security_architecture - lgcorzo/rust_CACD_autonomous_factory GitHub Wiki

iso_doc_type: "Description" iso_viewpoint: "SecurityView" type: "security" title: "Security Architecture โ€” Dark Gravity CA/CD Autonomous Factory" description: "ISO 42010 SecurityView / ISO 15289 Description documentation for Security Architecture โ€” Dark Gravity CA/CD Autonomous Factory." tags: ['iso42010', 'okf', 'security_view', 'zero_trust', 'governance'] timestamp: "2026-10-01T14:00:00Z" generated: "agent:okf-professional-documenter" verified: "true" last_verified_commit: "fbdc5848"

Security Architecture โ€” Dark Gravity CA/CD Autonomous Factory

Purpose: Comprehensive documentation of the Zero Trust security model, Ed25519 Non-Human Identity credentials, sandbox containment, SAST gates, and threat model mapping.

1. Zero Trust Security Model Overview

The Dark Gravity Factory operates under a Zero Trust security posture: no implicit trust is granted to any agent, service, or network path. Every action requires explicit cryptographic verification.

graph TB
    subgraph "Trust Boundary: External"
        GH["GitHub API"]
        GL["GitLab API"]
        JR["Jira API"]
    end
    
    subgraph "Trust Boundary: Network Layer"
        ZITI["OpenZiti Overlay"]
    end
    
    subgraph "Trust Boundary: Identity Layer"
        NHI["Ed25519 NHI Issuer"]
        VAULT["Vault Secret Store"]
        JIT["JitToken Manager"]
    end
    
    subgraph "Trust Boundary: Execution Layer"
        SAST["SAST Gate: score >= 8.0"]
        SANDBOX["gVisor Sandbox Pod"]
        CB["Circuit Breaker: max 3 retries"]
    end
    
    subgraph "Trust Boundary: Governance Layer"
        HITL["HITL 4-Vertex Mesh"]
    end
    
    GH --> ZITI
    GL --> ZITI
    ZITI --> NHI
    NHI --> SAST
    SAST --> SANDBOX
    SANDBOX --> CB
    CB --> HITL

    style ZITI fill:#2196F3,stroke:#1565C0,color:#fff
    style NHI fill:#4CAF50,stroke:#2E7D32,color:#fff
    style SANDBOX fill:#FF9800,stroke:#E65100,color:#fff
    style HITL fill:#9C27B0,stroke:#6A1B9A,color:#fff
Loading

2. Ed25519 Non-Human Identity (NHI) Lifecycle

Every autonomous mission requires a cryptographic identity. The NHI lifecycle ensures traceability from issue ingestion through code delivery.

sequenceDiagram
    participant Poller as PollerDaemonService
    participant NHI as NHI Issuer
    participant Vault as HashiCorp Vault
    participant Agent as ZeroClawAgent
    participant Validator as SecurityValidator

    Poller->>NHI: Request credential for mission
    NHI->>Vault: Retrieve Ed25519 signing key
    Vault-->>NHI: SigningKey (zeroized after use)
    NHI->>NHI: Sign mission payload with Ed25519
    NHI-->>Poller: VerifiableCredential (JSON-LD)
    
    Poller->>Agent: Dispatch task with NHI credential
    Agent->>Validator: validate_signature(data, sig)
    Validator->>Validator: Base64 decode + Ed25519 verify
    Validator-->>Agent: Verification result (bool)
    
    alt Signature Valid
        Agent->>Agent: Execute task in sandbox
    else Signature Invalid
        Agent->>Agent: REJECT - FactoryError::Security
    end
Loading

Key Security Types

Type Module Description
Ed25519SecurityValidator factory-core::security Implements SecurityValidator trait with ed25519-dalek
JitToken factory-core::security Ephemeral token with zeroize::ZeroizeOnDrop โ€” wiped from memory on drop
SecurityBounds factory-core::security Trait for token validation and JIT issuance
AuditResult factory-core::security Content audit output: is_safe, findings

3. Sandbox Containment Architecture

Agent code execution is isolated in gVisor-sandboxed Kubernetes pods with strict resource constraints.

graph TB
    subgraph "K8s Pod: Sandbox"
        subgraph "gVisor Runtime"
            APP["Application Container"]
            SIDE["Sidecar Container"]
        end
        NET["Network Policy: No Egress"]
        RES["Resource Quota"]
    end
    
    MCP["factory-mcp-server"] --> |"launch_sandbox_pod"| APP
    APP --> |"max 30 MiB RAM, 0.25 CPU"| RES
    SIDE --> |"max 20 MiB RAM, 0.10 CPU"| RES
    NET --> |"network_egress_allowed: false"| APP

    style APP fill:#FF9800,stroke:#E65100,color:#fff
    style NET fill:#f44336,stroke:#c62828,color:#fff
Loading

SandboxConstraint Enforcement

// From factory-core::security
pub struct SandboxConstraint {
    pub max_memory_mb: u32,      // Default: 30 MiB
    pub max_cpu_cores: f32,       // Default: 0.25 cores
    pub network_egress_allowed: bool, // Default: false
}

// ZeroClawAgent validates before every execution:
// - App container: max_memory_mb <= 30, max_cpu_cores <= 0.25
// - Sidecar container: max_memory_mb <= 20, max_cpu_cores <= 0.10
// - network_egress_allowed must be false for both

4. SAST Gate Integration

Every code change passes through a Static Application Security Testing gate before execution. The SastScanResult enforces a minimum score of 8.0/10.0.

flowchart TD
    DIFF["Code Diff"] --> SCAN["SastScanResult::inspect_diff"]
    SCAN --> CHECK1{"Hardcoded Secrets?"}
    CHECK1 -->|Yes| CRIT1["CRITICAL: -4.0 score"]
    CHECK1 -->|No| CHECK2{"Command Injection?"}
    CHECK2 -->|Yes| CRIT2["CRITICAL: -5.0 score"]
    CHECK2 -->|No| CHECK3{"SQL Injection?"}
    CHECK3 -->|Yes| CRIT3["CRITICAL: -4.0 score"]
    CHECK3 -->|No| CHECK4{"Unbounded Loop?"}
    CHECK4 -->|Yes| WARN1["WARNING: -2.5 score"]
    CHECK4 -->|No| PASS["Score: 10.0"]
    
    CRIT1 --> GATE
    CRIT2 --> GATE
    CRIT3 --> GATE
    WARN1 --> GATE
    PASS --> GATE
    
    GATE{"score >= 8.0 AND<br/>no critical vulns?"}
    GATE -->|Yes| ALLOW["โœ… PASS: Execute in Sandbox"]
    GATE -->|No| BLOCK["โŒ BLOCK: Execution Rejected"]

    style BLOCK fill:#f44336,stroke:#c62828,color:#fff
    style ALLOW fill:#4CAF50,stroke:#2E7D32,color:#fff
Loading

5. Circuit Breaker Anti-Deadlock (Aethelgard)

The Aethelgard circuit breaker prevents infinite agent retry loops:

stateDiagram-v2
    [*] --> ActiveExecution
    ActiveExecution --> TaskRetry: Execution/Test Error
    TaskRetry --> ActiveExecution: Attempt < 3
    TaskRetry --> DeadlockTripped: Attempt == 3 AND Diff Hash Stagnant
    DeadlockTripped --> AgentStuck: Trip Circuit Breaker
    AgentStuck --> HumanIntervention: Alert via HITL Vertex 3
    HumanIntervention --> ActiveExecution: Architect Override + Parameter Tune
    HumanIntervention --> MissionAborted: Mission Rejection
    MissionAborted --> [*]
Loading

6. STRIDE Threat Model Mapping

Threat Category Control Implementation
Agent impersonation Spoofing Ed25519 NHI credentials SecurityValidator::validate_signature()
Unauthorized code execution Tampering SAST gate + sandbox isolation SastScanResult::passes_gate() + gVisor
Mission data exposure Information Disclosure OpenZiti encrypted tunnels ZitiClient Zero Trust overlay
Denial of service via token burn Denial of Service FinOps budget hardstops DailyBudgetConfig::hardstop_threshold_ratio
Privilege escalation Elevation of Privilege Minimal sandbox constraints SandboxConstraint::validate_bounds()
Audit trail tampering Repudiation Cryptographic NHI claims Ed25519 signed verifiable credentials

Related: HITL Governance ยท Verification Triad ยท Compliance & Audit

โš ๏ธ **GitHub.com Fallback** โš ๏ธ