Physical-Layer Trust
Measuring Embodied Interaction in Distributed Robotic Systems
Dear SoTA,
Imagine if we had robot swarms for collective transport. You might be moving house with large furniture. You could hire a team of robots to carry your sofa for you. If you live in a flat, they could even bring it down the stairs. Like Uber, but for robot swarms. Maybe the robots are hired on-demand, contracted from different companies to form an ad-hoc team. Some are brand new and others a bit worse for wear. Maybe broken. If it’s hard to trust one robot, how much harder to trust five or more? To work as a team, the robots also have to “trust” each other.
This is an essential problem to solve to successfully integrate robots in the real-world. The goal: to enable teams of robots to operate in verifiably trustworthy ways. The scenario might be hypothetical but we are moving closer to this reality.
For three months at the beginning of 2026, I was lucky to participate in ARIA’s Trust Everything, Everywhere pre-programme discovery projects. In a small collaborative project with AutoDiscovery [2], we investigated the synergies between cryptography and physical-layer computation to measure trust in distributed robot systems. It was a great space to learn from the other project teams and experts coming from different backgrounds. In this letter, I share the learnings and insights we gained in the process, and some thoughts for the future.
Part 1: Swarms and distributed robot systems
As a researcher in swarm robotics based at the Technical University of Darmstadt, my interests span collective behaviour in biological and artificial systems, and complex systems across scales. I’m often solving problems in designing collective behaviours for robots to perform tasks, taking inspiration from biology.
Distributed robot systems find applications across many scenarios, including construction, transport delivery, agriculture, space exploration, and environmental monitoring [3]. They may consist of multiple different types of robots, with different roles and capabilities. A more futuristic system is a modular robot composed of components which may be attached and detached according to the desired functionality. Robot swarms are a special subgroup of distributed robot systems:
“Swarm robotics is the study of multi-robot systems that exhibit global emergent behaviours through local environmental and robot-robot interactions” [4]
The phenomena of emergent collective behaviour can be observed in biological and physical systems on a wide range of scales: think colonies of ants or bees, or even human societies. There is no central brain directing each individual in their actions and choices, and yet, the collective of individuals produces complex behaviours like foraging and nest-building, rich systems of culture and technology.
One challenge for researchers studying these systems is to understand how emergent behaviour arises from local interactions between individuals. Simple rules can produce complex behaviours. The famous Boids model is an example of this: it consists of three rules of motion for an individual agent and emulates the movement patterns of flocking of birds [5]. For swarm engineers, designing local rules to produce a target collective behaviour can be an elegant and minimalist approach. The upside is that it can be difficult to predict precisely what emerges which is a major challenge for safety verification and validation. This is one factor which can undermine trust.
The phenomenon of emergence is also interesting from the perspective of distributed computation. The swarm itself can be a substrate for “embodied computation”. The physical environment contributes to the dynamics of the system, in generating noise in sensor-action feedback loops for example. Physical signals therefore may be rich in information about the dynamics of the swarm. These are the two key ingredients driving this project – emergent behaviour and physical signals – which I will expand on next.
Part 2: Cryptography and trusted systems
What do we mean by trust in, or trustworthiness of, robots? Trust is multi-faceted: it’s composed of complex and intersecting factors. These include safety, security, proficiency, explainability, and predictability.
In this project, we focus on security and safety with the aim to understand how to measure and quantify trust more broadly by considering the physical layer of the system. By quantifying trust, we can provide guarantees about the behaviour of a system and make sure that it’s doing what we want and expect. There are two layers of trust we are interested in: trust between robots, and trust between humans and robots.
Trust in a single robot looks different to trust in a large number of robots. What’s more, distributed systems may be tightly coupled (e.g., robots explicitly coordinating movement) or loosely coupled (e.g., decentralized control in swarms). Multiple moving robots can produce effects of cognitive overload on human observers. Modelling the relationship between perceived trust and the behaviour of the system is an open problem. From a security perspective, important factors for trust are secure communication channels and authentication for robots in a team. Robot teams have various vulnerabilities such as impersonation, malware, tampering and communication jamming [6].
Towards scaling trust, we asked the question: how can we leverage the information generated by the interaction between robots (and with their environment) in order to verify trustworthiness and detect malicious behaviour? The interaction between multiple robots in distributed systems can produce non-linear, unpredictable behaviour, which could be exploited for cryptographic applications to secure the system itself.
At the same time, the characteristics of swarms, and distributed robots more generally, which may be advantageous for real-world operation (e.g. robustness), also present challenges for established cryptographic methods. For example, dynamic networks may require continuous authentication; limited onboard resources mean that lightweight security is advantageous; and methods may need to scale for very large numbers of interacting, connected robots.
Our goal is to leverage physical fingerprints as a tamper-proof and scalable measure of trust.
These fingerprints generated from the situatedness of robots in the environment, transformed through robot interactions, are difficult to manipulate by attackers. They are also difficult to reverse-engineer: it is not easy to deduce the local rules which produce a given emergent behaviour. This is promising in the context of verifying correct behaviour where a known mapping from local to global metrics can be leveraged in challenge-response protocols. The way that signals are generated means that computation comes “for free” from the physics of interaction dynamics. This is the key to potential scalability. Some observable signals come from the sensors available onboard each robot as well as internal components and processes. Possible observables include: motion (via cameras, infrared sensors), radio frequency, light, sound, forces (via tactile sensing, material deformation), and power (current, voltage).
One example of using radio frequency as a spoof-resilient signal comes from work by Gil et al. [7]. In this work, the physics of wireless signals is used to detect spoofed clients in a Sybil attack. The signals contain directional information which are used to verify the location of a client and determine if it’s legitimate or not.
In sum, physical-layer signals may be treated as trust primitives which can be used to build up more complex measures of trust, complementing established cryptographic methods.
Part 3: Towards protocols for physical-layer trust
The main goal of this project was to systematize knowledge on “Physical-layer trust”, to connect the dots between security, robot swarms, and physical-layer signals. Whilst the literature shows that physical signals have been exploited to secure such systems (e.g. Gil et al. [7]), there remains much room to develop methods which leverage the properties of uniqueness and tamper-resistance. Coming back to the emergent properties of swarms, the difficulty of reverse engineering collective behaviour can be leveraged to produce physical one-way functions [8]. The computation which occurs in swarm dynamics could therefore be exploited for cryptographic applications, to develop challenge-response protocols.
Future steps in this direction would be to investigate the robustness of identified trust primitives for a target robot platform and the generalisability of primitives across systems and scenarios. I’m excited for the continuation of this work and to find collaborators in related fields of research (any parallels to other complex systems?) to discuss and debate. If this sounds like you, feel free to drop me an email (suet [dot] lee [at] tu-darmstadt [dot] de). Thank you for reading this letter on physical-layer trust.
This project has been a collaboration with Autodiscovery: thank you to Aron Kisdi and Wayne Tubby who supported this project with their expertise on robot hardware and real-world deployment. Thank you also to colleagues at TU Darmstadt for their insights and feedback on this project. Lastly, thank you to the Trust Everything, Everywhere team at ARIA who are developing an awesome research programme, and for the opportunity to work on this pre-programme discovery project.
Yours,
Suet Lee
Suet Lee is a postdoctoral researcher in swarm robotics at the Technical University of Darmstadt, Germany. She earned a Master’s of Science in Mathematics (2015) from Imperial College London, and worked in industry in the fields of software development and robotics, before returning to academia to complete a PhD in Swarm Robotics (2025) at the University of Bristol. She also worked at the University of Konstanz as a postdoctoral researcher in swarms. Her research interests span embodied language for robot swarms and the embedding of such systems seamlessly and safely in real-world human environments; emergent interactive art; and complex systems across scales.
This letter is based on a blog post originally published here [https://autodiscovery.co.uk/aria-tee.html].
[1] https://autodiscovery.co.uk/aria-tee.html
[2] https://autodiscovery.co.uk
[3] Swarm Robotic Behaviors and Current Applications, Schranz, M. et al., 2020.
[4] Trustworthy Swarms, Wilson, J. et al., 2023.
[5] https://www.red3d.com/cwr/boids
[6] A Survey on Security of UAV Swarm Networks: Attacks and Countermeasures, Wang, C. et al., 2024.
[7] Guaranteeing Spoof-Resilient Multi-Robot Networks, Gil, S. et al., 2017.
[8] Physical One-Way Functions, Pappu, R. et al., 2002.
Write to the Society for Technological Advancement on letters@ilikethefuture.com.