Secure Systems Group
Menu×

Wiki Home

Remote attestation

Definition

Remote attestation is a protocol in which a prover convinces a remote verifier that it is running expected code on genuine hardware. A hardware root of trust, a component the verifier trusts by assumption because a key was embedded in the silicon at manufacture, signs a measurement: a hash of the code and configuration that was loaded. The verifier checks the signature against the hardware vendor’s certificate, confirming the report came from authentic hardware, and compares the measurement to the value it expects. A trusted execution environment supplies this naturally, since it already holds a hardware-protected key and can measure the program loaded into an enclave; the signed report is often called a quote. The guarantee is narrow. Attestation certifies which code loaded and that the hardware is real; it says nothing about whether that code is free of run-time bugs once it executes (De Oliveira Nunes et al., 2024).

Property-based attestation generalizes the goal. Rather than certify one exact configuration, the verifier wants to know that a configuration has a property. The verifier obtains the signed measurement from the prover and reference values for the property from a trusted authority, and concludes the property holds if the measurement matches a reference value (Sadeghi and Stüble, 2004). This is the one structural difference from standard remote attestation: the verifier consults an external authority for reference values instead of recomputing the expected measurement itself. Applied to machine learning, a measurer hashes the inputs and outputs of an operation such as training, evaluation, or inference, and the resulting attestations populate verifiable property cards, the model card, datasheet, and inference card a third party can check rather than take on faith.

Papers that use this concept

  • PAL*M: Property Attestation for Large Generative Models — the central mechanism: a confidential virtual machine and a TEE-aware GPU jointly measure an ML operation, and a hardware-signed quote attests properties of large generative models to a verifier who consults a trusted authority for reference values.

Variants and traps

  • Standard remote attestation certifies a fixed configuration; property-based attestation certifies a property, and for that the verifier must obtain reference values from a trusted authority rather than recompute them.
  • Attestation binds what code loaded, not that the loaded code is correct. A measured program can still contain a vulnerability exploited at run time.

See also

References

  • De Oliveira Nunes, I., et al. “Toward Remotely Verifiable Software Integrity in Resource-Constrained IoT Devices.” IEEE Communications Magazine, 62(7):58-64, 2024.
  • Sadeghi, A.-R. and Stüble, C. “Property-Based Attestation for Computing Platforms: Caring About Properties, Not Mechanisms.” Workshop on New Security Paradigms (NSPW), 2004.