Physical layer security (PLS) has long promised a way to secure wireless communications without relying entirely on upper-layer cryptographic keys. The core intuition of PLS is simple: leverage the physical properties of the wireless channel—such as fading, distance, and spatial diversity—to ensure that the legitimate receiver's channel is fundamentally better than the eavesdropper's channel.
To achieve this, transmitters use techniques like directional beamforming to focus energy toward the legitimate user, or inject Artificial Noise (AN) into the "null space" of the legitimate channel. This ensures that while the intended recipient hears a clean signal, an eavesdropper's reception is drowned out by noise.
However, conventional PLS frameworks operate under a dangerous assumption: they model eavesdroppers as isolated, independent entities. As Amal Hyadi et al. and Tiep M. Hoang et al. observe, evaluating security based only on the single most capable individual adversary severely underestimates the interception capabilities of real-world adversaries. When multiple eavesdroppers collude, the foundational assumptions of traditional physical layer security collapse.
The Mechanics of Collusion: The Virtual Antenna Array
Why does collusion change the game so drastically? The answer lies in how distributed nodes can pool their observations to act as a single, macro-scale entity.
Consider a transmitter attempting to hide a signal from several distributed, single-antenna eavesdroppers. Under traditional modeling, each eavesdropper faces an independent channel with its own noise floor and fading characteristics. If the transmitter uses spatial filtering or artificial noise, it can easily degrade the signal-to-noise ratio (SNR) at each individual node.
Traditional Assumption (Independent):
Transmitter ----> Legitimate User (Clean Signal)
\----> Eavesdropper 1 (Noisy/Nullified)
\----> Eavesdropper 2 (Noisy/Nullified)
The Reality of Collusion (Cooperative):
Transmitter ----> Legitimate User (Clean Signal)
\----> Eavesdropper 1 \ Joint Signal
\----> Eavesdropper 2 / Processing ---> Virtual High-Gain Array
When adversaries cooperate, they don't just look at their individual data streams; they share raw or processed observations. As ishmam zabir et al. explain, coordinated adversaries can pool their signals to function as a powerful virtual antenna array. By utilizing joint signal processing techniques such as Maximum Ratio Combining (MRC), they can coherently combine weak signal leakages from multiple distinct geographic locations.
This cooperative phenomenon isn't limited to traditional signal processing. In the domain of semantic and deep-learning-based communications, Seyyed Amirhossein Ameli Kalkhoran et al. demonstrate that when adversarial agents share internal prediction scores through ensemble learning techniques, their collective inference capability scales dramatically. By pooling observations, a colluding group can extract private data and sensitive attributes that would remain completely hidden from an isolated attacker.
Why Traditional Defenses Fail
The emergence of a virtual antenna array directly neutralizes traditional PLS defenses like Artificial Noise.
Normally, AN is designed to fall into the null space of the legitimate user's channel matrix, meaning it destroys the signal everywhere except at the precise spatial coordinates of the legitimate receiver. However, Xiao Chen et al. demonstrate that artificial noise matrices frequently contain "blind spots"—spatial regions where the noise power is naturally attenuated. If multiple eavesdroppers collude and share information, they can mathematically isolate these blind spots, performing joint spatial filtering to cancel out the artificial noise entirely and reconstruct the secret message.
Consequently, when multiple adversaries coordinate their efforts using techniques like MRC, they can bypass security protocols designed for isolated interceptors, as highlighted by Jiahui Li et al..
Designing Defenses for the Worst-Case Group
To counter cooperative eavesdropping, the paradigm must shift from defending against the best individual attacker to defending against the worst-case collective entity.
Zhu Han explored this adversarial dynamic through coalitional game theory, modeling how eavesdroppers self-organize into cooperative groups to maximize their interception rates. To defend against these self-organizing coalitions, network designers must adopt a highly conservative security baseline.
1. Collective Channel Modeling
Instead of optimizing beamforming vectors against individual adversarial channels, modern frameworks treat the entire network of distributed eavesdroppers as a single, massive multi-antenna receiver. For instance, Hongjiang Lei et al. propose a joint trajectory and power optimization framework that explicitly models distributed eavesdroppers as a unified cooperative entity. By maximizing the worst-case average secrecy rate against this aggregated threat, the system guarantees robust protection even if the attackers achieve perfect coordination.
2. Exploiting Near-Field Spherical Waves
As wireless systems move toward higher frequencies (like millimeter-wave and Sub-6GHz) and employ larger antenna arrays, the communication landscape shifts from the far-field to the near-field. In the far-field, waves are modeled as planar, meaning an eavesdropper only needs to match the angle of the transmission to intercept it. In the near-field, however, electromagnetic waves propagate as spherical wavefronts.
Far-Field (Planar): Eavesdropper anywhere along the angular path can intercept.
Transmitter =====> Angle θ =====> [Eavesdropper 1] =====> [Eavesdropper 2]
Near-Field (Spherical): Signal focuses on a specific 3D coordinate (Distance + Angle).
Transmitter ( ( ( Focus Point: Legitimate User ) ) ) [Eavesdropper is safe outside focus]
Jingjing Zhao et al. and Yiming Xu 0009 et al. demonstrate how to exploit this property using Integrated Sensing and Communications (ISAC). By using radar-like sensing to track the 3D trajectories of non-cooperative targets, the transmitter can use near-field beamforming to focus the data signal precisely at the specific distance and angle of the legitimate user. Because the energy rapidly defocuses outside this exact 3D patch, even a colluding network of distributed eavesdroppers cannot gather enough coherent signal energy to reconstruct the transmission.
Key Foundational Work
Secrecy capacity of a class of orthogonal relay eavesdropper channels (2008) — contributed 1 concept elements to this direction.