Tenant isolation and secure authority in hosted media infrastructure

cRUD Research TeamThe cRUD Company

Whitepaper designed for the web

Hosted media infrastructure handles powerful operations. It can create rooms, publish sources, compose outputs, record sessions and grant access to stored media. A secure design cannot treat possession of a public identifier or client token as broad authority. This whitepaper defines how tenant isolation, environment context, minimum capability and evidence shape the public security boundary of cRUD Morph.

Executive summary

  • Every resource belongs to one tenant and one environment established by the server.
  • Clients receive time-limited capability for an approved operation, not general provider or tenant authority.
  • Public identifiers do not grant access, and provider secrets remain protected references.
  • Media, state, recordings and replay follow the customer product's entitlement decision.
  • Audit evidence records material authority and access changes without copying secret values.

The security boundary follows fact ownership

The customer product owns customer users, business roles, registration, admission and entitlement. cRUD Morph owns media execution resources and their operating evidence. Security follows that boundary. The customer product authorises an outcome, and cRUD Morph grants only the media capability required to perform it.

A provider role or room membership cannot create a customer role. A media token cannot establish payment or consent. Keeping those facts with their authoritative owner reduces the chance that a technical credential becomes a business shortcut.

Tenant and environment context

Every cRUD Morph resource belongs to one tenant and one environment. Server-side context selects that scope before an operation reaches provider logic. A browser cannot cross the boundary by changing a resource identifier or environment parameter.

Mappings, provider credentials, webhooks, objects and evidence inherit the same scope. Sandbox, staging and live resources remain separate. Promotion creates reviewed configuration in the target environment rather than allowing lower-environment authority to continue operating.

Minimum capability, not broad role

A client needs only the capability required for its current approved action. A speaker may publish an authorised source. A viewer may receive selected media and state. A producer may request a composition change. None of them needs general provider credentials or unrestricted tenant access.

Capability is time-limited and bound to context. It can include resource, role, action and expiry. Revocation and policy change are server decisions. The client can request and observe, but it cannot extend its own authority by retaining an old token.

Public identity is not permission

Stable resource identifiers help products and operators correlate work. They should be safe to reference without becoming bearer credentials. Access checks still evaluate tenant, environment, capability and current policy.

This distinction supports links, logs and support conversations without relying on secrecy of identifiers. Guessing or discovering an identifier does not grant access to a session, recording, presentation asset or replay.

Secret references and provider credentials

Provider API keys, signing material and storage credentials remain inside the approved operating boundary. Public contracts and customer-facing evidence use secret references and versions, never values. Logs and errors should avoid copying secrets or complete sensitive payloads.

Rotation is an authorised transition with validation, activation, evidence and retirement. Ownership follows cRUD BYOP, cRUD MIP or the approved hybrid component. The model states who can rotate, who responds to compromise and which environment is affected.

Signed media and state access

Delivery authority should expire and remain specific to the approved resource and purpose. A viewer entitled to one replay does not receive general object-store access. A speaker authorised for one session does not retain permanent publication authority.

The customer product makes the entitlement decision. cRUD Morph issues or uses the bounded authority required for delivery. Expiry, revocation and replay policy remain explicit. Cached or prefetched assets do not override current permission.

Webhooks and event consumers

Webhook destinations need authentication, integrity, replay protection and an allowlisted purpose. Events are delivered at least once and carry stable identity so that a consumer can process duplicates safely. A webhook does not grant permission to call unrelated media operations.

Consumers should validate the event and retrieve current state when needed rather than trusting a long-lived payload as complete truth. Changes to destinations and signing references are material configuration events and should be auditable.

Audit without overcollection

Security evidence should identify material authorisation, access, configuration and operating changes. It needs tenant, environment, actor or service identity, action, resource, outcome and time. It does not need to store secret values or unrelated customer content.

Retention follows legal, privacy, security and operating purpose. High-volume telemetry and durable audit history can have different lifecycles. Access to audit evidence is itself an authorised and recorded action.

Threat and failure tests

Attempt cross-tenant identifiers, expired capability, wrong environment, repeated webhook, revoked participant, replaced secret reference and replay access after entitlement ends. Test provider errors to ensure that secret or internal type information does not enter the public response.

Also test recovery. A route or credential change should not create a second authority or expand scope. An unknown provider outcome should remain unknown until reconciled. Security includes safe failure, not only successful access control.

Governance through the security lifecycle

Security begins before a customer environment is activated. The technical engagement identifies tenant boundaries, environments, ownership model, approved capabilities, provider connections, data regions, retention purposes and support access. The resulting configuration should be reviewable as one operating boundary rather than a collection of unrelated secrets and dashboard roles.

Routine change follows the same model. A new webhook destination, replay audience, provider credential reference or operator privilege can change the exposure of the system. Each change needs an authorised actor, reason, bounded scope, effective time and durable outcome. Emergency access needs stricter expiry and subsequent review, not a permanent role created for convenience.

Detection should connect security signals to the media operation they affect. Repeated denied access, an unexpected environment, abnormal signed-delivery use, callback verification failures or unusual operator activity should identify the tenant, resource and operation without copying private customer content into the alert. The response can then revoke or narrow authority while preserving evidence.

Recovery includes proving that the boundary is restored. Rotating a credential is not enough if old authority remains accepted. Removing an operator is incomplete if active sessions still carry broader capability. Tests should verify expiry, revocation, cache behaviour, event consumers and retained audit access. The final record should explain the exposure window and the actions that closed it.

Customer-facing security statements must match this evidence. Architecture principles do not equal certification, and an internal test does not equal an independent assessment. Assure can make controls and outcomes explainable, while contractual commitments and certifications remain separately confirmed for the approved service scope.

A periodic access review should cover human operators, service identities, webhook consumers and provider connections. The reviewer needs purpose, current scope, last relevant use, expiry and owner. Dormant authority should be removed, not merely noted. Review results and remediation become durable evidence without revealing credential values.

Data lifecycle belongs in the same review. Operational telemetry, security audit, customer media and replay assets serve different purposes and can require different retention. Deletion and legal-hold behaviour should follow the approved responsibility model. The system must be able to show the outcome without retaining content merely to prove that deletion occurred.

Security review should include customer exit. New authority must stop, active authority must expire or be revoked, customer media must follow the approved disposition and necessary evidence must remain access controlled. A clear completion record prevents an ended commercial relationship from leaving unexplained operational access.

Named ownership, completion evidence and a final access review make that transition testable for both the customer and the service operator.

Conclusion

Secure hosted media infrastructure gives each operation enough authority and no more. Tenant and environment context, minimum capability, non-secret identifiers, protected credentials, bounded access and durable audit keep powerful media operations inside a reviewable product boundary. This model protects customer authority while cRUD Morph keeps its implementation private.

Basis and methodology

This whitepaper describes the public security principles of cRUD Morph. It is not a certification, penetration-test report or promise of a particular compliance status. Exact controls, retention, providers and contractual commitments are confirmed for the approved engagement.

For the public product boundary, review the cRUD Morph approach, capabilities and documentation.

Frequently asked questions

How does cRUD Morph separate tenants?
Every resource and operation belongs to server-established tenant and environment context, and capability is checked within that scope.
Does knowing a cRUD Morph resource identifier grant access?
No. Public identifiers are references, not bearer credentials. Access still requires current approved authority.
Do customers receive provider credentials?
Credential ownership follows cRUD BYOP, cRUD MIP or the approved hybrid model. Secret values are not exposed through public cRUD Morph contracts.
Is this whitepaper a security certification?
No. It explains public design principles. Certifications, assessments and contractual security commitments must be confirmed separately.

About this whitepaper. It explains the public product and operating principles of cRUD Morph. It does not disclose source code, private implementation, provider credentials or customer data.

Put your media architecture questions in writing.

Tell us what your product delivers and which media responsibility you want to evaluate.