An identity provider (IdP) is a system or service that performs authentication and supplies information about an authenticated subject to other systems. These systems, often called relying parties or service providers, use the information to establish sessions, associate users with accounts, and inform access decisions. An identity provider may serve employees within an organization, customers of an online service, or members of a federation spanning independent institutions. Its defining role is to vouch for an identity within an established trust relationship—not necessarily to prove a person’s legal identity.
Role in identity systems
An identity provider separates the process of authenticating a user from the applications the user accesses. Instead of each application collecting and verifying its own passwords, an application accepts an authentication result from a trusted provider.
Three roles commonly participate:
- Subject: The person or other entity being authenticated.
- Identity provider: The system that authenticates the subject and issues identity information.
- Relying party: The application or service that accepts that information under an agreed protocol and trust configuration.
In Security Assertion Markup Language (SAML), the receiving application is usually called a service provider. In OpenID Connect (OIDC), it is called a relying party, and the identity provider is formally an OpenID Provider.
An identity provider commonly maintains or connects to a directory containing account identifiers, authentication credentials, and attributes such as organizational affiliation. However, a directory and an identity provider are not the same thing: a directory stores and retrieves identity records, whereas an identity provider participates in authentication and issues results that another system can verify.
Authentication is also distinct from authorization. A provider may establish that a user controls a particular account and report the user’s group membership, but the application generally remains responsible for deciding which resources that user may access.
How authentication works
A typical browser-based federated sign-in proceeds as follows:
- A user attempts to access an application.
- The application directs the user’s web browser to a trusted identity provider with an authentication request.
- The provider authenticates the user, or recognizes an existing session that satisfies the request.
- The provider returns an authentication response, or a temporary code that the application exchanges for tokens.
- The application validates the response and associates its subject identifier with a local account.
- The application establishes its own session, often using an HTTP cookie.
The exact messages and validation rules depend on the protocol. The provider’s response typically identifies the issuer and subject, specifies the intended recipient, and includes validity information. It may also carry claims about the authentication event or the user.
These exchanges normally use Transport Layer Security to protect communications. Assertions or tokens may additionally use a digital signature, allowing the receiving system to verify their integrity and origin. A valid signature alone is insufficient: the application must also check the expected issuer, intended audience, expiration, and other protocol-specific conditions.
The identity provider normally does not disclose the user’s password to the application. Nevertheless, the application must still protect its own session credentials and any tokens it receives.
Federation and single sign-on
Identity federation allows one security domain to accept identity information issued by another. Trust may be established through bilateral configuration, a federation operator, or another arrangement for distributing trusted metadata and verification keys.
Federation does not require one universal account. An employee’s organizational account, for example, may be accepted by independently operated business applications, while those applications retain separate local profiles and permissions.
Single sign-on (SSO) is a related but distinct capability: a user authenticates once and subsequently accesses multiple applications without repeatedly presenting credentials. An existing identity-provider session can support this behavior, although a sensitive application may request fresh authentication or a stronger authentication method.
Federation can exist without a seamless SSO experience, and SSO can operate entirely within one organization. The two concepts therefore should not be treated as synonyms.
Identity brokering adds an intermediary. A broker may accept authentication from several upstream providers and present a consistent interface to downstream applications. It acts as a relying party toward the upstream provider and as an identity provider toward the downstream application.
Principal protocols
SAML
SAML is an XML-based standard for exchanging security assertions. SAML 2.0 was standardized in 2005 and became widely used for enterprise and institutional web SSO.
A SAML assertion can describe an authentication event, attributes, and related conditions. Browser SSO profiles specify how authentication requests and responses move between the provider and service provider. Configuration commonly includes entity identifiers, service endpoints, certificates, and rules governing accepted assertions.
OpenID Connect
OpenID Connect, finalized in 2014, defines an identity layer on top of OAuth 2.0. It allows an application to obtain an authentication result and identity claims through standardized flows.
Its central identity artifact is the ID token, represented as a JSON Web Token (JWT). An ID token identifies the issuer and subject, names its intended audience, and contains time-related claims. Other claims may describe the authentication event or user attributes. Additional user information may be available through a UserInfo endpoint.
The authorization-code flow is commonly used with Proof Key for Code Exchange (PKCE), which binds an authorization request to the later code exchange and helps protect against authorization-code interception and related attacks.
Relationship to OAuth
OAuth 2.0 is primarily an authorization framework for obtaining limited access to protected resources. It does not, by itself, define a general-purpose user-authentication protocol.
An OAuth access token and an OIDC ID token serve different purposes:
- An access token authorizes access to a resource server.
- An ID token communicates an authentication result to a particular client.
A deployment may combine the identity-provider and OAuth authorization-server roles in one service, but that operational combination does not eliminate the distinction.
Identity claims and assurance
Identity information is communicated through claims or attributes. Examples include a subject identifier, display name, email address, department, or group membership. Their significance depends on the issuer’s practices and the relying party’s trust policy.
Three processes must be distinguished:
- Identity proofing: Establishing a relationship between an account and a claimed real-world identity.
- Authentication: Establishing that a claimant controls one or more authenticators associated with an account.
- Federation: Conveying an authentication result and related information to another system.
A provider can authenticate an account reliably without having verified its holder’s legal name. Likewise, an email-address claim does not necessarily establish a person’s legal identity or continued ownership of that address.
Providers may support passwords, cryptographic authenticators, and multifactor authentication. Applications can use supported protocol mechanisms to request or evaluate authentication strength, but assurance claims require an agreed interpretation. Merely receiving a field labeled “MFA” is not a complete description of the methods used or their resistance to attack.
Account recovery is also part of the effective security boundary: weak recovery procedures can undermine otherwise strong authentication.
Account and session lifecycle
An identity provider is often part of a broader identity-management system, but authentication does not automatically create, update, or remove application accounts.
Provisioning may occur in advance, at first sign-in, or through System for Cross-domain Identity Management (SCIM). Applications may map received attributes to local roles, although those mappings remain separate policy decisions.
Disabling an account at the identity provider usually prevents future authentication through that account. It does not necessarily terminate:
- Existing application sessions.
- Previously issued access tokens.
- Application-specific credentials.
- Access paths that bypass the provider.
Logout is similarly distributed. Signing out of one application may leave the identity-provider session active, while signing out of the provider may not immediately invalidate every application session. Coordinated logout depends on protocol support, implementation, and session policy.
Security, privacy, and operational limits
Centralizing authentication can reduce duplicated credential handling and make authentication controls more consistent. It also concentrates risk. Compromise of a widely trusted provider, its signing keys, or its administrative accounts can affect many connected applications. An outage can prevent new sign-ins across otherwise functioning services.
Important failure modes include accepting tokens issued for another application, inadequate issuer validation, replaying authentication responses, unsafe account linking, and misinterpreting claims. Phishing remains relevant when users can be diverted to fraudulent sign-in pages or tricked into disclosing credentials.
Identity providers also occupy a significant position in data privacy. A provider may observe authentication requests across multiple services, while released attributes can enable correlation or reveal unnecessary personal information. Pairwise subject identifiers can give the same user different identifiers at different relying parties, reducing straightforward cross-service correlation. They do not prevent every form of tracking or disclosure.
Operational trust includes more than protocol compatibility. Participating systems must agree on identifier stability, attribute meaning, authentication assurance, key rotation, account recovery, and incident handling. Two products can implement the same protocol yet remain incompatible in practice because their configurations or identity semantics differ.
An identity-provider assertion is therefore evidence within a defined trust framework, not a universal guarantee that an account is legitimate, a device is safe, or a requested action should be permitted.