ArmaraOS: SOC2-style (TSC) control mapping
Maps AICPA Trust Services Criteria (CC6 logical access, CC7 system operations) to ArmaraOS in-product controls vs operator responsibilities. For shared responsibility and questionnaires — not legal or audit advice.
ArmaraOS: SOC2-style (TSC) control mapping (product vs operator)
This page mirrors the canonical copy in the ArmaraOS repository: docs/operations/SOC2_CONTROL_MAPPING.md. Prefer that path when linking from other repositories.
Sister page (AINL language & runtime): AINL: SOC2-style (TSC) control mapping — canonical on GitHub: ainativelang/…/AINL_SOC2_CONTROL_MAPPING.md.
This document maps common AICPA Trust Services Criteria (TSC) control themes to what ArmaraOS provides in software versus what deployers and enterprises must own in their environment, process, and monitoring stack. Use it for shared responsibility statements and customer questionnaires. Align criterion IDs to the exact TSC publication year your auditor uses; numbering and labels can differ slightly by edition.
Related (ArmaraOS on GitHub): Security systems · Security policy / reporting · API reference (audit routes)
Shared responsibility (summary)
ArmaraOS implements in-process controls: agent least-privilege (capabilities, tool policy), kernel RBAC for configured human/API use, a tamper-evident Merkle audit trail (with HTTP inspect/verify; see the API reference above), and other layers documented in docs/security.md.
Organizational access (e.g. tenants, org-wide IdP, SCIM), network and host boundaries, encryption at rest and in motion for a given deployment, log retention in a compliance or WORM store, and SIEM correlation (including human user identity and client IP for every event) are typically customer or operator responsibilities unless you run a managed service that explicitly includes them.
CC6 — Logical and physical access controls
| Criterion (typical intent) | In ArmaraOS (software) | Operator / customer | Notes |
|----------------------------|------------------------|------------------------|--------|
| CC6.1 — Logical access: software, infrastructure, architecture | Capabilities; RBAC (owner / admin / user / viewer); path / SSRF / tool policy; WASM sandbox; OFP wire auth (see SECURITY.md / docs/security.md) | OS hardening, segmentation, install/run policy, endpoint protection | Multi-tenant control plane (orgs, teams, IdP) is not a first-class product object. |
| CC6.2 — Register / authorize before system credentials | User config, channel bindings, api_key_hash | IdP provisioning, secrets manager, API key lifecycle | SSO/SCIM is not built-in. |
| CC6.3 — Remove or deactivate access | Edits to user config; key changes | Offboarding, rotation, HR-driven deprovisioning | Deprovisioning is process + automation around config, not a cloud user directory. |
| CC6.4 — Physical access to facilities | N/A (software) | Device / facility / colo physical security | |
| CC6.5 — Logical access to protected information | Capability grants; child cannot exceed parent; secret handling (see docs/security.md) | Data classification, DLP, encryption of home dir / backups | Classification of business data is customer. |
| CC6.6 — Restrict logical access to sensitive data | Per-agent memory; tool allow/deny; path restrictions | Host filesystem permissions, FDE, backup access | Host OS may access the same paths as the process. |
| CC6.7 — Restriction of data during transmission (e.g. encryption) | TLS / mTLS where used for admin or wire; SSRF mitigations | Cert management, TLS for all admin surfaces, perimeter controls | End-to-end story is the stack. |
| CC6.8 — Malware and unauthorized software | Ed25519 manifests, skill scanning, supply-chain awareness in policy | Change control, allowlisted installs | Governance of third-party skills is process. |
CC7 — System operations (detection, monitoring, response)
| Criterion (typical intent) | In ArmaraOS (software) | Operator / customer | Notes |
|----------------------------|------------------------|------------------------|--------|
| CC7.1 — Detect and communicate anomalies and events | Merkle audit log; GET /api/audit/recent, GET /api/audit/verify; graph-memory JSONL audit (separate surface) | SIEM, alerting, runbooks | Integrity of the in-product log is strong; actor (human) + IP are not the default uniform fields on every line — enrich in SIEM. |
| CC7.2 — Monitor system components | Health endpoints (redaction per security doc); usage persistence (see architecture/monitoring docs) | Infra APM, disk/CPU, log shipping, SLOs | No managed 24/7 NOC. |
| CC7.3 — Evaluate security events | verify_integrity() for audit chain; policy deny outcomes in logs | SOC playbooks, correlation | Detection logic is customer SIEM. |
| CC7.4 — Respond to security incidents and breaches | Hardening and safe defaults; SECURITY.md contact | IR plan, comms, forensics | Product is one component. |
| CC7.5 — Vulnerabilities and system defects (identify, log, track) | Supported versions, disclosure process, development practices | Patching, VM, image updates | Dependency and image cadence is operator. |
Suggested control ownership (matrix row)
| Control area | ArmaraOS (design) | Operator (run time) | |--------------|-------------------|---------------------| | Agent permission model | Yes (capabilities, tool policy) | Tunes manifests, skills, policy | | Human RBAC | Yes (kernel config) | Provisions users, keys, service restarts | | Tamper-evident audit | Yes (Merkle chain + verify API) | Export, retain, monitor, alert | | Enterprise IAM, tenancy, ABAC | Not by default | IdP, namespaces, separate installs, wrappers | | Immutable WORM / legal-hold log archive | Not default | Object lock, secondary store, log pipeline |
Disclaimer
This is not legal or audit advice. A SOC2 report is issued for a system description and control objectives in scope; map this document to your auditor’s criteria list and to any customer-specific deployment architecture.