Concepts

How Security Labels Work

From physical document stamps to cryptographically-enforced digital labels — a practical introduction to NATO label-based access control.

The Access Control Problem

In government and defence organisations, information is sensitive in different ways and to different degrees. A battlefield report, a diplomatic cable, and a public press release all require different handling — yet they may all live on the same network and be processed by the same systems.

The physical world solved this with stamps and envelopes: a document marked SECRET in red ink could only be handled by personnel who had been granted the appropriate clearance. Personnel carried an identity badge that expressed what they were permitted to access. Access was controlled by comparing the document's marking to the person's badge — a simple matching rule.

The fundamental question is unchanged in the digital world: does the person requesting access to this information hold a clearance that permits them to see it?

The challenge in electronic systems is that this check must happen automatically, consistently, and at every point where information could escape to an unauthorised consumer — not just at the front door.

What Is a Security Label?

A security label is a structured, machine-readable descriptor attached to a piece of information that expresses its sensitivity and the conditions under which it may be accessed. It is the electronic equivalent of the red SECRET stamp on a paper document.

A security label has three components:

Classification Level

A position in an ordered hierarchy — from UNCLASSIFIED to TOP SECRET. A clearance must reach at least this level for access to be permitted.

Categories

Additional, non-hierarchical restrictions: topic-based compartments, coalition releasability groups, geographic restrictions, or other organisational boundaries.

Security Policy

The domain policy (SPIF) that defines which label values are valid, what they mean, and the rules for deciding whether a label permits a given action.

Critically, in a data-centric model the label is not a separate database entry or a property of a storage location — it is bound to the content itself with a cryptographic signature. It travels with the data wherever the data goes.

Classification Levels

Classification levels form a strict hierarchy. Each level implies containment of all levels below it: a person cleared to SECRET may also access CONFIDENTIAL and below. The hierarchy defines a total ordering that makes the basic access decision straightforward: if your clearance level is at least as high as the document's classification level, the level criterion is satisfied.

UNCLASSIFIED
No restrictions on distribution
RESTRICTED
Limited distribution; not for public release
CONFIDENTIAL
Unauthorised disclosure could damage national security
SECRET
Serious damage expected from unauthorised disclosure
TOP SECRET
Exceptionally grave damage from disclosure
NATO uses this hierarchy.
Each NATO member nation's security policies define which levels apply within their domain. The NATO CONFIDENTIAL / NATO SECRET / NATO TOP SECRET classification marks are defined by the Brussels Accord and implemented in the domain's SPIF.

In a networked environment a classification ceiling is also defined per network or domain — the highest classification permitted to transit that network. Information whose label exceeds the domain ceiling must not enter that domain.

Categories and Caveats

Classification level alone is insufficient. A document may be SECRET and yet must only be shared with a specific coalition, or with a specific Community of Interest (COI) who have an operational need to know. Categories add these granular, non-hierarchical restrictions on top of the classification level.

Common category types in NATO practice include:

Releasability Groups

Restrict information to specific allied nations or coalitions. RELIDO CCEB NATO mark what national groupings may receive the information.

Compartments

Special Access Codes (SAC) or Sensitive Compartmented Information (SCI) isolate particularly sensitive information within a clearance level — clearance alone does not grant compartment access.

Handling Instructions

Operational caveats such as EYES ONLY NO FOREIGN ORCON that further restrict distribution or handling beyond the classification level and releasability groups.

Exclusive Restrictions

Single-valued flags such as CRYPTO or NOFORN that apply a specific requirement universally — access is denied unless the clearance explicitly grants that restriction.

Categories are represented in the SPIF as named bit-positions. A label carries a bit for each category; a clearance carries an authorisation bit for each category it permits. The access decision requires that every set bit in the label's category mask is also set in the clearance's authorisation mask.

Security Clearances

A security clearance is the mirror image of a security label. Where a label describes what a piece of information requires for access, a clearance describes what a person (or system) is authorised to access.

A clearance consists of:

  • A clearance level — the highest classification level the holder may access
  • A set of category authorisations — which categories the holder may access

The access decision is a straightforward match:

Document Label

SECRET RELIDO EYES ONLY
must be covered by

User Clearance

TOP SECRET RELIDO ✓ EYES ONLY ✓

Access is granted if and only if:

  1. The clearance level is at least the document's classification level, and
  2. Every category bit set in the document's label is also set in the clearance's authorisation.

Both conditions must be satisfied simultaneously. A TOP SECRET clearance does not override a compartment restriction; the compartment must be explicitly authorised.

The Security Policy — SPIF

A Security Policy Information File (SPIF) is a machine-readable file that defines everything required to evaluate and enforce a security policy within a domain:

  • The complete set of valid classification levels and their ordering
  • All valid categories, their type (permissive, restrictive, or requirement), and their bit positions
  • Marking instructions that turn label values into human-readable strings
A label is only valid if it can be validated against a SPIF. An information flow carrying a label that references values not present in the destination domain's SPIF is rejected — the domains are operating under incompatible policies and no access decision can be made.

In cross-domain scenarios, two SPIFs are relevant: the originating domain's SPIF (to validate the label as issued) and the destination domain's SPIF (to verify the information is releasable under the destination's policy). Both must agree before information is allowed to cross the domain boundary.

SPIFs are defined in the XMLSPIF format — a NATO-standardised XML schema available in versions 2.1 and 3.0. Tessera supports both.

Where Enforcement Happens

A label that is only checked in one place provides weak protection. An end-user application might be bypassed; a network relay might not enforce it; a storage system might not inspect it. Robust label-based access control requires multi-point enforcement — the label is checked at every boundary crossing.

Creation
Label applied & bound
Client App
UI enforcement
Gateway / Guard
Dual-SPIF check
KAS
Key release decision
Destination
Delivery check

In the Tessera model:

  • At the client: Office add-ins and Explorer integration ensure a label is always applied before a file leaves the desktop.
  • At the gateway: The Guard evaluates the label against both domain SPIFs before any cross-domain transfer proceeds.
  • At decryption: The Key Access Service re-evaluates the ABAC policy and the requester's clearance attributes in real time before releasing the Data Encryption Key. Even if an encrypted file is forwarded to an unauthorised party, the KAS will refuse the key request.
Attribute-Based Access Control (ABAC)

Traditional access control asks: "Is this user on the access list?" ABAC asks: "Do this user's attributes satisfy every condition in the policy?"

A user's attributes — their classification clearance, their category authorisations (compartments, COIs, releasability markings), their nationality — are evaluated against the ABAC policy embedded in the .ztdf archive. Every condition must be satisfied simultaneously. A TOP SECRET clearance alone does not grant access to a compartmented document; the compartment authorisation must also be present.

Critically, this evaluation happens fresh on every decryption request — there is no cached approval, no pre-authorised user list, and no way to pre-position a key. If a user's clearance is revoked or a category authorisation is withdrawn, the next access attempt is denied, regardless of whether they successfully decrypted the same file before.

From Paper to Protocol

The NATO standards suite translates the physical-world model into a concrete, interoperable electronic implementation:

STANAG 4774

Confidentiality Metadata Labels

Defines the XML schema for a Confidentiality Label — the electronic equivalent of the classification stamp. Specifies how classification level, categories, and policy references are encoded in a structured, parsable format.

STANAG 4778

Binding Data Object (BDO)

Defines how a Confidentiality Label is cryptographically bound to a specific document. The Binding Data Object (BDO) is a CMS SignedData structure that includes the label, a hash of the content, and a digital signature — making tampering detectable.

ADatP-5636

Object Classification Label (OCL)

Provides the structured handling instructions that accompany a Confidentiality Label — the JSON OCL record whose portioning marking string is displayed to users and printed in document headers and footers.

ACP-240

Zero Trust Data Format

The ACP-240 standard defines the .ztdf encrypted archive format that packages payload encryption, ABAC access policy, and STANAG label binding into a single, self-describing container that enforces policy at the point of decryption.

Taken together, STANAG 4774 labels what the information is, STANAG 4778 proves the label is authentic and unaltered, ADatP-5636 tells humans how to handle it, and ACP-240 encrypts the content so that only a KAS — after verifying all three — will release the decryption key.