What is attribute-based access control (ABAC)?
Attribute-based access control (ABAC) is an access model that decides each request by evaluating attributes of the user, the resource, the action and the environment against policy rules. Instead of a fixed role, a rule might allow access only for users in the same region as the record during business hours. NIST describes ABAC in Special Publication 800-162.
For example, a bank lets loan officers read applicant files, but only for applicants in their own branch and only while the application is open. A loan officer in Austin opens an Austin file and it loads. The same officer tries a Dallas file and is turned away. Same role, different attributes, different result.
How ABAC works
ABAC policies combine four kinds of attributes:
- Subject: who is asking, such as department, clearance level or employment type.
- Resource: what is requested, such as data classification, owner or region.
- Action: what the request wants to do, such as read, edit or export.
- Environment: the conditions, such as time of day, device or network location.
A policy decision point evaluates the rules for each request, and a policy enforcement point applies the result. XACML is a common standard for writing these policies, and many cloud platforms offer their own ABAC rules through resource tags.
ABAC vs RBAC
| ABAC | RBAC | |
|---|---|---|
| Access decided by | Rules over many attributes | The user’s role |
| Changes per request | Yes | No, until the role changes |
| Good fit | Large, varied access needs | Stable job functions |
| Main cost | Writing and testing rules | Managing many roles |
Most organizations run both: RBAC for broad access and ABAC rules for sensitive cases. A hospital might give every clinician the clinician role, then add an ABAC rule that limits records to patients on the clinician’s current care team. The role keeps the setup simple, and the rule handles the exception that a role alone cannot express.
ABAC and AI agents
ABAC fits AI agents better than plain roles because attributes can include the AI agent’s identity, the task and the model a request is going to. But ABAC still decides on labels around the data, not on what is inside it. A document tagged “internal” may contain a customer’s diagnosis in paragraph four. In AI workflows, the content of the prompt or the retrieved chunk is what reaches the model, so a check on the content itself matters. Context-based access control adds that check.
ABAC rules are also only as good as the attributes behind them. If data labels are missing or out of date, the rules make the wrong call. Many teams pair ABAC with automated data classification so the resource attributes stay accurate as new files and records arrive.
How Protecto helps
Protecto adds a content check to attribute-based policies. Its agentic detection and classification engine finds sensitive values inside prompts, documents and tool outputs, and policy decides who sees real values. Your application passes the user, AI agent and task with each request, and anyone outside the policy sees tokens. See data access control for AI agents.
Frequently asked questions
What are examples of ABAC attributes?
User department, clearance level, data classification, record owner, region, time of day and device type. A rule combines several of them into one decision.
Is ABAC better than RBAC?
ABAC is more flexible but harder to set up and audit. RBAC is simpler and works well for stable job functions. Many teams use RBAC as the base and ABAC for the sensitive cases.
Can ABAC and RBAC be used together?
Yes. The user’s role can itself be one of the attributes an ABAC rule checks. Read more in attribute-based access control.
Related terms
Role-based access control (RBAC) · Context-based access control (CBAC) · Least privilege access · Data classification · AI agent