Resource Classification Configuration

Field	Value
Document ID	ASCEND-GOV-008
Version	1.0.0
Last Updated	March 27, 2026
Author	OW-KAI Engineering
Publisher	OW-KAI Technologies Inc.
Classification	Confidential
Compliance	SOC 2 CC6.1, NIST AC-3, PCI-DSS 7.1

Reading Time: 10 minutes | Skill Level: Advanced

Overview

Resource Classification Configuration allows administrators to define sensitivity tiers and risk score modifiers for each resource type in their organization. When an AI agent submits an action, ASCEND looks up the resource type against the admin-defined classification table to determine the correct sensitivity tier and risk multiplier.

This implements a defense-in-depth model: the platform trusts admin-defined classifications over agent-declared sensitivity. Agents cannot downgrade the sensitivity of a resource type that an administrator has classified as critical.

Key principles:

• Admin authority: Administrator-defined classifications always override agent-declared sensitivity levels.
• Fail-secure default: Any resource type not found in the classification table defaults to CRITICAL sensitivity with a 1.5x risk multiplier.
• Tenant isolation: Classifications are scoped per organization via Row Level Security (RLS). One tenant's classifications are invisible to another.
• Audit trail: Every create, update, and deactivation is logged as an immutable CONFIG_CHANGE event with before/after state.

How It Works

When an agent submits an action with a resource_type field, the classification resolution engine follows this flow:

The resolution follows three paths:

Condition	Behavior	Risk Impact
resource_type present and classification found	Admin-defined tier and modifier applied	Modifier from classification (0.1x to 3.0x)
resource_type present but no classification found	Fail-secure: CRITICAL tier, 1.5x modifier	Elevated risk score until admin configures the type
resource_type absent from action payload	No override applied; existing flow unchanged	Standard risk calculation without resource modifier

Sensitivity Tiers

Each classification maps a resource type to one of four sensitivity tiers. The tier determines the data_sensitivity field injected into the risk pipeline and controls whether the contains_pii flag is set.

Tier	Risk Posture	contains_pii	Typical Use
LOW	Minimal sensitivity; routine operations	false	Public documentation, non-sensitive logs, static assets
MEDIUM	Moderate sensitivity; standard business data	false	Internal APIs, application configuration, workflow state
HIGH	Elevated sensitivity; personal or financial data	true	Customer records, payment data, HR systems
CRITICAL	Maximum sensitivity; regulated or privileged data	true	Encryption keys, secrets managers, production databases, PII stores

The risk_score_modifier field applies a multiplier (0.1x to 3.0x) to the weighted risk score calculated by the risk assessment engine. For example, a resource classified as CRITICAL with a 2.0x modifier will double the computed risk score for any action targeting that resource type.

Fail-Secure Default

When an agent references a resource type that has no active classification in the tenant's configuration, ASCEND applies fail-secure defaults:

Parameter	Default Value	Rationale
sensitivity_tier	critical	Unknown resources are treated as maximally sensitive
risk_score_modifier	1.5	Elevated multiplier ensures human review for unclassified resources
contains_pii	true	Conservative assumption until admin classifies the resource

This design ensures that newly introduced resource types cannot bypass governance controls. The fail-secure behavior persists until an administrator explicitly creates a classification for the resource type.

Deactivating an existing classification also triggers fail-secure behavior. The deactivation confirmation in the UI warns administrators that actions using the deactivated resource type will default to CRITICAL sensitivity.

API Reference

All endpoints require admin role (admin, org_admin, or super_admin) and enforce tenant isolation via the organization_id derived from the authenticated session.

Base path: /api/v1/resource-classifications

List Classifications

GET /api/v1/resource-classifications

Query parameters:

Parameter	Type	Default	Description
sensitivity_tier	string	—	Filter by tier: low, medium, high, critical
is_active	boolean	—	Filter by active status
limit	integer	100	Maximum results (1-1000)
offset	integer	0	Pagination offset

Response:

{
  "classifications": [
    {
      "id": 1,
      "organization_id": 4,
      "resource_type": "database",
      "display_name": "Relational Database",
      "description": "Production RDS instances",
      "sensitivity_tier": "critical",
      "risk_score_modifier": 2.0,
      "is_active": true,
      "created_at": "2026-03-27T14:30:00+00:00",
      "updated_at": "2026-03-27T14:30:00+00:00",
      "created_by": "admin@acme.com",
      "updated_by": "admin@acme.com"
    }
  ],
  "total": 1
}

Create Classification

POST /api/v1/resource-classifications

Request body:

{
  "resource_type": "database",
  "display_name": "Relational Database",
  "description": "Production RDS instances",
  "sensitivity_tier": "critical",
  "risk_score_modifier": 2.0
}

Validation rules:

Field	Constraint
resource_type	Required, 1-100 characters, normalized to lowercase, unique per organization
display_name	Required, 1-255 characters
sensitivity_tier	Required, must be low, medium, high, or critical
risk_score_modifier	Required, float between 0.1 and 3.0

Response: 201 Created with the full classification object.

Error responses:

Status	Condition
409 Conflict	Resource type already exists for this organization
422 Unprocessable Entity	Validation failure (invalid tier, modifier out of range, empty fields)
403 Forbidden	Caller does not have admin role

Update Classification

PUT /api/v1/resource-classifications/{id}

All fields are optional. The resource_type field cannot be changed after creation.

{
  "display_name": "Production Database",
  "sensitivity_tier": "high",
  "risk_score_modifier": 1.8,
  "is_active": true
}

Response: 200 OK with the updated classification object.

Deactivate Classification

DELETE /api/v1/resource-classifications/{id}

Performs a soft delete by setting is_active = false. The classification record is retained for audit purposes.

Response:

{
  "success": true,
  "message": "Classification 'database' deactivated. Actions using this resource type will default to CRITICAL sensitivity (fail-secure)."
}

Configuration via UI

The Resource Classifications management interface is accessible from Settings > Resource Classifications in the ASCEND admin console.

Adding a Classification

1. Navigate to Settings and select the Resource Classifications tab.
2. Click + Add Classification.
3. Enter the resource type identifier (e.g., database, s3, lambda). This value is normalized to lowercase and must be unique within your organization.
4. Provide a display name and optional description.
5. Select the sensitivity tier from the dropdown.
6. Set the risk score modifier (0.1x to 3.0x).
7. Click Create to save.

Editing a Classification

Click Edit on any active classification to modify its display name, description, sensitivity tier, or risk score modifier. The resource type identifier cannot be changed after creation. All changes are recorded in the audit trail with before and after values.

Deactivating a Classification

Click Deactivate on any active classification. A confirmation dialog warns that actions using this resource type will default to CRITICAL sensitivity under fail-secure rules. Deactivated classifications can be reactivated by toggling the Show deactivated checkbox and clicking Reactivate.

Caching

Resource classification lookups are cached in Redis with a 5-minute TTL to minimize database load during high-volume action processing.

Behavior	Detail
Cache key format	rc:{organization_id}:{resource_type}
TTL	300 seconds (5 minutes)
Negative caching	Cache misses are stored as a __miss__ sentinel to prevent repeated database lookups for unrecognized resource types
Invalidation	All cached classifications for the organization are invalidated on any create, update, or deactivate operation
Graceful degradation	If Redis is unavailable, lookups fall back to direct database queries with no impact on correctness

Cache invalidation uses a SCAN-based key deletion pattern (rc:{org_id}:*) to ensure all entries for the affected organization are cleared without impacting other tenants.

Audit Trail

Every classification operation generates an immutable audit event via the ImmutableAuditService. These events are hash-chained and cannot be modified or deleted.

Operation	Event Type	Risk Level	Recorded Data
Create	CONFIG_CHANGE	MEDIUM	resource_type, sensitivity_tier, risk_score_modifier, display_name
Update	CONFIG_CHANGE	MEDIUM	Full before/after state for all modified fields
Deactivate	CONFIG_CHANGE	HIGH	resource_type, sensitivity_tier, soft_deleted flag, fail-secure effect statement

All audit events include:

• actor_id: Email of the administrator who performed the action
• organization_id: Tenant scope
• ip_address: Source IP from the request
• compliance_tags: ["SOX", "CONFIG_MANAGEMENT", "AUDIT_TRAIL"]

The elevated HIGH risk level on deactivation events reflects the security impact of removing a classification, which causes the affected resource type to default to fail-secure CRITICAL handling.

Compliance Mapping

Requirement	Standard	Implementation
Least privilege access control	SOC 2 CC6.1	Admin role required for all classification operations; RLS enforces tenant isolation
Access enforcement	NIST AC-3	Admin-defined classifications override agent-declared sensitivity; fail-secure defaults prevent bypass
Role-based access control	PCI-DSS 7.1	Only admin, org_admin, and super_admin roles can manage classifications
Configuration change management	SOX	All changes logged with before/after state in immutable audit trail
Audit trail generation	SOC 2 CC7.2	Hash-chained audit events for every create, update, and deactivate operation
Data classification	NIST SI-12	Sensitivity tiers map resource types to data protection requirements
Fail-secure design	NIST SC-24	Unrecognized resource types default to maximum sensitivity to prevent governance bypass