Architectures for controlling what your systems are permitted to do.
Authority drifts
In distributed systems, authority drifts when decisions accumulate wherever logic was convenient to add. Gateways start enforcing policy. Middleware starts holding context. Coordination layers become decision points. Device signals, model outputs, logs, and ledger state start carrying more meaning than they should.
The middle of the path becomes the authority — just because it was there.
No component has to fail. No control has to be bypassed. The system works exactly as built and still permits more than anyone intended.
What this costs you
Misplaced or poorly enforced authority creates four recurring risks:
- A compromised component can exercise the authority it holds. An attacker who controls an intermediary can misuse whatever that component was trusted to decide or initiate.
- A replayed request can trigger another execution. Treating consumed permission as fresh authority makes reuse possible. Single-use permission must remain spent after it is consumed.
- Lost connectivity exposes unclear permission boundaries. A disconnected component needs to know which operations remain authorized, within what limits, and when to stop. Previously granted authority can support offline operation; loss of connectivity must not expand it.
- An interrupted operation can leave authority ambiguous. Recovery must distinguish an unfinished action from unused authority. A use already consumed must stay consumed.
Retries, outages, and restarts are part of operating distributed systems. Compromise must also be part of the threat model.
Where authority belongs
Requests may move. Artifacts may verify. Hubs may route. Devices may signal. Models may advise.
Execution authority remains provider-controlled.
The party responsible for the operation establishes the authority and its bounds. Components along the path carry requests and enforce the limits already set. Their role does not include creating, enlarging, substituting, or regenerating that authority.
Observing or forwarding a request does not create additional authority.
It starts before identity
Authentication answers who is making a request. It does not, by itself, establish what that requester is permitted to do.
The design starts by defining who can authorize the operation and where its limits are enforced. Identity and policy controls then operate within those explicitly defined authority boundaries.
Constraint is the design principle
Security doesn't always improve by adding more. Every new decision point is another place authority can migrate. Every service that interprets a request is another place meaning can change.
The less an intermediary is allowed to know, decide, or keep, the less it can do when the traffic turns hostile.
A method, not a dependency
XerØtrust provides architectural designs for implementation inside your own stack.
- No agent in the request path
- No library you have to import
- No vendor-hosted service or proprietary runtime
- No required proprietary software in your supply chain
You choose the language, libraries, and deployment environment while preserving the defined authority constraints.
Neutral Python reference: Apache 2.0 · about 1.8 KB of source · 0 third-party packages.
The reference uses only the Python standard library and is small enough to inspect directly.
The thing that constrains your architecture is itself constrained. Less to trust is less to defend.
We show the work
Claims follow the evidence.
Published architectural descriptions, reference implementations, and test packages make the work available for inspection. Each test record identifies its configuration, procedure, observations, and limitations.
Design property. For persistent single-use enforcement, authority must be durably consumed before the protected command is issued. Recovery must not restore a consumed use. An interruption can leave authority spent without a completed action.
Observed in testing. In the tested configurations, at-most-once authority consumption was preserved across reboot, power loss, and injected crashes. Missing, corrupt, and truncated state failed closed. Hostile-relay tests rejected altered authority fields and an untrusted provider signature. Valid authorities were each accepted once, and replay produced no second observed command.
Supported findings and observed implementation gaps belong in the same record. A demonstrated result applies to its documented configuration and conditions.
GitHub carries the evolving source and evidence record, including procedures for reproducing documented configurations.
Intellectual property
XerØtrust's architecture and constraint mechanisms are documented in filed patent applications. Commercial licensing agreements define the specific designs, rights, and supporting materials included. Research licenses available upon request.
What you receive
XerØtrust licenses architecture and constraint modules to organizations implementing these designs within their own systems.
The package for your selected modules includes:
- Applicable patent documentation
- Architectural diagrams and technical descriptions
- Available reference implementations
- Available test evidence, with stated limitations
Each agreement identifies the covered modules, permitted uses, supplied materials, and any included unpublished material or subsequent updates.
Your engineering team implements the design within its own stack and retains control over deployment and operation.