Security teams should treat confidential computing as a way to process sensitive data in a protected enclave, not as a replacement for encryption. Data remains encrypted in transit and at rest, then is decrypted only inside isolated hardware-backed memory. The key controls are attestation, restricted operator access, transparent code publication, and trusted client communication so the cloud can compute without gaining broad data visibility.
What confidential computing changes in a cloud encryption model
Confidential computing adds a protected execution layer between the cloud operator and the data being processed. The practical shift is that security teams can keep ciphertext protected in transit and at rest, then permit decryption only inside a hardware-backed trusted execution environment. That means the cloud can execute code on sensitive inputs without granting the cloud platform broad visibility into the underlying plaintext.
The important boundary is that confidential computing protects data during use, but it does not eliminate the need for strong encryption design. The trust model still depends on attestation proving the enclave is the expected workload, key delivery being tightly controlled, and the application being written so only the minimum necessary plaintext exists inside the protected boundary.
Transparent code publication matters because the client side has to know what code will receive the decrypted data. If the workload is opaque, attestation becomes weaker as a trust signal and customers have less basis for deciding whether to release keys or sensitive inputs. That is why confidential computing is best treated as a trust reduction mechanism, not a blanket permission to relax end-to-end encryption expectations.
Operational controls that preserve end-to-end encryption
Teams should separate cryptographic protection from operational access. The cloud service should not receive decryption material by default, operator access should be tightly constrained, and trusted client communication should be explicit about what the enclave can and cannot see. When the data owner can verify the workload identity through attestation, keys can be released just-in-time for a very specific processing event rather than being broadly available.
Key management is usually the hard part. If keys, policy decisions, or client trust checks are loosely implemented, confidential computing can become a convenience layer that still exposes data to administrators, downstream services, or code paths that were never meant to handle plaintext. The design goal is to keep ciphertext outside the trust boundary except for the shortest possible processing window inside isolated hardware-backed memory.
- Use attestation as a gate before any key release or sensitive payload transmission.
- Keep the processing code auditable so the data owner can validate the runtime trust target.
- Limit operator, debugger, and maintenance access paths that could bypass the enclave model.
- Minimise what the application decrypts, processes, or logs inside the trusted boundary.
For cloud control mapping, the relevant expectations align with the CSA Cloud Controls Matrix, which covers cloud IAM, data security, and governance, and with ISO/IEC 27001:2022 Information Security Management, which supports access control, authentication, cryptography, and cloud security requirements. At the implementation level, teams can also use NIST SP 800-57 Key Management to structure key lifecycle and cryptoperiod decisions around enclave use.
Risk and Threat Considerations
Confidential computing reduces exposure, but it also creates a new trust concentration in attestation, enclave code integrity, and key-release policy. If those controls are weak, organisations can end up with a system that still exposes plaintext during use, only now behind a more complex trust boundary that is harder to inspect.
Failure mechanism: A malicious or poorly reviewed workload, weak attestation policy, or overbroad operator privilege can cause keys to be released to the wrong runtime, or allow sensitive data to be handled outside the intended protected boundary.
Impact: The cloud provider may still be unable to see the plaintext directly, but a compromised runtime, misconfigured trust chain, or excessive administrative access can defeat the practical confidentiality goal and undermine end-to-end protection.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
CIS Controls v8, NIST CSF 2.0 and NIST Zero Trust (SP 800-207) set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| CIS Controls v8 | 6 — Access Control Management | Restricts who can administer enclaves and key-release paths. |
| 3 — Data Protection | Covers protecting sensitive data in transit, at rest, and during use. | |
| 4 — Secure Configuration of Enterprise Assets and Software | Applies to attestation, workload hardening, and trustworthy enclave configuration. | |
| Recommendation — Limit administrative access to enclave and key-management paths to approved, least-privilege roles. Protect sensitive data end-to-end and minimise plaintext exposure inside trusted execution paths. Harden enclave workloads and lock configurations to the exact measured runtime you trust. | ||
| NIST CSF 2.0 | PR.AC — Identity Management, Authentication and Access Control | Attestation-gated key release depends on strict access decisions and trusted identities. |
| PR.DS — Data Security | Confidential computing is a data-security control for protecting data in use. | |
| GV.RM — Risk Management Strategy | The trust shift to attestation and enclave integrity is a governance risk decision. | |
| Recommendation — Bind key release and sensitive processing to authenticated, authorised enclave identities. Use hardware-backed isolation to keep sensitive data encrypted except inside the approved enclave. Define when confidential computing is sufficient and when additional controls are required. | ||
| NIST Zero Trust (SP 800-207) | SC-2 — Device Authentication | Attestation serves as a trust signal for the enclave runtime before secrets are released. |
| SC-7 — Continuous Verification and Least Privilege | Supports limiting access to the minimum processing authority inside the trusted boundary. | |
| Recommendation — Require runtime attestation before issuing decryption material to the enclave. Continuously verify enclave state and keep processing privileges narrowly scoped. | ||
Practitioner Guidance
What to verify: Verify that attestation is tied to an exact build, exact measurement, and exact policy for key release. If the same enclave identity can be reused across materially different workloads, treat the trust design as too loose for sensitive processing.
Common mistake: Teams often assume confidential computing makes encryption optional during processing. It does not. The right question is whether plaintext is exposed only to the smallest possible code path, for the shortest possible time, under a trust decision the client can independently validate.
Decision rule: If a workload cannot prove its runtime state to the data owner, do not release production keys into it. If it can prove state but the code is not transparent or tightly governed, reduce the data scope before expanding the trust boundary.
Practitioner takeaway: Use confidential computing to narrow who can see data in use, not to broaden who can be trusted by default; the security gain comes from attested, minimal, time-bounded plaintext exposure.
Related resources from NHI Mgmt Group
- How should security teams implement data encryption alongside data loss prevention in cloud and SaaS environments?
- How should security teams implement cloud data loss prevention in Google Cloud environments without losing control of sensitive data elsewhere?
- How should security teams secure AI agents in private cloud and hybrid environments without weakening control boundaries?
- How should security teams onboard code analysis for GitHub Enterprise Cloud data residency environments without creating extra operational drag?
Deepen Your Knowledge
Reviewed and updated by the NHIMG editorial team on September 18, 2026.
NHI Mgmt Group — the #1 independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org
