Jorge Quitério Search
IT & Engineering

Digital sovereignty and the role of verification

Europe cannot build every component of its critical infrastructure. It can, however, ensure that it is able to inspect, assure and replace every component it depends on. A dependency simulation shows why verification is the most effective lever.

Digital sovereignty is one of the most-used phrases in European policy, and one of the least defined. It is invoked for industrial policy, data residency, cloud procurement and space programmes. The debates often assume that sovereignty means ownership, and that the goal is to make everything in Europe.

For complex infrastructure that goal is unattainable and, as this article argues, unnecessary. A modern system contains thousands of components from hundreds of suppliers. If not every component can be owned, the question is what makes a system sovereign, and why the number of components matters so much.

The EU Cybersecurity Act (Regulation (EU) 2019/881) created a framework for European cybersecurity certification schemes. The first, EUCC (Implementing Regulation (EU) 2024/482), is based on the Common Criteria (ISO/IEC 15408) and allows products to be evaluated by accredited laboratories against published protection profiles. Software transparency is addressed by software bills of materials and reproducible builds. The arithmetic of many independent components is classical reliability theory: a series system works only if every component does.

Ownership, control and verification

  • Ownership: who holds the asset, the company or the intellectual property. It receives most public attention and changes least often.
  • Control: who can operate the system, change its configuration and stop it without asking permission from outside the jurisdiction. It depends on keys, update mechanisms and people.
  • Verification: who can establish, with evidence, what a component does and does not do. Inspecting designs and code, testing independently, certifying, detecting change.

Verification is the foundation. Ownership without verification means acquiring an asset whose behaviour must be taken on trust. Control without verification means operating a system that is not understood. Verification without full ownership still allows an informed decision to depend on a component, and to replace it when that trust is no longer justified.

A household's heating system shows the same distinction on a small scale. The owner does not need to manufacture the boiler to be in charge of the heating. What matters is that an independent engineer can inspect it, that the owner can switch it off without the manufacturer's permission, and that it can be replaced by another brand if it proves unsafe. Who owns the factory that built it makes no difference to any of these.

Requirements for verification

  • Independent evaluation: EUCC and Common Criteria, accredited labs, published protection profiles.
  • Transparency of software: source access or, at least, reproducible builds and an SBOM tying what runs to what was reviewed.
  • Open standards and interfaces, so components can be tested and replaced.
  • Cryptographic agility: the ability to change algorithms and keys without redesign, which matters now for the post-quantum migration.
  • Skills: a jurisdiction that cannot staff evaluation labs and security engineering cannot verify, whatever its laws say.

Space systems

Satellites operate for fifteen years or more and cannot be inspected once launched. Supply chains are global and ground systems run large, changing software stacks. Security must be assured before launch and maintained for decades. Europe's secure satellite communication programmes reflect this: their requirements rest less on building every part in Europe than on being able to evaluate, accredit and, if necessary, replace the parts that matter.

Dependence on many components

If a system relies on NN components and each independently carries a small probability qq of an undetected compromise, the probability that all of them are clean is

(1−q)N≈e−Nq. (1-q)^N \approx e^{-Nq} .

With 1,000 components and q=0.1%q = 0.1\%, that is only about 37%. Modern systems have far more than a thousand dependencies once libraries, firmware and services are counted. The formula leaves two levers: reduce NN, or reduce qq for each component through evaluation and assurance.

The simulation below illustrates the arithmetic. With many components, a small risk per component becomes a large risk for the system, and lowering that risk through verification moves the result far more than any realistic cut in the number of components.

  1. Default settings. 1,000 components at 0.1% each. Press "Draw 1,000 worlds": about 37% come out clean, matching the formula.
  2. Single draws. Draw one world several times. Most contain one or two compromised components, which is the situation a sovereign operator must be able to detect and fix.
  3. Verification. Verify 80% of components, cutting their risk tenfold. The share of clean worlds rises to about 76%.
  4. Fewer components. Instead, halve the number of components to 500 without verifying anything. Clean worlds rise only to about 61%. Cutting dependencies helps; verifying them helps more, and it is usually the only option.

Where the approximation comes from. (1−q)N=eNln⁡(1−q)(1-q)^N = e^{N \ln(1-q)}, and ln⁡(1−q)=−q−q2/2−⋯≈−q\ln(1-q) = -q - q^2/2 - \dots \approx -q for small qq. So (1−q)N≈e−Nq(1-q)^N \approx e^{-Nq}: what matters is the product NqNq, the expected number of compromised components. With verification of a share vv that divides the risk by ten, the exponent becomes Nq (1−0.9v)N q\,(1 - 0.9 v).

A system can be described as sovereign to the degree that those accountable for it can verify what it does, operate it without external permission, and replace any component that is no longer trustworthy. Ownership contributes to this. Verification enables the other two, and the arithmetic shows that it remains effective as systems grow.

References

  1. Regulation (EU) 2019/881 (Cybersecurity Act), Official Journal of the European Union.
  2. Commission Implementing Regulation (EU) 2024/482 establishing the European Common Criteria-based cybersecurity certification scheme (EUCC).
  3. ISO/IEC 15408, Information security, cybersecurity and privacy protection - Evaluation criteria for IT security (Common Criteria).
  4. R. E. Barlow and F. Proschan, Mathematical Theory of Reliability, Wiley, 1965.