uncertainty:
1. Friendly Jamming and Artificial Noise
Rather than relying on ambient environmental noise, a network can actively generate interference to blind the warden. For example, Hongjiang Lei et al. in Trajectory and Power Design for Aerial Multi-User Covert Communications utilize a dual-UAV system where one drone transmits data while a second "friendly jammer" drone dynamically alters its power output to mask the primary signal.
In modern Integrated Sensing and Communications (ISAC) systems, this concept is extended to Dual-Functional Artificial Noise (DFAN). As shown by Runzhe Tang et al. in Dual-Functional Artificial Noise (DFAN) Aided Robust Covert Communications in Integrated Sensing and Communications, a base station can emit a structured noise signal that simultaneously acts as a radar probe to track targets and as an obfuscation blanket to hide a concurrent data stream from a warden.
2. Cover Traffic
Another approach is to hide a secret signal inside an existing, legitimate transmission. By embedding a low-power covert stream within the frequency and time slots of a standard user, the transmitter forces the warden to try to separate the two. As analyzed by Shuai Ma 0002 et al. in Robust Beamforming Design for Covert Communications, this "cover user" strategy provides a permanent baseline of legitimate power variations, enabling a positive constant transmission rate without alerting the warden.
Exploiting Spatial Degrees of Freedom
Wireless channels are inherently spatial. By controlling where energy is directed, transmitters can minimize the signal footprint reaching the warden's physical location.
[Transmitter] ====(High Power Beam)====> [Legitimate Receiver]
\\
\\ (Attenuated/Phase-Shifted)
\\
v
[Warden] <-- Sees only noise-like fluctuations
Early work in this domain focused on directional antennas. Xiaofeng Lu et al. in Minimizing Detection Probability Routing in Ad Hoc Networks Using Directional Antennas demonstrated that by modeling the geometric detection zones of potential interceptors, routing protocols can steer signals through paths that minimize exposure to hostile observers.
Modern architectures leverage more granular spatial controls:
- Intelligent Reflecting Surfaces (IRS): By deploying surfaces capable of altering the phase of reflecting waves, operators can intentionally distort the signal path. Xinying Chen et al. in UAV-IRS Assisted Covert Communication: Introducing Uncertainty via Phase Shifting demonstrate that rapid, pseudo-random phase shifts introduced by an IRS create artificial channel fading specifically at the warden's location, destroying their ability to establish a reliable baseline for hypothesis testing.
- Flexible Antenna Placement: Beyond phase shifting, physically manipulating antenna geometry can optimize null-steering toward adversaries. The authors of Pinching-Antenna System (PASS)-Enhanced Covert Communications show that dynamically adjusting antenna positions creates highly localized destructive interference zones, ensuring that even if a warden has a significant noise-figure advantage, the received signal power drops effectively to zero.
Network-Wide Scaling and Routing
When moving from a single transmitter-receiver pair to a large-scale ad hoc or multi-user network, covert communication introduces severe throughput trade-offs.
In a broad network, capacity-boosting techniques like virtual MIMO or hierarchical cooperation naturally create bursty, high-power signal signatures that wardens can easily spot. To counter this, Kang-Hee Cho in Throughput Scaling of Covert Communication over Wireless Adhoc Networks proves that networks must enforce "preservation regions" around known or suspected warden locations. This spatial smoothing forces a modification of classic network scaling laws, constraining the overall throughput to protect node identities.
Furthermore, because network environments are dynamic and heterogeneous, finding a covert path cannot rely on static routing. To address this, Han-Bae Kong et al. in Covert Routing in Heterogeneous Networks propose decentralized reinforcement learning frameworks (such as Q-covert). These algorithms allow independent nodes to collaboratively discover multi-hop paths and select optimal transmission modalities (e.g., balancing RF, acoustic, or optical links) that minimize the global detection probability across the entire network topology.
Ultimately, covert communication works by shifting the security paradigm from a mathematical barrier (decryption hardness) to a statistical one (indistinguishability). By combining information-theoretic limits with spatial engineering—such as artificial noise, beamforming, and phase manipulation—modern wireless systems can successfully transmit data right under an adversary's nose without ever triggering an alarm.
Key Foundational Work
Communication Through Jamming Over a Slotted ALOHA Channel (2008) — contributed 1 concept elements to this direction.
Minimizing Detection Probability Routing in Ad Hoc Networks Using Directional Antennas (2009) — contributed 1 concept elements to this direction.