Hybrid PQC Signing
import { Steps, Aside, Tabs, TabItem } from ‘@astrojs/starlight/components’;
Hybrid Post-Quantum Cryptographic (PQC) Signing
Section titled “Hybrid Post-Quantum Cryptographic (PQC) Signing”AgentV v1.6.2 introduces support for Hybrid Post-Quantum Cryptographic (PQC) Signing. This protocol is designed to protect industrial forensic traces against emerging quantum computing threats while maintaining strict data privacy through the Zero-Exposure Signing (ZES) pattern.
This guide is intended for Auditors, Integrators, and Evaluators who need to ensure the long-term non-repudiability of agentic evaluation results.
🏗️ Architecture
Section titled “🏗️ Architecture”The hybrid protocol combines classical elliptic curve cryptography with modern lattice-based algorithms, ensuring that the system remains secure even if one of the layers is compromised.
- Classical Layer: Ed25519 (SHA-512 + Curve25519) - Used for high-performance, universally compatible signing.
- Post-Quantum Layer: ML-DSA-65 (Module-Lattice-based Digital Signature Algorithm) - Aligned with NIST’s FIPS 204 standard for quantum resistance.
- Hybrid Binding: Both signatures are mathematically bound to the same Verification Certificate (VC) v3.0.0 and stored in the
provenance_chain.
🧬 Zero-Exposure Signing (ZES) Protocol
Section titled “🧬 Zero-Exposure Signing (ZES) Protocol”To maintain the privacy of industrial evaluation data, AgentV implements the Zero-Exposure Signing (ZES) pattern. This ensures that raw traces, trajectories, and sensitive logs never leave the project’s security jail.
The ZES Flow
Section titled “The ZES Flow”-
Generate Manifest (VC v3) The harness assembles the
run_manifest.jsoncontaining the trace hash and forensic evidence ledger. -
Compute SHAKE-256 Digest The manifest is hashed locally using SHAKE-256. This creates a fixed-length (32-byte) cryptographic condensation of the data.
-
Secure Transmission Only the resulting 32-byte digest is transmitted to the PQC provider (CycleCore) along with your API Key. The raw trace data never leaves your environment.
-
Remote Signing The provider signs the digest using the ML-DSA-65 algorithm and returns the signature hex.
-
Seal Certificate The signature is appended to the
provenance_chainand the Verification Certificate is sealed for final audit.
⚙️ Configuration (Environment Variables)
Section titled “⚙️ Configuration (Environment Variables)”Enable and configure Hybrid PQC using the following parameters. You can set these in your .env file.
| Variable | Default | Description |
|---|---|---|
PQC_ENABLED | false | Set to true to activate the hybrid signing pipeline. |
PQC_PROVIDER | cyclecore | The cryptographic provider (e.g., cyclecore). |
PQC_STRICT_MODE | false | If true, fails evaluation if PQC signing fails. |
| Variable | Default | Description |
|---|---|---|
CYCLECORE_API_KEY | (None) | Your ZES API key (Required for remote signing). |
CYCLECORE_IDENTITY_ID | default | The identity name for the PQC signature. |
🛠️ Troubleshooting
Section titled “🛠️ Troubleshooting”Common CycleCore API Errors
Section titled “Common CycleCore API Errors”| Error Symptom | Potential Cause | Resolution |
|---|---|---|
Authentication Failure | Invalid or expired CYCLECORE_API_KEY. | Check your environment variables and CycleCore dashboard. |
Rate Limit Exceeded | Too many concurrent signing requests. | Implement ADAPTER_RETRY_DELAY or contact support for higher limits. |
Connection Timeout | Network/Firewall blocking egress to the PQC provider. | Ensure your host can reach the CycleCore API endpoints. |
ImportError: cyclecore-pq | The cyclecore-pq package is missing. | Run pip install cyclecore-pq or check your pyproject.toml. |
Verification Failures
Section titled “Verification Failures”- Signature Mismatch: Ensure the
PQC_IDENTITY_IDused during signing matches the one configured during verification. - Hash Inconsistency: If the local
SHAKE-256digest differs from the one recorded in the manifest, verification will fail automatically to prevent tampering.
⚖️ NIST Alignment
Section titled “⚖️ NIST Alignment”The Hybrid PQC implementation is explicitly aligned with NIST AI-100-1 and NIST FIPS 204, providing a non-repudiable bridge between raw evaluation data and regulatory-grade compliance reports.