Attribute-Based Access Control for Agents

Attribute-based access control (ABAC) evaluates access decisions based on attributes of the subject, the resource, the action, and the environment. Unlike RBAC, which assigns static roles, ABAC can express dynamic conditions that depend on the current context of the request.

Attributes in Agent Systems

Four categories of attributes are relevant:

Subject attributes: Properties of the requesting agent, such as trust level, owning team, model version, and creation date.

Resource attributes: Properties of the target resource, such as sensitivity classification, data residency region, and owning department.

Action attributes: Properties of the requested action, such as whether it is read-only, whether it modifies external state, and its risk classification.

Environment attributes: Context that is not tied to the subject or resource, such as current time, system load, and active incident status.

Policy Expression

ABAC policies express rules as conditions over attributes:

rules:
  - conditions:
      subject.trust_level: "high"
      resource.classification: "confidential"
      action.type: "read"
      environment.incident_active: false
    effect: "allow"

This rule allows a high-trust agent to read confidential resources, but only when no active incident is in progress.

Advantages Over RBAC

ABAC eliminates role explosion. In RBAC, supporting combinations of team, function, and environment requires creating a role for every combination. ABAC expresses the same logic as attribute conditions without creating any roles.

ABAC also handles dynamic context natively. A rule that depends on current system load or active incident status cannot be expressed in pure RBAC because roles do not change based on environment.

Combining RBAC and ABAC

In practice, most systems use both. RBAC handles the coarse structure (this agent is a reader, this one is an executor), and ABAC handles the fine-grained conditions (this reader can access confidential data only during business hours).

Performance Considerations

ABAC evaluation is more computationally expensive than RBAC because it evaluates multiple attribute conditions per rule. Authensor's policy engine keeps evaluation synchronous and fast by using indexed attribute lookups and short-circuit evaluation on condition chains.

ABAC gives you the expressiveness to write policies that match your actual security requirements, not a simplified approximation.