Confidential Computing: Protecting Data While It Is In Use

The third state of data

Security guidance has long covered two states of data: at rest, protected by disk and database encryption, and in transit, protected by TLS. There is a third state that traditional controls do not address. To process data, a system generally has to decrypt it into memory, where it sits in plaintext and is readable by anything with sufficient privilege on that machine.

What confidential computing does

Confidential computing uses hardware-based trusted execution environments, sometimes called enclaves, to process data in a region of memory that is encrypted and isolated. Even a privileged administrator on the host, or the cloud provider operating the hardware, cannot read what is inside.

A related capability is attestation: cryptographically verifying that the code running inside the enclave is what you expect, and that it is running on genuine trusted hardware.

Where it earns its complexity

• Regulated data processed on infrastructure you do not physically control
• Multi-party analysis where organisations want a combined result without exposing their inputs to each other
• Handling cryptographic keys and secrets in ways that resist host-level compromise
• AI workloads processing sensitive material, where model inputs would otherwise be readable in memory

Hardware roots of trust

Confidential computing sits within a broader category of hardware security. Secure boot verifies that a system starts with untampered firmware and operating system. Trusted platform modules provide tamper-resistant key storage. The common principle is that software protections rest on assumptions about the layer beneath them, and hardware roots of trust are what make those assumptions defensible.

For most organisations this is not an immediate priority. It becomes relevant when regulatory obligations or commercial sensitivity mean trusting the infrastructure operator is not acceptable.

Related in the Knowledge Base

• Cloud and Hybrid Security: Understanding Shared Responsibility
• Secret Management and Container Security: The Basics
• AI Governance and Shadow AI: Two Problems That Share a Solution