Tenzro Testnet is live. Get testnet TNZO

TEE Security

Hardware-secured computation with Intel TDX, AMD SEV-SNP, AWS Nitro, and NVIDIA Confidential Computing

Trusted Execution Environments provide hardware-level isolation for sensitive computation. Tenzro integrates with four major TEE platforms for attestation, confidential AI inference, and secure key management.

TEE-secured nodes can prove they are running unmodified code inside genuine hardware enclaves, enabling trustless verification without relying on the operator's honesty.

Key Features

Runtime Detection

detect_tee() automatically identifies available hardware: Intel TDX, AMD SEV-SNP, AWS Nitro, or NVIDIA GPU

Remote Attestation

Cryptographic proofs that code is running inside genuine hardware enclaves. Verify without trusting the operator

Confidential AI

Run inference inside TEEs. Model weights and user prompts never leave the enclave. Ideal for healthcare, legal, financial

Secure Key Management

MPC wallet key shares generated and stored inside TEEs. Keys never exist in plaintext outside hardware

Attestation Registry

On-chain registry of TEE providers with verified attestations. 2x consensus weight for TEE-attested validators

NVIDIA GPU TEE

Confidential computing on Hopper/Blackwell/Ada Lovelace GPUs. NRAS attestation with 24-hour report validity

Supported Platforms

Intel TDX

Trust Domain Extensions on 4th and 5th generation Xeon processors. Memory encryption with per-VM keys ensures complete isolation from the hypervisor and host OS.

Feature flag: intel-tdx

  • • Hardware-encrypted memory
  • • Per-VM cryptographic keys
  • • Remote attestation via Intel SGX DCAP
  • • Supports confidential containers

AMD SEV-SNP

Secure Encrypted Virtualization with Secure Nested Paging on AMD EPYC processors. GHCB (Guest-Hypervisor Communication Block) protocol for secure VM communication.

Feature flag: amd-sev-snp

  • • Memory encryption at VM level
  • • GHCB secure communication
  • • AMD PSP attestation reports
  • • Integrity protection against hypervisor attacks

AWS Nitro

Isolated compute on Amazon EC2 Nitro Enclaves. No persistent storage, no interactive access, no external networking from within the enclave. Perfect for confidential workloads.

Feature flag: aws-nitro

  • • Isolated CPU and memory
  • • Cryptographic attestation documents
  • • No SSH, no persistent storage
  • • Integration with AWS KMS

NVIDIA GPU

Confidential computing on Hopper H100/H200, Blackwell B100/B200, and Ada Lovelace architectures. NVIDIA Remote Attestation Service (NRAS) provides GPU attestation with 24-hour report validity.

Feature flag: nvidia-gpu

  • • GPU memory encryption
  • • Confidential AI inference
  • • NRAS attestation reports
  • • 24-hour attestation validity window

Code Examples

Detect and Attest

use tenzro_tee::{detect_tee, TeeProvider};

// Auto-detect available TEE hardware
let provider = detect_tee()?;
println!("TEE: {:?}", provider.platform());

// Generate attestation
let attestation = provider.attest(&report_data)?;

// Verify attestation (can be done on any node)
let valid = provider.verify(&attestation)?;

Build with TEE Features

# Build with Intel TDX support
cargo build --features intel-tdx

# Build with NVIDIA GPU support
cargo build --features nvidia-gpu

# Build with all TEE providers
cargo build --features intel-tdx,amd-sev-snp,aws-nitro,nvidia-gpu

Register TEE Provider On-Chain

use tenzro_tee::{TeeRegistry, detect_tee};

// Initialize registry
let registry = TeeRegistry::new();

// Generate attestation
let provider = detect_tee()?;
let attestation = provider.attest(&report_data)?;

// Register on-chain (2x consensus weight for validators)
registry.register(node_id, provider.platform(), attestation)?;

// Check if provider is registered
let is_registered = registry.is_registered(&node_id)?;

Confidential AI Inference

use tenzro_tee::detect_tee;
use tenzro_model::InferenceRouter;

// Run inference inside TEE
let tee = detect_tee()?;

// Model weights and prompts never leave hardware enclave
let result = router.route_request_with_tee(
    &request,
    &tee,
    RoutingStrategy::TeeOnly, // Only route to TEE providers
)?;

// Result includes TEE attestation proving execution environment
assert!(result.tee_attestation.is_some());

Deploy TEE-Secured Infrastructure

Run validators, AI inference, or key management with hardware attestation. TEE providers earn 2x consensus weight and premium fees for confidential computation.