Privacy isn’t only about where data sits. It’s about what happens to data while it’s being processed, the “hot” moment when most systems leak. Trusted Execution Environments (TEEs) fix that moment.

A TEE is a secure section of a processor that is isolated from the rest of the machine. Code and data enter this protected enclave, run there, and stay sealed from the operating system, the hypervisor, and other apps. Even an admin on the host cannot peek inside.

How it works - with iExec

1. Prepare inputs on the user side‍

Sensitive fields are encrypted before they leave the device. Access rules specify which job may use them.

2. Verify the enclave before running‍

The job is scheduled to a worker that supports a TEE. Before anything runs, the enclave proves to iExec’s Secret Management Service (SMS) that it’s genuine and that it will execute the expected code. The SMS verifies this attestation and only then allows the job to proceed; the requester later receives a proof bound to that verified run.

3. Compute privately‍

Inside the enclave, inputs are decrypted and the computation happens. Keys and plaintext never appear outside the protected space. The OS, logs, and other processes only see opaque buffers.

4. Return results with proof‍

The enclave produces outputs and a verifiable receipt of how the run happened. You can expose only the result, keep intermediates sealed, and share the receipt so downstream services can trust the outcome without seeing raw inputs.

Benefits of protecting data during processing

For privacy

Shifting the sensitive work into a TEE reduces the attack surface at the exact moment leaks occur. You can prove integrity without sharing secrets, which makes collaboration safer and audits simpler.

For builders

You don’t have to rework your whole stack. Add one confidential step where it counts, keep the rest unchanged, and expand from there.

What you can ship right now

• Private alerts that compute eligibility on encrypted data and notify only the right wallets.
• Risk & scoring engines that return a signed score without revealing inputs.
• Protected data feeds and paid results that never expose the underlying dataset or model.

A simple mental model

Think of a sealed room inside the processor. You slide a locked briefcase in, the work happens inside the room, and a stamped result comes out. Nobody outside the room touches the documents, everyone can trust the stamp, and once the job is done, the room “self-destructs” (memory is cleared), leaving no trace of what was inside.

If you want a hands-on path, start with a short guide that shows you how to encrypt inputs, verify attestation, and return an attested result.