SentioCP Attestation & Verification Framework (SAVF)

SentioCP introduces a cryptographic attestation framework designed to verify the provenance, integrity, and authenticity of outputs generated by decentralized AI agents. This framework leverages digital signatures, zero-knowledge proofs, and deterministic state representations to achieve complete verifiability—without relying on any centralized authority.


1. Formal Representation of Agent State

Let a SentioCP Agent be defined as a deterministic function:

fθ:X→Yf_\theta : X \to Yfθ​:X→Y

where:

  • XXX is the input space

  • YYY is the output space

  • θ∈Rn\theta \in \mathbb{R}^nθ∈Rn represents the fixed model parameters

To cryptographically bind an output to a specific agent instance, we derive a state commitment using a collision-resistant hash function HHH:

C=H(fθ ∥ θ ∥ m)C = H(f_\theta \, \| \, \theta \, \| \, m)C=H(fθ​∥θ∥m)

where mmm is the input message or prompt, and ∥\|∥ denotes byte-level concatenation.


2. Attestation Signature Scheme

Each agent instance is initialized with a public-private key pair (pk,sk)(pk, sk)(pk,sk).

Upon generating an output y=fθ(m)y = f_\theta(m)y=fθ​(m), the agent signs the tuple (m,y,C)(m, y, C)(m,y,C) using a digital signature algorithm such as EdDSA:

σ=Signsk(m,y,C)\sigma = Sign_{sk}(m, y, C)σ=Signsk​(m,y,C)

The attestation package (m,y,C,σ,pk)(m, y, C, \sigma, pk)(m,y,C,σ,pk) can be independently verified:

Verifypk(m,y,C,σ)=trueVerify_{pk}(m, y, C, \sigma) = \text{true}Verifypk​(m,y,C,σ)=true


3. Deterministic Output Verification

To eliminate nondeterminism in generative model outputs, SentioCP enforces seed-locked generation:

y=fθ(m;s)y = f_\theta(m; s)y=fθ​(m;s)

where sss is a shared pseudorandom seed. This ensures that the same input and seed always yield the same output:

fθ(m1;s)=fθ(m2;s)  ⟹  m1=m2f_\theta(m_1; s) = f_\theta(m_2; s) \implies m_1 = m_2fθ​(m1​;s)=fθ​(m2​;s)⟹m1​=m2​

The seed is included in the attestation hash:

C=H(fθ ∥ θ ∥ m ∥ s)C = H(f_\theta \, \| \, \theta \, \| \, m \, \| \, s)C=H(fθ​∥θ∥m∥s)


4. Zero-Knowledge Proof of Execution

For proprietary or sensitive models, SentioCP supports optional zk-SNARK-based attestation.

A prover generates a succinct proof π\piπ such that:

π=Prove(Rf,m,y)\pi = Prove(R_f, m, y)π=Prove(Rf​,m,y)

where RfR_fRf​ encodes the execution circuit of fθf_\thetafθ​. A verifier can confirm correctness without revealing the model internals:

VerifyZK(π,m,y)=trueVerify_{ZK}(\pi, m, y) = \text{true}VerifyZK​(π,m,y)=true

This allows third parties to trust outputs while keeping model parameters and architecture confidential.


5. Blockchain Anchoring

To ensure immutability, SentioCP periodically commits attestation hashes to a blockchain smart contract (e.g., Ethereum):

SubmitAttestation(C,t)→on-chainSubmitAttestation(C, t) \to \text{on-chain}SubmitAttestation(C,t)→on-chain

where ttt is a timestamp or block height. This provides a tamper-proof audit trail for all model responses, verifiable by any participant.


6. Security Assumptions

The security of the SentioCP Attestation & Verification Framework relies on standard cryptographic assumptions:

  • Collision resistance of the hash function HHH

  • Unforgeability of the digital signature scheme under chosen-message attacks (UF-CMA)

  • Soundness and completeness of the zk-SNARK proof system

  • Determinism of the agent fθf_\thetafθ​ under fixed seeds

Last updated