Most security frameworks are built on a comforting but increasingly fragile assumption: the "lone wolf" adversary. Whether modeling a jammer in a wireless network or a spoofer targeting a drone, we typically assume a single attacker acting in isolation.
However, a growing body of research across wireless communications, robotics, and game theory suggests that this assumption creates a dangerous "security gap." When multiple adversaries coordinate their actions or pool their data—a phenomenon known as collusion—they don't just become twice as strong; they fundamentally change the geometry of the problem, neutralizing traditional defenses like beamforming, spatial filtering, and outlier detection.
The Power of Collusion: Virtual Antenna Arrays
In physical layer security (PLS), the goal is to ensure that the legitimate receiver gets a clear signal while the eavesdropper (often called Eve) receives only noise. This is usually achieved through beamforming—shaping the signal so it "nulls out" in the direction of the eavesdropper.
As ishmam zabir and Gaojie Chen demonstrate in Secrecy of Multi-Antenna Transmission With Full-Duplex User in the Presence of Randomly Located Eavesdroppers, this defense crumbles when eavesdroppers collude. By sharing their intercepted signals and processing them jointly, a group of spatially distributed, single-antenna eavesdroppers can function as a powerful virtual antenna array.
This collusion allows them to:
- Aggregate Signal Power: They can coherently combine weak signals that would be indecipherable to any single node.
- Perform Spatial Filtering: They can "steer" their collective attention to bypass artificial noise injected by the transmitter, as noted by Arthur S. de Sena and Jiguang He in their work on Holographic MIMO Networks.
Hongjiang Lei and colleagues argue in Trajectory and power design for aerial CRNs with colluding eavesdroppers that treating a distributed network of attackers as a single cooperative entity is the only way to provide a truly "conservative and robust" security baseline.
Coordination in Physical Space: Overwhelming the Sensors
While wireless collusion focuses on information sharing, coordination in robotics and autonomous vehicles focuses on simultaneous physical disruption.
In Drones in Distress: A Game-Theoretic Countermeasure for Protecting UAVs Against GPS Spoofing, AbdelRahman Eldosouky and Walid Saad highlight a critical vulnerability: most UAV outlier detection systems can handle one fake GPS signal, but they struggle when multiple "coordinated" spoofers target a fleet simultaneously. If attackers synchronize their fake signals, they can make the "outlier" look like the "truth," leading the vehicle's control system to make catastrophic corrections.
This shift from single to multi-attacker models is also vital for ground vehicles. Katherine Rose Driggs-Campbell and Mykel Kochenderfer show in Improved Robustness and Safety for Autonomous Vehicle Control with Adversarial Reinforcement Learning that training agents against a single adversary is insufficient. To bridge the "sim-to-real" gap, agents must be trained in "semi-competitive" games that simulate the complex, multi-actor disturbances found in real-world traffic.
The Game Theory of Shared Defense
When multiple attackers coordinate, the defense often becomes a Multi-Defender Security Game. In these scenarios, the challenge isn't just the attacker—it's the lack of coordination among the protectors.
Jian Lou 0002 and Yevgeniy Vorobeychik explore this in Multidefender Security Games. They identify a "negative externality": if one agency protects its asset well, it might simply push a coordinated attacker toward a neighbor's less-protected asset. This strategic interaction creates a "Price of Anarchy" where independent defenders over-invest in selfish security but leave the overall system vulnerable to diverse, concurrent threats.
Emerging Frontiers: Moving the Goalposts
To counter coordinated threats, researchers are moving toward dynamic, reconfigurable defenses that "break" the attacker's ability to coordinate.
- Movable Antennas (MA): Instead of fixed arrays, researchers like Jingze Ding are using Movable Antennas to dynamically reshape the channel. By constantly changing the physical position of antennas, the defender makes it impossible for colluding eavesdroppers to maintain a stable "virtual array."
- Integrated Sensing and Communication (ISAC): Yiming Xu proposes Adaptive Sensing to actively track potential eavesdroppers. If you can sense where the attackers are, you can "focus" signals away from them with surgical precision, even if they are trying to pool their measurements.
- Rate-Splitting Multiple Access (RSMA): In RSMA-Enabled Covert Communications, Xinyue Pei and Xingwei Wang use complex signal splitting to hide transmissions from a "cloud" of randomly located wardens, recognizing that invisibility is the only defense when observers share data.
The transition from single-actor to coordinated-adversary models represents a maturation of the field. By acknowledging that "the bad guys talk to each other," researchers are developing systems that are not just theoretically secure, but practically robust against the complexities of a networked world.