Research · as of
Research
303 Overwatch A.S.B.L is a Luxembourg non-profit working on the cybersecurity and resilience of space systems: the satellites, ground stations, communication links, and user equipment that space services depend on. This page is for researchers, collaborators, funders, and publishers who want to know what we study, what we have published, and how to work with us.
Legilux
303 Overwatch A.S.B.L publishes its research positions here so that a researcher, a funder or a publisher can verify them and cite them. Accredited by the Luxembourg Ministry of National Education.
Positions as of the date above; this page changes in the same commit as the work. Research enquiries: workforce.development@303overwatch.org
Sovereign and independent research and development
Our research department exists to do three things: produce peer-reviewed research on securing space systems, carry that research into international standards so it changes how systems are built, and give independent and non-profit researchers a recognised organisation to publish under.
RCS F13274
303 Overwatch A.S.B.L is a Luxembourg non-profit association, RCS F13274 (EUID LURCSL.F13274), registered 10 May 2021 and governed under Luxembourg law.
Our research agenda is set by the association. It is not set, directed, tasked or funded by any government, armed force, intelligence service, or defence prime contractor. We hold no contract obliging us to deliver a capability to any party. That is what sovereign and independent means here: the question is ours to choose, the result is ours to publish, and we publish it, in peer-reviewed venues and in open standards work.
We do not develop weapons
We do not research, design, develop, produce, integrate, test, deliver, operate, maintain, or export space cyber weapons platforms. This holds in all three forms in which the term is used, and we define each so the commitment is testable rather than decorative.
Active. A system built to produce a commanded effect against another party's space system, its communication link, or its ground segment. We build none and hold none.
Passive. A system that produces an effect on another party's space system by presence, persistence or latency rather than by command. This covers dormant access, resident implants, and sustained interference. We build none and hold none.
Unintentional. A system whose design produces effects on another party's space system that were not intended, including uncontrolled emission, spectral spillover, and interference with shared orbital or spectrum resources. We fly no hardware and we operate no emitter. Our work runs in models.
What we do instead
We model. Every finding is produced against a digital twin, a virtual instrument, or a synthetic environment that we built ourselves. We do not test against live, third party, or client space systems. We request no sensitive system data from any operator. Adversary behaviour is modelled so that a defender can price it, and the output is an engineering requirement, a detection, a taxonomy entry, or a paper.
Research into how systems break is how systems are built not to break. The IEEE standard we edit, the taxonomy we publish, and the credentials we teach all rest on someone having worked out what an adversary could actually do. Doing that honestly means stating where the line is and staying on our side of it.
Legal position
2021/821
We operate under the Union dual-use regime, Regulation (EU) 2021/821 of 20 May 2021, and under Luxembourg export control law, the Law of 27 June 2018. The competent national authority is the Office for Export, Import and Transit Control (OCEIT) at the Ministry of the Economy. Where any output could fall within the Annex I controls, including the entries covering intrusion software and related technology, we treat it as controlled and seek authorisation before any transfer. Our default is publication, which is the opposite posture to a controlled transfer, and we design the work so that it stays that way.
OST
Art. IV
We conduct our research consistent with the peaceful purposes principle of the 1967 Outer Space Treaty, and we do not contribute to any capability prohibited under Article IV.
Keeping this honest
This statement describes what we do today. If the work changes, this page changes in the same commit as the work. If you believe anything we publish sits on the wrong side of the line described here, write to us and we will answer.
Research areas
Our work centres on multidomain converged platforms, where the orbital, spectrum, network and human layers of a space system are one attack surface and one engineering problem rather than four. We design security in from the component level up, an approach the field calls secure-by-design, and the scope runs from that engineering through to the workforce development and the tooling that teaches it. Within that, we focus on:
- Secure-by-component design for space missions. Decomposing a mission into its components and data flows, enumerating the attack surface of each, and deriving engineering requirements that prevent mission failure.
- Link segment security. Protecting the radio frequency and optical communication links between spacecraft, ground stations, and aerial platforms, including how environmental conditions and adversaries jointly degrade link confidentiality.
- Resilient heterogeneous networks. Securing multi-layer architectures that combine satellite constellations with high-altitude and low-altitude platforms for critical applications such as wildfire management.
- Minimum cybersecurity requirements. Establishing the smallest set of design principles that preserve access to space by preventing the permanent loss of control of a spacecraft.
- Software supply chain transparency for space. Applying software bill of materials practice to space systems, so operators know what runs on the systems they depend on.
- Workforce development for converged platforms. Determining what a practitioner has to be able to do against a converged space platform, and building the virtual instruments that teach it through experiential learning.
- Tooling that carries the method. Turning the research into the taxonomies, harnesses and labs an operator can run without us, published open source.
Applied research
The Center of Excellence teaches a method. This research decides what the method has to cover next.
We run four research lines against the converged platform described above. Each one states what it is, which doctrine it answers to, where it sits in the wider landscape, and what it produces. All four run in synthetic environments, and all four feed the workforce development and the tooling, which is where a finding becomes something a practitioner learns by doing. We do not touch client or third party flight hardware, and nothing on this page is offered as an operational service. The key beside each line names a joint publication in its doctrinal basis.
JP 3-14
Virtual hardware for resilient cyber operations
Why an instrument that can be broken on purpose teaches resilience, and a simulator does not.
Definition. Virtual hardware is a software model of a space system component that is faithful at the interface: the learner drives the real control surface, reads the real telemetry format, and produces a file the real tool will open.
The distinction that matters is between a simulator and a virtual instrument. A simulator reproduces an outcome. A virtual instrument reproduces a behaviour, including how the part fails. We build the second, because a model that only ever works teaches nothing about resilience.
Workforce development for space cyber cannot wait on flight hardware. A physical bench ties up an operator, costs money per seat, and cannot be handed to a hundred learners at once. A virtual bench can, and it can be broken deliberately, which a flight asset cannot.
Doctrinal basis. JP 3-12, Joint Cyberspace Operations (2022), describes cyberspace in three layers: physical network, logical network, and cyber-persona. A training bench that models only one of those teaches a third of the problem, so our instruments carry all three. JP 3-14, Joint Space Operations (2023), gives the segment structure the bench has to reproduce: the spacecraft, the link, and the ground.
Landscape we track. Persistent, reusable cyber ranges are now the standard instrument for building operational cyber forces, from national training environments through to alliance-level exercises such as NATO Cyber Coalition and the CCDCOE Locked Shields series. USSOCOM's SOFWERX operates the same idea in acquisition form, running assessment events under three lines of effort, Discover, Engage and Accelerate, to pull non-traditional capability toward SOF problems. Our position is that the space sector has the range gap those efforts closed elsewhere.
What we produce. The H3 harness, an instrumented virtual rack, and the labs and credentials that run on both. Published open source.
JP 3-12
Digital twin technology for space system exploitation
Where a platform breaks, what it takes to reach the break, and how a defender prices that.
Definition. A digital twin, in this line, is a behaviourally faithful model of one specific space system, bound to that system's stated configuration, on which vulnerability research, exploit development and command and control technique development can be carried out without touching the flying asset.
The line answers three questions in order. Where does this platform break. What does it take to reach the break. Once reached, what can actually be commanded, how would that access be sustained across contact windows, and what would give it away.
We describe this as establishing the art of the possible, and we mean it literally. The output is an upper bound on what an adversary could achieve against a given architecture, expressed so a defender can price it. It is not a delivered capability.
Adversarial command and control in space is not a generic C2 problem wearing an orbital badge. It competes with the legitimate telemetry, tracking and command path, inside a link budget, on an orbital contact schedule, against a ground segment that may be the only place anyone is watching. Those constraints are the research.
Doctrinal basis. JP 3-12 separates cyberspace exploitation from cyberspace attack, and sets out the effects vocabulary we use for twin findings: deny, in its degrade, disrupt and destroy forms, and manipulate. It also describes operational preparation of the environment, which is the activity a twin lets a defender study safely. JP 3-14 covers space superiority and space control, offensive and defensive, which is the mission language a twin finding has to translate into if an operator is going to act on it.
Landscape we track. The Aerospace Corporation's SPARTA matrix gives the sector a shared vocabulary for space system attack techniques. The Hack-A-Sat series, and the Moonlighter cubesat flown as an on-orbit testbed for it, established that adversarial research against real space systems can be done in the open and that the sector benefits when it is. Space ISAC is where findings belong. We contribute to Space ISAC and to the IEEE Space Systems Cybersecurity Working Group.
Boundaries. Twins are built from configuration the operator chooses to state. We request no sensitive system data. Findings go to the operator, and to Space ISAC where the operator agrees.
Status. Research direction.
JP 2-01.3
Environmental tradecraft modelling with METEORSTORM
Model the environment first, then derive what an adversary could do inside it without being seen.
Definition. Environmental tradecraft modeling is the practice of modeling the operational environment first, deriving what an adversary could do inside it without being seen, and then enumerating the observables that would give that adversary away.
Most threat modeling starts with the adversary and asks what they might do. This inverts it, because in space the environment does most of the constraining. Orbital geometry decides when a contact window exists at all. The link budget decides what an emitter can be and where it can sit. Ground segment topology decides which network paths physically exist. An evasion technique that ignores those is a theory about nothing.
METEORSTORM is our open framework for this, published as a MISP taxonomy. Its first layer is the Primary Capability Environment, the environment itself, and every element below it names its parent. The analytic layer anchors each threat, attack path, detection and resilience measure to exactly one enumerated element. That anchoring is the point: a theoretical evasion technique is stated against a named element in a named environment, and it is auditable because of it. A technique that cannot name what it evades is excluded on the record.
Multidomain is not decoration here. A space platform is simultaneously an orbital object, an electromagnetic emitter, a network, and a set of people with credentials. A technique that buys concealment in one of those layers routinely pays for it in another, and the framework exists to make that trade visible.
Doctrinal basis. JP 2-01.3, Joint Intelligence Preparation of the Operational Environment, is the direct ancestor: define the environment, describe its effects, evaluate the adversary, then determine adversary courses of action, in that order. JP 3-85, Joint Electromagnetic Spectrum Operations (2020), supplies the spectrum half of the environment. JP 3-0, Joint Campaigns and Operations (2022), frames the multidomain problem. Signature and emissions discipline sits with joint operations security doctrine, JP 3-13.3.
Landscape we track. SOF AT&L's science and technology work is organised around three SOF Hard Problems: Small Unit Dominance, Mission Assured Communications, and Signature Management. This line answers the third, moved into the space domain. PEO SOF Warrior states its operating environment as denied and non-permissive, with low probability of intercept and detection, and sensor defeat as standing requirements, which is the same problem an operator faces when their platform is the loudest object in the sky. On the alliance side, NATO tested Maven Smart System capabilities during Exercise STEADFAST DETERRENCE 2026 at the Joint Warfare Centre, including automated scripting of opposition force responses, which is exactly the modeled-adversary function this line builds. NATO's data labelling and metadata binding standards, STANAG 4774 and 4778, are the interoperability target for any taxonomy that wants to be shared across an alliance rather than inside one organisation.
What we produce. The METEORSTORM taxonomy and nomenclature, published open source, and the modeling method taught through the Full Spectrum credential.
JP 3-04
AI driven edge computing for on-orbit capability
What a spacecraft can decide for itself when the ground is not reachable, and what changes when it can.
Definition. Edge autonomy research asks what a spacecraft can decide for itself when the ground is not reachable, and what changes, offensively and defensively, when that decision authority moves on board.
The link is the whole constraint. A satellite has contact windows, and between them either the spacecraft decides or nothing is decided. Moving inference on board cuts both ways. It shortens the loop for anything already resident on the platform, and it removes the ground station as the single place where intent can be observed. Both of those are research findings a space operator needs before the capability arrives, not after.
We study the enabling stack in the open: constrained inference on radiation-tolerant compute, model integrity and update paths that survive an unattended contact gap, and what a bounded autonomy envelope has to look like before anyone should trust it with a command.
Doctrinal basis. JP 3-14 establishes the dependence of space control on the link segment, which is the dependence on-board autonomy is meant to break. JP 3-04, Information in Joint Operations, frames information as a joint function rather than a support activity, which is the argument for treating on-board decision quality as a mission property. JP 3-12 supplies the effects vocabulary for what resident autonomy could be made to do.
Landscape we track. NATO has now taken AI to the operational edge on accredited networks. The NATO Communications and Information Agency finalised acquisition of the Palantir Maven Smart System NATO for Allied Command Operations on 25 March 2025. NATO Maven Smart System reached full technical operational capability on 22 June 2026, and the NATO Security Accreditation Board granted it full accreditation to operate on NATO's classified network for exercises, missions and activities. Task Force Maven, the small team behind it, reports more than fifty applications built and deployed across the operational force during 2025, work on data ontology for AI models, and compliance with STANAG 4774 and 4778. The direction of travel is settled.
On the special operations side, USSOCOM has been pushing the same idea toward the individual for years. The Hyper Enabled Operator concept was framed as cognitive overmatch at the edge, where the edge is the small unit or single operator in a remote and austere environment. That work now sits alongside PEO SOF Digital Applications, stood up in 2020 to deliver SOF software and data capability, and USSOCOM's J24 Intelligence Data Science Team, which has been assessing GPU infrastructure for large language model workloads through SOFWERX.
A spacecraft between contact windows is the most austere disconnected edge node anyone operates. Our position is that the lessons are transferable in both directions, and that the space sector should be doing this research in the open before it has to do it under pressure.
Status. Research direction.
Doctrinal references
We use these as our shared vocabulary. Where our terms differ from them, we say so.
- JP 3-0, Joint Campaigns and Operations (2022)
- JP 3-04, Information in Joint Operations
- JP 3-05, Special Operations
- JP 3-12, Joint Cyberspace Operations (2022)
- JP 3-13.3, Operations Security
- JP 3-14, Joint Space Operations (2023)
- JP 3-85, Joint Electromagnetic Spectrum Operations (2020)
- JP 2-01.3, Joint Intelligence Preparation of the Operational Environment
- JP 5-0, Joint Planning
- AJP-3.20, Allied Joint Doctrine for Cyberspace Operations, Edition A Version 1 (January 2020), recorded under STANAG 6514
- STANAG 4774 and STANAG 4778, NATO confidentiality labelling and metadata binding
Affiliation and boundaries
This page names external doctrine, programmes and organisations to state where our research sits. It does not claim affiliation with any of them. 303 Overwatch A.S.B.L is a Luxembourg non-profit, RCS F13274. Where we do have a standing relationship, we name it plainly: we contribute to the Space ISAC and to the IEEE Space Systems Cybersecurity Working Group.
All four lines run against models, twins and synthetic environments that we build. We do not conduct operations, we do not test against live or third party space systems, and we request no sensitive system data from anyone. Research outputs are published, or disclosed to the operator concerned, or both.
Sovereign and independent research. Our agenda is set by the association, not by any government, armed force or defence contractor. We develop no active, passive or unintentional space cyber weapons platforms. All work runs against models, twins and synthetic environments we build ourselves, never against live or third party space systems. Full statement above.
Peer-reviewed publications
Research by 303 Overwatch is published in peer-reviewed IEEE venues. The affiliation appears in print as 303 Overwatch (A.S.B.L), Luxembourg City, Luxembourg.
- N. Benchoubane, O. Ben Yahia, W. Ferguson, G. Gür, S. Chakravarty, G. Falco, and G. Karabulut Kurt, “Securing Heterogeneous Network (HetNet) Communications for Wildfire Management: Mitigating the Effects of Adversarial and Environmental Threats,” IEEE Journal of Radio Frequency Identification, vol. 9, pp. 713 to 726, 2025. DOI: 10.1109/JRFID.2025.3601843
- O. Ben Yahia, W. Ferguson, S. Chakravarty, N. Benchoubane, G. Karabulut Kurt, G. Gür, and G. Falco, “Securing Satellite Link Segment: A Secure-by-Component Design,” 12th IEEE International Conference on Wireless for Space and Extreme Environments (WiSEE), 2024, pp. 177 to 182. DOI: 10.1109/WiSEE61249.2024.10850060. Preprint: arXiv:2411.12632
- G. Falco, N. Boschetti, A. Viswanathan, B. Bailey, C. Maple, G. Karabulut Kurt, J. Slay, E. Birrane, D. Logsdon, S. Bennett, W. Ferguson, et al., “Minimum Requirements for Space System Cybersecurity: Ensuring Cyber Access to Space,” 2024 IEEE 10th International Conference on Space Mission Challenges for Information Technology (SMC-IT), 2024, pp. 78 to 88. DOI: 10.1109/SMC-IT61443.2024.00016. Published under our founder’s earlier project affiliation.
Our co-authors on these works are at Polytechnique Montréal, Cornell University, Zurich University of Applied Sciences, Kennesaw State University, and partner institutions across government, industry, and academia.
Standards work at IEEE
P3536
Research matters most when it changes how systems are engineered. Our founder, William O. Ferguson, Senior Researcher and Primary Investigator and an IEEE Senior Member, carries our research perspective into the IEEE Standards Association:
- Technical Editor, IEEE P3536 Space Systems Cybersecurity (2023 to present). IEEE P3536 is the emerging standard for space system cybersecurity design, setting baseline security requirements for how space systems are engineered. The role has represented more than 500 hours of technical writing, consensus building, and engineering analysis, coordinating contributions from government, industry, and academic stakeholders across four continents and carrying the draft from scoping through ballot.
- Lead, Space Systems Integration Subgroup, IEEE P3536 (2026 to present). The integration layer is where ground control systems, communication links, spacecraft, user terminals, and third-party services interconnect, and where adversaries exploit trust between segments to move laterally. The subgroup specifies security requirements for these interfaces, extending to cislunar communications and commercial space stations.
- Developer, IEEE Space Systems Resilience Engineering pilot cohort. A standard changes engineering behaviour only when practitioners are trained to apply it. This IEEE SA training effort translates the standard into taught practice for working engineers, addressing a documented workforce shortage in the sector.
- Contributor, IEEE SA P3349 Space System Cybersecurity Working Group, whose minimum-requirements methodology appears in our SMC-IT publication above.
Community and professional service
- Co-Chair, Space Systems Software Bill of Materials Working Group, Space ISAC (2026 to present)
- Space ISAC member organisation; earlier service includes co-chairing the Information Sharing Working Group (2020 to 2022)
- Aerospace Contributor, Hardware Special Interest Group, MITRE (2024 to present)
- Co-Founder, Cyber Deception Village (2025 to present)
Selected presentations
- Panelist, “Digital Foundations: Securing the Software-Defined Space Enterprise,” Space ISAC Value of Space Summit VII, Colorado Springs, September 2026
- Presenter, CyberSat Summit, Washington DC area, November 2026
- “Space Collective Defense: Collective Response,” CyberSat Summit, Washington DC area, 2025
- Presenter, CYSAT, Paris, 2025
Publish with us
Part of our purpose is to give independent and non-profit researchers who lack a university affiliation a recognised, resolvable organisation to publish under. If you are researching space system security without an institutional home, we would like to hear from you.
Researchers publishing with us list the affiliation exactly as:
303 Overwatch (A.S.B.L), Luxembourg City, Luxembourg
Organisation identifiers
- Legal registration
- Registre de Commerce et des Sociétés Luxembourg, RCS F13274 (EUID LURCSL.F13274), registered 10 May 2021
- Accreditation
- Ministerial act of 12 June 2026, Mémorial B2710
- ISNI
- 0000 0005 3061 5177
- Ringgold ID
- 854039
- Websites
- 303overwatch.org, 303ow.team
Work with us
Research enquiries go to the address below.
workforce.development@303overwatch.org
An operator who wants the method taught, or an exercise run, starts with the Center of Excellence.
Talk to the Center of Excellence