Skip to content
IAM Insights

Glossary

Our research sits where cybersecurity meets AI, and that intersection mints acronyms faster than anyone can absorb them. This page defines every term we use, in plain English, and each term's page lists the briefings that cite it. Across the site, dotted-underlined terms link to those pages — hover for the short version.

Some terms also have explainers, exploring how they work and why they matter. Look for Explainer →.

795 terms

Where security meets AI

The vocabulary being invented right now — identity concepts that only exist because AI agents do.

A-Commerceagent commerce

Shopping done by AI agents on a person's behalf; a term coined by the Customer Futures newsletter.

AAuth (Hardt)AAuth · AAuth Protocol · draft-hardt-oauth-aauth-protocol

An individual Internet-Draft that defines the AAuth protocol for agent-to-resource authorization, using signed requests from a cryptographic agent identity instead of bearer tokens.

AAuth agent providerAP

In AAuth, the service that vouches for an agent by signing its identity token.

AAuth Budgetsdraft-hardt-aauth-budgets

An individual IETF draft by Dick Hardt that lets an access token carry a spending ceiling the resource enforces as a hard cap, so an agent can buy metered services without seeing the person's overall limit.

AAuth Rich Resource Requests (R3)draft-hardt-aauth-r3 · AAuth R3

An individual Internet-Draft from the Internet Engineering Task Force (IETF) that lets an AI agent ask for access to named operations of an API, described in a document the resource publishes and identifies by hash.

Action classaction classes · r/w/x class

A label for what an API call does: read-only, changes only the caller's own data, or reaches beyond the caller or cannot be undone.

Action-spaceaction space

The range of actions an AI agent can take, the tools, systems and permissions it is allowed to use, as distinct from its autonomy (how much it decides for itself); the two together set an agent's risk profile.

Agent Audit Traildraft-sharif-agent-audit-trail

An individual IETF draft defining a hash-chained log record for each action an AI agent takes.

Agent Authorization Envelope (AAE)draft-kroehl-agentic-trust-aae · Agent Authorization Envelope · AAE

Individual Internet-Draft defining a JWS-secured W3C Verifiable Credential with mandate, constraints and validity blocks for agent authorization.

Agent Authorization Profile (AAP) for OAuth 2.0draft-aap-oauth-profile · OAuth AAP

An individual IETF draft defining structured claims and validation rules for AI agents in OAuth 2.0 and JWT; revision -01 expired in August 2026.

Agent Authorization Use Cases and Gap Analysisdraft-chen-oauth-agent-authz-use-cases

An individual IETF informational draft that walks eleven AI-agent scenarios through OAuth and lists where it falls short.

Agent CardAgent Cards · A2A Agent Card · AgentCard

The JSON document an A2A agent publishes to describe itself: name, endpoint, skills, supported transports and how callers must authenticate. Anyone who fetches it sees all of it unless it is protected.

Agent checksumagent_checksum · agent checksums

A Secure Hash Algorithm 256 (SHA-256) fingerprint of an AI agent's system prompt, tool definitions and model settings, which an identity server can compare with a registered value before issuing a token.

Agent delegationdelegated authority

One party acting with another's authority: a person lets an agent act for them, and that agent may hand part of the job to a sub-agent.

Agent hijackingAI agent hijacking

Taking over an AI agent, usually by planting instructions in content it reads, so it acts for the attacker.

Agent identityAI agent identity · AIdentity

A unique, verifiable identifier assigned to an AI agent itself, distinct from the human or system it acts for, so the agent can authenticate, be tied to an accountable owner, and have its actions tracked and revoked like any other identity.

Agent Identity Management SystemAI Identity Management System

Agent Identity Management System: the model an Internet Engineering Task Force (IETF) draft defines for identifying AI agents, giving them credentials and authorizing them, built on Workload Identity in Multi System Environments (WIMSE).

Agent kill switchkill switch

An immediate, whole-identity revocation for an autonomous AI agent, cutting it off the instant it looks compromised or malicious rather than waiting out a session timeout a fast-acting agent could exploit in the meantime.

Agent Operation Authorization (draft-liu-agent-operation-authorization)Agent Operation Authorization · draft-liu-agent-operation-authorization

Individual IETF draft from Alibaba/Cisco authors for AS-issued JWT tokens carrying user-confirmation evidence and an OPA policy reference for agent operations.

Agent PKI (APKI)draft-sharif-apki-agent-pki · APKI · Agent Public Key Infrastructure

Agent Public Key Infrastructure: an individual IETF proposal extending X.509 certificates with agent capabilities, trust scores, delegation constraints and model provenance.

Agent registryagent registration · centralized agent registration

A single inventory of the AI agents an organisation runs, recording which agents exist, who owns each one and what it may do.

Agent sprawlagentic sprawl · genteck sprawl · AI agent sprawl

The uncontrolled growth in number of AI agents created by employees across platforms, many unused or unowned, which outpaces the organisation's ability to inventory and govern them.

Agent Transfer Protocol (AGTP)draft-hood-independent-agtp · AGTP · Agent Transfer Protocol

An individual Internet-Draft for a transport-level protocol for agent-to-agent communication, with carriers for agent identity, delegation, session state and audit evidence.

Agent traps

Websites, content or crafted inputs built to subvert an AI agent, for example to make it ignore its instructions, leak private data or spend money it should not.

Agent Trust eXtensionATX

In the OpenA2A proposals, a registry-signed credential attesting to an agent’s identity, code integrity, capabilities and trust level.

Agentic control planeagent control plane

Okta's term for one place to discover, govern and secure AI agents.

Agentic identity

Giving AI agents first-class identities — who they act for, what they are allowed to do, and how that delegated authority is verified and revoked.

Agentic JWTAJWT · agentic JWT

A proposed JWT extension that binds an AI agent's specific intent — the task it was delegated — into the access token, so services can verify not just who an agent acts for but what it was authorized to do. An arXiv preprint and IETF individual draft, not a ratified standard.

Agentic payments

Payments initiated by an AI agent rather than a person at checkout; the card networks and payment processors are defining how a merchant knows the agent is legitimate and what the person actually authorised.

Agentic transaction indicatorAgentic Transaction Indicators · agentic indicators · agent indicator · agent-initiated indicator

A flag or field in a payment message telling banks, networks and merchants that an AI agent started or carried out the transaction for a consumer.

agentic_ctx claimagentic_ctx · agentic context claim · agentic context propagation

A JWT claim the transaction-tokens-for-agents draft adds to a transaction token to carry chain-level agent facts: the current agent, the originating agent, and a hop count.

AgentPassAgent Passport

CyberSecAI's open-source Trust Authority for AI agents, which rates and vouches for an agent's trust; its reference code ships as the @proofxhq/agentpass package.

AI agentAI agents · agentic AI · autonomous agent · autonomous agents

Software that uses an AI model to decide its own next steps and then acts through tools and APIs, rather than answering a single prompt. Because it acts, it needs an identity of its own.

AI Agent Authentication and Authorizationdraft-klrc-aiagent-auth · draft-klrc-aiagent-auth-01

A proposed IETF Internet-Draft on authentication and authorization patterns for AI agents acting on a user's behalf, since replaced by the WIMSE working group's draft-ietf-wimse-aims. A draft, not a standard.

AI-BOM

AI bill of materials: an inventory of the models, datasets and components inside an AI system, like an SBOM for software. It lets teams check where an agent's parts came from and what it is built on.

ANSANS v2 · Agent Name Service · Agent Naming Service · draft-narajala-ans · draft-narajala-courtney-ansv2

A proposed scheme for naming and verifying AI agents by a version number and a domain the publisher controls, checked by a registration authority. An individual IETF Internet-Draft, not a standard.

ASTRAAuthorization with Semantic Task-based Restricted Access

A public dataset and data-generation pipeline (Cisco/Outshift, Oct 2025) for benchmarking semantic matching between an AI agent's task and the OAuth scopes it requests, including deliberately wrong and null scope matches.

Attenuated Delegation Profile for Automated Agentsdraft-hamr-oauth-agent-delegation · Agent-Delegation header · Attenuated Delegation Profile

An individual IETF draft that carries a chain of delegation links in an HTTP header and requires the verifier to check that each link is no wider than its parent.

Attenuating Authorization Tokenattenuating agent token

A signed credential an agent can narrow offline into a child token with equal or fewer tools and tighter argument limits.

Attenuating Authorization Tokens for Agentic Delegation Chainsdraft-niyikiza-oauth-attenuating-agent-tokens

An individual IETF draft for OAuth tokens that an agent can narrow before passing them on.

Attestationprovenance attestation

A signed statement that something is what it claims to be: which model an agent runs, what code it was built from, or which human owns it.

AUDIT BoF (Agent Use of Delegation and Interaction Traceability)AUDIT BoF · AUDIT working group

Approved Internet Engineering Task Force (IETF) Birds of a Feather (BoF) request for a session on tamper-evident, cross-party records of what an agent did, for whom, and under what authority.

Authenticated workflows

Signing every inter-entity agent invocation with its policy so a receiving enforcement point can verify authenticity before executing.

AuthZEN Access Request and Approval ProfileAARP · Access Request and Approval Profile

AuthZEN profile defining how applications, agents and governance platforms request, track, satisfy and re-evaluate prerequisites (approval, consent, attestation) when policy cannot yet authorize an action.

Client ID Metadata DocumentsCIMD · Client ID Metadata Document

A draft OAuth method letting a client identify itself by publishing a metadata document at an HTTPS URL it controls, instead of registering with each authorization server first.

COAZCOAZ-MCP

AuthZEN Profile for Model Context Protocol Tool Authorization (COAZ): an OpenID Foundation working group draft that maps source information models, such as MCP tool calls, onto AuthZEN's subject, action, resource and context structure.

Context-aware authorization

Authorizing an agent's action using the task, trust relationships and policy context, not only the instruction text.

Credential brokercredential brokering · credential vaulting and brokering · secrets broker

A service that holds real credentials on an agent's behalf and hands out short-lived or proxied access instead, so the agent never keeps a long-lived key.

Credential owneraccountable owner

The individual or organization accountable for an agent and able to countersign its credential request.

Delegate SD-JWTdSD-JWT · dSD-JWT+KB · Delegated SD-JWT · draft-gco-oauth-delegate-sd-jwt · draft-gco-oauth-delegate-sd-jwt-00

An Internet-Draft that lets the holder of a Selective Disclosure JSON Web Token (SD-JWT) delegate it onward, so a verifier can walk the chain back to the issuer.

Delegated vault tokenvault token · delegated payment token

In the Agentic Commerce Protocol, a token a merchant's payment processor issues for one card that an agent platform may use only inside a stated allowance: one merchant, one checkout session, a maximum amount and an expiry.

Delegation chain

An ordered list of tokens from a root grant to the token presented, each derived from the one before.

Delegation Chain for OAuth 2.0 (draft-liu-oauth-chain-delegation)draft-liu-oauth-chain-delegation · delegation_chain claim

Individual Internet-Draft defining a delegation_chain JWT claim that records each delegation hop, AS-signed and optionally delegator-signed, beside the act claim.

Delegation Grant (DG)Delegation Grant · Delegation Grants

A signed, revocable authorization artefact, distinct from a VC, that transfers bounded authority and may only narrow scope along a chain.

Delegation Server

A service that holds a user's credentials in a vault, issues delegation tokens to AI agents, enforces attenuation and revocation, and proxies the agents' calls.

Direct prompt injectiondirect prompt injection · direct prompt injections

Prompt injection where the attacker edits the instruction the user gives the agent, rather than text the agent later reads from a tool or document.

Dual LLM patterndual-LLM · dual LLM architecture

A prompt-injection defense that splits a privileged model, which holds the tools, from a quarantined model, which reads untrusted content and never calls tools; proposed by Simon Willison in 2023.

Enforcement boundaryenforcement point

The first component on an action's path that actually refuses or permits it; a component that only advises does not count.

Enhanced Tool Definition Interface (ETDI)ETDI

Proposed MCP extension: cryptographically signed tool manifests, immutable versions, and OAuth-scoped tool definitions to stop tool squatting and rug pulls.

ERC-8004ERC 8004

A proposed Ethereum token standard for an on-chain trust layer for AI agents: identity, reputation, and validation registries anchored to a blockchain so one organization's agents can be identified and vetted by another's.

Human Delegation Provenance (HDP, draft-helixar-hdp-agentic-delegation)Human Delegation Provenance · HDP protocol · HDP

Helixar's token scheme recording a human authorization and signed agent hops

Human in the loop (HITL)HITL · human-in-the-loop

A design where a person must approve certain agent actions before they run.

IBACintent-based access control

Intent-based access control: a research design that checks each AI-agent tool call against an authorization store derived from the user's stated intent before the call runs.

Identity debtidentity technical debt

Technical debt of the identity kind: past choices (SMS codes, voice biometrics, a different login product per business unit) that were fine for an earlier era and now leave an organisation exposed to AI-driven attacks.

Infocomm Media Development Authority (IMDA)IMDA · Infocomm Media Development Authority

Singapore's converged infocomm, media and digital-economy regulator; publishes the Model AI Governance Framework series that many enterprises use as a de facto AI governance checklist, most recently the 2026 MGF for Agentic AI.

Intent bindingintent digest · spt_intent_digest

Tying a token to a hash of the exact tool, parameters and target it was issued for, so the checking point denies any call that differs.

Intent Provenance Protocol (draft-haberkamp-ipp)Intent Provenance Protocol

Draft for Ed25519 append-only intent provenance chains

Intent ServiceIntent Services · Intent Service Provider · Intent Service Providers

In EMVCo's agentic-payments framework, a hosted service run by a payment system or issuer that records what a consumer authorised an agent to buy, gives it a stable reference and tracks its state. It does not decide whether a payment is approved.

Know Your AgentKYA · Know-Your-Agent

The ability of a bank, card network or merchant to identify the AI agent in a payment and tie it to stable, shared details such as an identifier, version and registry entry, so the payment can be risk-scored, audited and disputed. Named after the banks' Know Your Customer duty.

KYA-OSMCP-I · KYA-OS Protocol · Know Your Agent OS

An open specification, contributed by Vouched to the Decentralized Identity Foundation as MCP-I and renamed in 2026, that gives AI agents a Decentralized Identifier and uses verifiable credentials to prove which person delegated what to the agent.

Least agencyLeast Agency

OWASP term extending least privilege to agents: restrict what each agent tool can do, how often and where, not only what it can access.

LOKA ProtocolLOKA

Proposed Universal Agent Identity Layer plus Decentralized Ethical Consensus Protocol for agent ecosystems (arXiv 2504.10915).

MAESTROMAESTRO framework · Multi-Agent Environment, Security, Threat, Risk, & Outcome

A threat-modeling framework from the Cloud Security Alliance for agentic AI that examines risk across seven layers, from the foundation model up to the agent ecosystem.

MCP Guardian

Research layer for MCP systems that adds user authentication, rate limiting and web application firewall protections (Kumar et al. 2025).

MCPS (Sharif)draft-sharif-mcps-secure-mcp · MCPS · MCP Secure

An individual Internet Engineering Task Force (IETF) draft, MCPS (Cryptographic Security Layer for the Model Context Protocol), that adds per-message signing, replay protection and tool-integrity checks to MCP.

Memory poisoningmemory poisoning · knowledge base poisoning · RAG poisoning · knowledge-base poisoning

Planting malicious entries in an agent's long-term memory or retrieval store so they are fetched later and steer its actions.

MI9

A runtime governance framework for AI agents, from a 2025 arXiv paper, that keeps watching an agent after approval through telemetry, authorization monitoring, drift detection and graduated containment.

Microsoft CopilotM365 Copilot

Microsoft's AI assistant built into Microsoft 365 apps such as Word, Outlook and Teams. It works with the files and data the signed-in user can already access.

Mission shapingMission Shaping Problem · Mission Proposal · Approved Mission · Mission Authority Model

In Karl McGuinness's Mission-Bound OAuth framework, the step that turns a human's approved intent into a machine-readable authority artifact for policy engines. It comes before authorization, not instead of it.

Mission-Bound OAuth

A multi-part framework proposed by IAM practitioner Karl McGuinness for scoping an AI agent's OAuth-issued authority to a specific, human-approved "Mission" — a bounded, versioned task grant — rather than a broad standing scope.

OASNT-CAIDdraft-thallapelly-oasnt-caid · OASNT · draft-thallapelly-oasnt

An individual Internet Engineering Task Force (IETF) draft that defines how to derive a Canonical Action Identifier (CAID) from an OASNT (Attested Action Authorization Token) and bind it to a named human.

OAuth Profile for Delegated AI Agent Authorization (DAAP)draft-mishra-oauth-agent-grants · DAAP

An individual IETF draft profiling OAuth for AI agents, with user consent, sender-constrained tokens and narrowing through token exchange.

OpenA2A Agent Identity Protocoldraft-fane-opena2a-aip · OpenA2A AIP

An individual IETF draft giving AI agents an Ed25519 key, a did:web identifier and a registry-computed trust score.

OpenID Connect for AgentsOIDC-A

A proposal (an arXiv paper, not an OpenID Foundation specification) to extend OpenID Connect with claims, capability metadata and discovery for LLM-based agents.

OpenID Foundation AI Identity Management Community GroupAI Identity Management Community Group · Artificial Intelligence Identity Management Community Group · AIIM CG

The OpenID Foundation community group whose members produced the Identity Management for Agentic AI whitepaper.

OWASP Top 10 for Agentic Applications 2026OWASP Agentic Top 10 · Agentic Top 10 · ASI Top 10 · OWASP ASI Top 10

OWASP's December 2025 list of the ten highest-impact security risks for agentic AI applications (ASI01-ASI10), from goal hijack to rogue agents.

par_hashpar_hash claim

A token claim holding a Secure Hash Algorithm 256 (SHA-256) digest of the parent token's signing input, binding a child token to one parent instance.

Per-call authorizationper-call proposal · per-call access mode · r3_per_call

Authorizing each individual invocation of an operation against its concrete parameters, with a single-use grant, rather than granting the operation in advance.

Person serverPS

In the AAuth drafts, the server that stands for the person behind an agent: it issues the agent's tokens, asks the person for consent and holds their limits.

Personhood credentialPHC · PHCs · personhood credentials

A digital credential that proves its holder is a real, unique person without saying who they are; one issuer gives each person only one.

Plan-then-execute patternplan-then-execute

A prompt-injection defense where the agent fixes its list of tool calls before reading any untrusted data, so injected text cannot add new actions, though it can still change what the planned calls carry.

Prompt injectionprompt-injected · prompt manipulation · indirect prompt injection

Hiding instructions in text an AI model reads, such as an email or a web page, so the model does something its operator never asked for.

Proof of personhoodproof of human · proof-of-human · proof of person

Showing that an account belongs to a unique living person, without necessarily revealing which person.

PSEA Token Profile: An EAT Profile for Action-Bound, User-Verification-Gated Transaction-Confirmation Evidencedraft-yossif-psea · PSEA · PSEA Token Profile

An individual Internet Engineering Task Force (IETF) draft defining an Entity Attestation Token (EAT) profile for signed evidence that a human, verified at an authenticator, approved one specific action.

Runtime authorityagentic runtime authority · runtime identity

Akeyless term: authorise each agent action at execution time, with intent evaluated and a credential minted per request, instead of granting a session-long token.

Runtime authorizationruntime enforcement · runtime authorisation

Deciding whether to allow each individual action at the moment it happens, using the current context and purpose, instead of granting broad access in advance and reviewing it later.

SCIM Agent Governance Extension (draft-kushwaha)draft-kushwaha-scim-agent-governance-00 · SCIM Agent Governance Extension

An individual IETF draft (Oracle's Saurabh Kushwaha, July 2026) that adds governance metadata to the SCIM Agent resource: a five-state lifecycle, an autonomy level, a validity window and a pointer to credential discovery.

SCIM Agentic Identity Schemadraft-wahl-scim-agent-schema · AgenticIdentity · SCIM agent schema

An individual IETF draft that adds an AgenticIdentity resource type to SCIM so AI agents can be provisioned and removed like employees, with owners and group memberships.

SCIM Agents and Agentic Applications Extensiondraft-abbey-scim-agent-extension · draft-abbey-scim-agent-extension-00 · Abbey and Cohen draft

An expired IETF individual draft (October 2025, authors Macy Abbey and Okta's Rafael S. Cohen) that modelled an AI agent as a property of an 'agentic application' with owned, authorized or guest relationships, rather than as its own SCIM resource.

SCIM AI Agent Resource Extensiondraft-wzdk-scim-agent-resource · draft-wzdk-scim-agent-resource-00 · wzdk draft · AI Agent Resource Extension for SCIM

An individual Internet-Draft that defines a resource type for AI agents in the System for Cross-domain Identity Management (SCIM), so agents can be provisioned and deprovisioned between systems the way user accounts are.

Scope attenuationattenuation of scope · recursive delegation · offline attenuation

Narrowing the permissions at every hand-off, so a sub-agent that receives a task gets only what that task needs rather than everything its parent held.

SD Agent (SD-Card)SD-Card · SD-Cards · SD Agent · draft-nandakumar-agent-sd-jwt · draft-nandakumar-agent-sd-jwt-02

An expired 2026 Internet-Draft from the Internet Engineering Task Force (IETF) that encodes an Agent2Agent (A2A) Agent Card as a Selective Disclosure JSON Web Token (SD-JWT), so callers see only some capabilities.

Shadow AI

AI tools and agents that employees adopt without the security team knowing. It is the AI-era version of shadow IT.

Shadow MCP servershadow MCP servers · shadow MCP

A Model Context Protocol server a developer installs locally without security review or central inventory, running with that developer's permissions and credentials (OWASP MCP Top 10 item MCP09).

Skill-based prompt injectionskill injection · SkillInject · Skill-Inject

Prompt injection carried in a third-party agent skill file, where malicious lines hide among legitimate instructions.

Standing authority

Authority to permit a class of actions that a person holds before an oversight act, conferred by a source other than that act.

Task-scoped authorizationtask-scoped authorization · Precise Task-Scoped Implicit Authorization

Authorization derived from the user's concrete task, so each call (operator plus operands) is checked against what the task implies rather than against a broad scope.

Token passthroughtoken pass-through

A server forwards the access token a client sent it, unchanged, to a downstream API. Model Context Protocol (MCP) authorization forbids this because the token was issued for the server, not the downstream service.

Tool filtertool filtering

A prompt-injection defence that fixes which tools an AI agent may use for a task before it reads any untrusted data.

Tool poisoningtool poisoning attack · poisoned tool · poisoned tools · tool-poisoning

Hiding malicious instructions in the description or parameters of a tool an AI agent can call, so the model is steered into leaking data or taking actions its user never intended.

Tool squattingtool name squatting · package name squatting

Publishing a tool or server whose name or description mimics a legitimate one so an AI agent selects the attacker's by mistake.

TPX (Token Pony Express)TPX · Token Pony Express

An OAuth 2.0 profile that lets a person grant an app a metered LLM inference budget described as a damage cap, not a payment.

Transaction Tokens for Agentsdraft-araut-oauth-transaction-tokens-for-agents

An individual IETF draft that adds an actor chain and agent context claims to OAuth transaction tokens.

Transaction-Bound Authorization Tokens for Software and AI Agents (SPT-Txn)draft-coetzee-oauth-spt-txn-tokens · SPT-Txn

An individual IETF draft that mints a short-lived token for one declared action on one resource, chained from a root through narrowing delegations and checked offline.

Trust FabricTrust Fabric

A five-layer design for an open agent ecosystem: identity and discovery, access control, enforcement, evaluation, and economic incentives.

Trusted Surfacetrusted surfaces

In AP2, the one component that must not be an AI agent: the screen or device a person actually sees and approves on, which then signs the mandate. It is where informed consent is supposed to happen.

UMA for Agents (U4A)U4A · UMA for Agents · UMA4Agents · UMA 4 Agents

An early open-source profile and extension of UMA for third-party agents: an agent that belongs to someone else asks a resource owner's authorization server for access. The resource owner sets the conditions, not the agent's owner.

Unified Identity GovernanceUIG

A proposed architecture for governing human and non-human identities — workloads and AI agents included — under one framework: unified registry, policy enforcement, monitoring, and tiered remediation. From a 2026 IJFMR practitioner paper, not a shipping standard.

Verifiable Attenuated Delegation for AI Agent Chainsdraft-asor-wimse-agent-delegation-chain · Verifiable Attenuated Delegation Chains for AI Agents

An individual IETF draft that chains signed JWTs, one per hop, so a verifier can check offline that each agent's authority is a subset of the one that delegated to it.

Verifier-Side Evaluation Semantics for Delegated Authority Chainsdraft-jackson-wimse-evaluation

An individual IETF draft proposing fail-closed rules for how a verifier judges a presented delegation chain.

Web Bot Authweb bot auth

A proposed web standard, backed by Cloudflare, that lets an automated client prove which bot it is by signing its requests, so sites can let trusted agents in and keep others out.

Workload Authorization Grantdraft-carleton-workload-authz-grant · draft-carleton-workload-authz-grant-01 · Workload Authorization Grant

A proposed JSON Web Token (JWT) bearer grant by which a platform-signed assertion naming one workload is accepted by an authorization server configured once to trust that platform.

World IDWorldcoin · World App

Tools for Humanity's proof-of-personhood system: an iris scan at an Orb device issues a credential that a person is a unique human, which the holder can show online.

ZeroIDHighflame ZeroID

Highflame's open-source identity service for AI agents: each agent has its own identifier, and delegation narrows scope at each hop using OAuth token exchange.

Identity & security

Established identity-security standards, disciplines, and the bodies that steward them.

3GPP3rd Generation Partnership Project

The standards group for mobile networks (4G, 5G and the 6G studies), which writes the specifications that operators and handset makers build to.

AAGUIDauthenticator attestation GUID

A 128-bit identifier a FIDO authenticator reports for its make and model, which lets a site allow or block particular passkey providers or security keys.

ABACPBAC · attribute-based access control · policy-based access control

Access decisions made from attributes of the requester, the resource, and the moment (department, data sensitivity, risk score) rather than from a fixed role. PBAC is the same idea written as policy rules.

Access control listACLs · access control lists

Access control list: the stored record of which users or groups may perform which actions on a resource.

Access reviewuser access review · UAR · UARs · access certification

The periodic check, usually by a manager or system owner, that each person or account still needs the access it holds.

Access tokenaccess tokens

An OAuth 2.0 credential, usually an opaque string, representing a specific scope and lifetime of access that the resource server enforces when a client presents it.

Account recovery

The process for regaining access after losing a sign-in factor. Attackers aim at it because it is often weaker than the sign-in it replaces.

ACMEAutomatic Certificate Management Environment · ACME DNS-01 · ACME HTTP-01

The Automatic Certificate Management Environment (RFC 8555), the protocol behind Let's Encrypt: a certificate authority checks that a requester controls a domain, by a DNS record or an HTTP file, then issues a certificate automatically.

act claimactor claim · act chain · actor_token

The OAuth token-exchange claim that names who is acting for the token's subject; nested act claims record earlier actors.

Active Directory

Microsoft's directory service for users, computers and groups on Windows networks, used for sign-in and access. Entra ID is Microsoft's separate cloud identity service.

Adaptive attackadaptive attacks

An attack designed after the defense is fully disclosed, with the attacker knowing the defense and tuning to defeat it; the standard test of whether a defense is robust.

Adversary-in-the-middleAitM · AiTM · adversary in the middle · adversary-in-the-middle

A phishing attack where the attacker's server sits between the victim and the real sign-in page, passing everything through so it can capture the password, the one-time code and the session that results.

AESAdvanced Encryption Standard

A symmetric block cipher standardized by NIST, using a 128-bit block and 128-, 192- or 256-bit keys.

AES-256AES-256-GCM

The Advanced Encryption Standard with a 256-bit key, a symmetric cipher for protecting data at rest and in transit.

Amazon Verified PermissionsAVP

AWS's managed authorization service: an application asks whether a principal may take an action on a resource in a context, and the service answers allow or deny by evaluating Cedar policies.

API keyAPI keys · static API key · static API keys

A long-lived secret string that lets software call a service; anyone who copies it can use it until someone revokes it.

Assertion Framework for OAuth 2.0 Client Authentication and Authorization GrantsRFC 7521 · RFC7521

Generic framework for using assertions as authorization grants and client credentials

ATOaccount takeover

An attacker taking control of someone's existing account, usually with stolen or phished credentials, rather than breaking in through a software flaw.

Audience injection attackaudience injection attacks

Attack where a malicious authorization server publishes an honest server's token endpoint so a client's signed assertion can be replayed there.

Audience restrictionsingle audience · audience claim

Limiting a token to one named recipient through its aud claim, so the recipient cannot replay it elsewhere.

Authenticator assurance level (AAL)AAL · AAL1 · AAL2 · AAL3 · Authenticator Assurance Level

NIST SP 800-63 scale (AAL1 to AAL3) for how strongly a sign-in proves the user holds their authenticator, from a password alone up to a phishing-resistant hardware key. It sits beside the identity assurance level (IAL) and federation assurance level (FAL).

Authorization Code grantAuthorization Code Grant · Authorization Code flow · authorization code flow

The OAuth flow in which an app sends the user to sign in, receives a one-time authorization code, and exchanges it at the token endpoint for an access token.

AuthZENOpenID AuthZEN · Authorization API 1.0 · AuthZEN Authorization API

An OpenID Foundation working group standardizing how applications ask an external policy engine authorization questions (“may this user do that?”) instead of hard-coding the rules.

BBS signaturesBBS · BBS+

BBS (Boneh-Boyen-Shacham) signatures: a signature scheme that lets a holder reveal only some fields of a signed credential and keeps repeated uses from being linked to each other.

Bearer tokenbearer tokens · bearer credential

A token that works for whoever holds it, so a stolen copy can be replayed until it expires.

Behavioral analyticsbehavioural analytics

Spotting fraud or account takeover by comparing how a user types, moves and acts now with how they usually do.

Biometricsbiometric

Using a body trait, such as a face or fingerprint, to identify or verify a person.

Bitstring Status ListBitstring Status List v1.0 · W3C Bitstring Status List

A W3C Recommendation (May 2025) for publishing the revocation or suspension status of verifiable credentials as a compressed bit array, so a verifier checks one bit without the issuer learning which credential was checked.

Blast radiusblast-radius

The potential damage if an agent or component fails or is compromised.

BLS signaturesBLS · Boneh-Lynn-Shacham · draft-irtf-cfrg-bls-signature

A signature scheme whose signatures from many signers merge into one short value that still checks against all of them; it is not safe against quantum computers.

Bulk revocationmass revocation · bulk token revocation

Revoking every token and grant tied to one user, client or task in a single call instead of one token at a time.

BYODbring your own device

Staff using their own phones and laptops for work, so the company has to secure access from devices it does not own or fully manage.

C2PACoalition for Content Provenance and Authenticity

A content-provenance standard that attaches tamper-evident metadata to digital files so you can check where they came from and how they were changed.

CAEPContinuous Access Evaluation Profile

An OpenID standard for one system telling another, in real time, that a session should be re-checked or ended, for example when a device is found compromised.

CaMeL

Research system that tracks data provenance within one agent run to keep untrusted content from steering privileged actions.

Capability tokensBiscuit tokens · object-capability · OCap

Tokens such as Biscuits and Macaroons that carry their own limits, so the holder can hand on a narrower copy offline without asking the issuer.

CAPTCHACAPTCHAs

Completely Automated Public Turing test to tell Computers and Humans Apart (CAPTCHA): a challenge meant to tell a person from a bot, such as picking images or typing distorted text.

Cascade revocationcascade revocation · ancestor revocation walk

Revoking one credential in a delegation chain automatically denies every credential issued beneath it.

Cedar

AWS's open-source policy language and evaluation engine for fine-grained authorization decisions.

Certificate authorityCA · certificate authorities

An organization or service that signs digital certificates, vouching that a public key belongs to a named server, person or workload.

Certificate Management over CMSCMC

Certificate management protocol built on Cryptographic Message Syntax (CMS).

Certificate Management ProtocolCMP

Protocol for certificate lifecycle between end entities, registration authorities (RAs) and certificate authorities (CAs), from the PKIX (Public Key Infrastructure using X.509) standards.

Certificate revocation listCRL · CRLs

A signed list of revoked certificate serial numbers that a certificate authority publishes and verifiers download.

Certificate TransparencyRFC 6962 · CT log

A public append-only Merkle log of issued certificates that outsiders can audit.

Certificate Transparency Version 2.0RFC 9162

Experimental RFC defining CT v2 Merkle log structures and inclusion proofs.

Certificate-bound access tokencertificate-bound access token · certificate-bound access tokens · certificate-bound token · certificate-bound tokens · mTLS-bound access token

An OAuth access token tied to the TLS client certificate its holder used when getting it, so a stolen copy fails when presented without the matching private key. RFC 8705 carries the certificate's SHA-256 hash in the token's cnf claim.

CIAMCustomer Identity and Access Management

Customer Identity and Access Management — registration, login, and profile management for a company's customers rather than its workforce.

CIBAClient-Initiated Backchannel Authentication

Client-Initiated Backchannel Authentication — an OpenID flow where sign-in is approved out-of-band on another device (say, a push to your phone) rather than in the requesting browser.

CIEMCloud Infrastructure Entitlement Management

Tools that find and trim excess permissions in cloud accounts.

CISACybersecurity and Infrastructure Security Agency

The US federal agency for civilian cybersecurity defense and critical-infrastructure protection; publishes the standard definition of insider threat used across board-governance guidance.

Client assertionclient_assertion · client assertion authentication

A signed assertion a client sends to an OAuth token endpoint, in the client_assertion parameter, to prove its identity instead of sending a shared client secret.

Client AttesterClient Attester

In OAuth attestation-based client authentication, the backend party that vouches for a client instance and its key by issuing a signed Client Attestation.

Client credentials grantclient credentials flow

An OAuth 2.0 grant in which a confidential client authenticates with its own credentials and gets an access token for resources it controls or has pre-arranged, with no resource owner involved.

Client to Authenticator Protocol (CTAP)CTAP · Client to Authenticator Protocol

The Fast Identity Online (FIDO) protocol between a client, such as a browser, and an authenticator, such as a security key.

Cloud Native Computing FoundationCNCF

The vendor-neutral foundation that hosts Kubernetes and related cloud-native projects, including SPIFFE and SPIRE.

Cloud Security AllianceCSA

A nonprofit that publishes cloud-security research, best practices and certifications, including work on AI-agent identity.

CMSCryptographic Message Syntax

Cryptographic Message Syntax, the standard format (RFC 5652) for signing or encrypting data, used in S/MIME and in some attestation drafts. In web publishing it can also mean a content management system.

Command and controlC2 · C&C

The channel an attacker uses to send instructions to malware on an infected machine and receive its results.

Command injectioncommand ejection · OS command injection

Untrusted input reaches a shell or command line and runs as a command.

Common Weakness EnumerationCWE · Common Weakness Enumeration

MITRE's catalog of software weakness types, e.g. CWE-798 hard-coded credentials.

Confidential clientconfidential clients · confidential client

An OAuth client that can keep a secret, such as a server-side app, and so can authenticate to the token endpoint with a client secret.

Confidential Computing

Running a workload inside a hardware-based trusted execution environment so data in use stays isolated from the host.

Confirmation (cnf) claimcnf · cnf.jwk · confirmation claim

A field in a token naming the key whose holder may use it; chaining one token's cnf key to the signer of the next is how delegation chains are built.

Confused deputyconfused deputy attack · confused deputy attacks · confused deputy problem

A classic security flaw where a system holding legitimate elevated privileges is tricked by a less-privileged or malicious party into misusing those privileges on that party's behalf; applied to AI agents that hold real permissions but can be manipulated through their prompts or inputs.

Contextual integrity

Privacy framework: whether an information flow is appropriate depends on the norms of its context.

Continuous Access Evaluation

Revoking or re-checking access in real time when a risk signal fires, instead of waiting for a token or session to expire on its own; the general goal that standards such as OpenID CAEP and the Shared Signals Framework implement.

Control-flow integritycontrol flow integrity

A guarantee that a program only follows the steps it was meant to follow. For agents, it means injected text cannot add or reorder the tool calls the agent planned.

COSECBOR Object Signing and Encryption · RFC 9052

The compact binary counterpart of JOSE: how to sign and encrypt CBOR data, used in device attestation and mobile credentials.

Credential exchangeCredential Exchanger · credential exchanger

Moving a credential from one system or provider to another, such as a passkey between password managers, or trading a token from one trust domain for a local one.

Credential leakagecredential leakage · secret leakage

Exposure of authentication material to a recipient or channel the software neither declares nor needs.

Cross App AccessXAA · Cross Application Access · ID-JAG

Okta's OAuth extension that lets an app request access to another app for a user, with user, agent and resource app bound cryptographically. It uses ID-JAG, a JSON Web Token (JWT) authorization grant.

Cross-JWT confusioncross-JWT confusion

A substitution attack in which a JSON Web Token (JWT) issued for one purpose is accepted in another context; explicit typing, distinct claims, keys, audiences or issuers help prevent it.

Cross-site scripting (XSS)XSS · cross-site scripting

An attack where a site is tricked into running an attacker's script in a visitor's browser, which can then read tokens held in the page and act as the visitor.

CSOChief Security Officer · chief security officer

Chief Security Officer: the executive in charge of an organization's security. Often used as a variant of CISO, the chief information security officer.

CSPContent Security Policy

Content Security Policy: a browser header that limits which scripts and sources a web page may load, a standard defence against injected code.

CSPMCloud Security Posture Management · CSPMs

Cloud security posture management: tooling that pulls cloud configuration through provider APIs, compares it with known-good states and compliance frameworks such as CIS and PCI, and flags misconfigurations; some definitions add scanning of infrastructure-as-code before deployment.

CSRcertificate signing request · certificate signing requests · CSRs

A certificate signing request: the message a device or service sends to a certificate authority asking it to sign a certificate for its public key.

CSRFcross-site request forgery

Cross-site request forgery: a page tricks a signed-in user's browser into sending a request to another site, which acts on it because the browser attaches the user's cookies.

Cursor-based Pagination of SCIM Resourcesdraft-ietf-scim-cursor-pagination · RFC 9865

A Proposed Standard from the Internet Engineering Task Force (IETF), RFC 9865, that adds cursor query parameters and nextCursor/previousCursor response attributes to the System for Cross-domain Identity Management (SCIM) protocol.

CVECVEs · Common Vulnerabilities and Exposures · CVE-2026 · CVE-2025 · CVE

A public catalogue number for a known, specific software or hardware vulnerability, used to track and reference security flaws consistently across vendors and researchers.

CVSSCommon Vulnerability Scoring System · CVSS score

The Common Vulnerability Scoring System, a 0 to 10 severity score for a vulnerability; 7.0 and above is high, 9.0 and above critical.

CWTCBOR Web Token · CBOR Web Tokens

CBOR Web Token: a compact signed token format like JWT, but encoded in binary so it suits small devices and tight bandwidth.

Cybench

Cybersecurity agent benchmark built from professional-level capture-the-flag (CTF) tasks.

DASTdynamic application security testing

Dynamic application security testing: probing a running application from the outside for flaws such as injection. It complements SAST, which reads the source code.

Data exfiltration

Unauthorised transfer of data out of a system.

Deanonymizationdeanonymization · de-anonymization · re-identification

Linking a pseudonymous or anonymized record or account to a real identity or to another account.

Decentralized identitydecentralized ID · self-sovereign identity · SSI

Identity built from credentials the holder keeps in a wallet and presents to a verifier, who checks the signed claims without calling the issuer. Decentralized Identifiers name the holder.

Decentralized Identity FoundationDIF

A Linux Foundation-hosted standards group for decentralized identity; its Trusted AI Agents working group publishes KYA-OS, an identity and delegation spec for AI agents.

Decoy digestsdecoy digest

Extra digests in a Selective Disclosure JSON Web Token (SD-JWT) that match no Disclosure, added to hide how many claims exist.

Deepfakedeepfakes

Audio, video or images generated by AI to look like a real person. In identity work they are used to fool video checks, call-center staff and verification tools.

Denial of serviceDoS · DDoS

An attack that makes a service unavailable, usually by flooding it with requests or exhausting a resource.

Device-bound passkeydevice-bound passkeys · device bound pass keys · device-bound passkey · device bound passkeys

A passkey whose private key is created on one piece of hardware, such as a security key, and cannot leave it, unlike a synced passkey that is copied between devices.

DID / Verifiable CredentialsDID · VC · verifiable credentials · Decentralized Identifiers

W3C standards for user-held digital credentials: Decentralized Identifiers (DIDs) name the holder, Verifiable Credentials carry cryptographically signed claims that can be checked without calling the issuer.

did:keydid key method

A Decentralized Identifier method in which the identifier is the public key itself, so it needs no registry or server to resolve and cannot be rotated or revoked; changing the key means a new identifier.

did:webdid web method

A Decentralized Identifier method that resolves an identifier like did:web:example.com to a document served over HTTPS from that domain, so it inherits the web's domain ownership and certificates.

did:webvhdid:tdw

A Decentralized Identifier method that anchors an identifier to a web domain, as did:web does, and adds a signed log of every change so a verifier can check the identifier's history.

DIF Presentation ExchangePresentation Exchange · PEX · presentation_definition

A DIF format for a verifier to say which credentials it wants and for a wallet to say which ones it is sending; a forerunner of OpenID4VP's query language.

Differential privacy

Formal privacy guarantee bounding what an output reveals about any one input, set by epsilon and delta

Digital Credentials APIW3C Digital Credentials API · DC API

A W3C browser interface that lets a website ask for a credential, such as a digital driver's license, from the wallet on the user's device.

DISPATCH (IETF)SECDISPATCH

Internet Engineering Task Force (IETF) working groups that take proposals for new work and recommend where each goes: an existing group, a new one, or no further action. SECDISPATCH does this for security topics.

DLPdata loss prevention

Data loss prevention: tools that spot sensitive data leaving, such as in email or uploads, and block or flag it.

DNS rebindingMCP rebinding · DNS rebinding attack

An attack in which an attacker-run Domain Name System (DNS) server re-points a website's domain at a local address, so scripts in the victim's browser can reach services meant to be local only.

DNS-ID

A Domain Name System (DNS) identifier: a domain name listed in the subject alternative name field of a TLS (Transport Layer Security) certificate, which a client compares with the name it meant to reach.

Dolev-Yao modelDolev-Yao · Dolev–Yao · Dolev-Yao adversary

Classic formal adversary model in which the attacker controls the network (intercept, replay, synthesize messages) but cannot break the cryptography

DORADigital Operational Resilience Act

An EU law, in force since January 2025, that makes banks, insurers and their IT suppliers prove they can withstand and recover from cyber incidents.

Downgrade attack

An attack that pushes a user from a strong sign-in method to a weaker fallback, such as an SMS code, that the attacker can beat.

DPoPRFC 9449 · Demonstrating Proof of Possession · sender-constrained tokens

A way to bind an OAuth access token to the specific key that requested it: the client signs a fresh proof for every call, so a stolen bearer token is useless to whoever steals it without also holding the matching private key.

DSSEDead Simple Signing Envelope

Dead Simple Signing Envelope: a signing wrapper that binds a payload's type and bytes to the signature.

Dynamic Client RegistrationDCR · RFC 7591 · RFC7591

An OAuth mechanism (RFC 7591) letting a client register itself with an authorization server at runtime and receive a client ID, instead of being provisioned by hand.

ECDHElliptic Curve Diffie-Hellman

A key-agreement method in which parties use elliptic-curve public keys and their private keys to derive a shared secret. It does not authenticate the parties by itself.

ECDH-ESECDH-ES+A256KW

Elliptic-curve Diffie-Hellman Ephemeral-Static, the JOSE key agreement that lets a sender encrypt a JWT to a recipient's public key. Implementation bugs, such as accepting points on the wrong curve, have made it a recurring source of JWT encryption flaws.

ECDSAElliptic Curve Digital Signature Algorithm

The elliptic-curve signature algorithm behind most JWT and TLS certificate signatures today; like RSA, it is not safe against a large quantum computer.

Ed25519EdDSA Ed25519

A digital-signature scheme on an elliptic curve, used to sign messages and tokens in Secure Shell (SSH), Transport Layer Security (TLS) and identity protocols.

EdDSA

Edwards-curve Digital Signature Algorithm: an elliptic-curve signature scheme, usually used as Ed25519, with short keys and deterministic signatures.

EDRendpoint detection and response

Software on laptops and servers that watches for malicious behavior and can stop it on its own.

Egress controlegress controls

Limiting which outbound destinations a process may reach, to stop exfiltration.

eIDAS RegulationeIDAS

The EU regulation on electronic identification and trust services; it sets low, substantial and high assurance levels for digital identity, and eIDAS 2.0 added the EU Digital Identity Wallet.

EKUExtended Key Usage

A field in an X.509 certificate that limits what the certificate may be used for, such as client or server authentication.

Endorsement KeyTPM Endorsement Key

Endorsement Key (EK): a TPM key derived from the chip's endorsement seed, whose manufacturer-issued certificate vouches that the TPM is genuine hardware.

Entity Attestation Token (RFC 9711)RFC 9711 · RFC9711 · Entity Attestation Token

Token format from the Remote ATtestation procedureS (RATS) work of the Internet Engineering Task Force (IETF), encoded in Concise Binary Object Representation (CBOR) or JSON, carrying claims about an attested device or workload.

Ephemeral credentialsephemeral tokens · ephemeral access · short-lived credentials

Credentials issued for one task and set to expire in minutes or hours, so a stolen copy is soon useless.

ETSIEuropean Telecommunications Standards Institute

European standards body that publishes telecom, digital-signature and security specifications.

EU Digital Identity WalletEUDIW · eIDAS 2.0 · eIDAS2 · EUDI Wallet · EUDI

The digital identity wallet each EU member state must offer its citizens by 24 December 2026 under the revised eIDAS (electronic identification and trust services) regulation, holding identity data and credentials in a common format.

Execution isolation

Running untrusted components in separate constrained processes or contexts so they interact only through mediated interfaces.

Executive Order 14412EO 14412 · Securing the Nation Against Advanced Cryptographic Attacks

A US executive order of 22 June 2026 setting federal deadlines for post-quantum cryptography: key establishment by the end of 2030, digital signatures by the end of 2031.

Fail closedfail-closed · fails closed

Deny by default when a check cannot be completed, rather than let the request through.

Fail openfails open · failing open · fail-open

Let the request through when a check cannot be completed, the opposite of failing closed.

FAPIFAPI 2.0 · Financial-grade API · FAPI 2.0 Security Profile

OpenID's Financial-grade API Security Profile: a high-assurance OAuth/OpenID Connect profile originally built for banking APIs, now referenced wherever a deployment needs OAuth hardened beyond the baseline.

FATFFinancial Action Task Force

Financial Action Task Force, the global body that sets anti-money-laundering standards. Its travel rule makes payment firms pass sender and receiver details along, which matters for agent payments.

Federation assurance levelFAL · FAL1 · FAL2 · FAL3

NIST's levels for how strongly a federated login assertion is protected, from bearer assertions up to ones bound to the user's key.

FedRAMPFederal Risk and Authorization Management Program

The US program that authorizes cloud services for federal agency use.

FIDO AllianceFIDO

The industry consortium behind FIDO2 and passkeys — the push to replace passwords with phishing-resistant credentials.

FIDO2 / passkeyspasskeys · passkey · WebAuthn · FIDO2

Standards from the Fast Identity Online (FIDO) Alliance and World Wide Web Consortium (W3C) behind passkeys: phishing-resistant sign-in with public-key credentials instead of passwords, on one device or synced across devices.

FIPSFederal Information Processing Standards

Federal Information Processing Standards: the US government's technical standards for computer systems, including the approved cryptography rules.

FIPS 203 (ML-KEM)ML-KEM · FIPS 203

NIST's post-quantum standard for agreeing an encryption key over an open network.

FIPS 204 (ML-DSA)FIPS 204

NIST's main post-quantum digital signature standard, the replacement for RSA and ECDSA signatures.

FIPS 205 (SLH-DSA)SLH-DSA · FIPS 205

NIST's hash-based post-quantum signature standard, slower than ML-DSA but resting on fewer assumptions.

Formal verificationformal verification · formal security analysis

Proving mathematically that a protocol or system meets a stated security property inside an explicit model, rather than testing it; the result holds only within the model's assumptions.

FTCFederal Trade Commission

The US consumer-protection regulator that tracks reported fraud losses (its Consumer Sentinel Network) and enforces rules against deceptive and impersonation-based scams.

GDPRGeneral Data Protection Regulation

The EU's General Data Protection Regulation, which obliges organisations processing personal data (including data touched by software agents) to support pseudonymization, consent, and demonstrable access-control accountability.

GLEIFGlobal Legal Entity Identifier Foundation

The Swiss foundation that runs the global Legal Entity Identifier (LEI) system and the verifiable LEI (vLEI), overseen by a committee of financial regulators.

Global Hardware Trust StoreTrust Store

Curated list of hardware attestation roots accepted for agent identity enrollment.

Global Token Revocationdraft-parecki-oauth-global-token-revocation

An IETF draft for telling every service a user's tokens and sessions are revoked at once, not one application at a time.

GNAPGrant Negotiation and Authorization Protocol

An IETF authorization protocol designed to replace OAuth 2.0's grant model with a single, extensible negotiation between client and authorization server, folding in capabilities (like PAR and device flow) that OAuth bolted on as separate extensions.

Hash chainhash chaining · hash-chained · chain hash

A log where each entry carries the hash of the one before it, so changing any earlier entry breaks every later link.

High value assetHVA · HVAs · high value assets

A US federal designation for the systems and data whose compromise would do the most harm; they migrate first.

HIPAAHealth Insurance Portability and Accountability Act · HIPAA's Privacy Rule

The US Health Insurance Portability and Accountability Act, which sets privacy and security rules for protected health information.

HMACHMAC

Hash-based message authentication code (HMAC): a keyed hash that proves a message came from someone holding a shared secret; both sides use the same key, so it cannot prove which of them signed.

HMAC-SHA-256

Hash-based message authentication code (HMAC) built on the SHA-256 (Secure Hash Algorithm, 256-bit) hash: it combines a shared secret key with the hash to check message integrity and possession of that key.

Holder bindingsubject binding

Proof that the party presenting a credential is the one it was issued to, not someone who copied it.

HS256

The JWT signing algorithm that uses HMAC with SHA-256 and one shared secret, so anyone who can check a token can also mint one.

HSMhardware security module · HSMs

A dedicated hardware device that generates and stores cryptographic keys and performs signing/encryption operations inside a tamper-resistant boundary, so private keys never leave the device.

HTTP Message SignaturesRFC 9421 · HTTP Signatures · RFC 9421 HTTP Message Signatures

An IETF standard (RFC 9421) for cryptographically signing parts of an HTTP request or response, so a receiver can verify who sent it and that it was not altered in transit — an application-layer alternative to relying on TLS alone.

HTTP Signature Keysdraft-hardt-httpbis-signature-key · Signature-Key header

An IETF draft for carrying the key that signed an HTTP request alongside the request itself, which AAuth builds on.

Hybrid Public Key Encryption (HPKE, RFC 9180)HPKE · RFC 9180

IETF standard for hybrid public-key encryption to a named recipient.

IAMidentity and access management · identity & access management

The discipline and tooling for deciding who, or what, can sign in and what they may do afterwards: accounts, authentication, authorization, and the lifecycle around them.

IANAInternet Assigned Numbers Authority

The Internet Assigned Numbers Authority: the body that keeps the official registries of protocol names and numbers, so that two standards do not use the same label for different things.

ID TokenID Tokens · id_token

A signed JSON Web Token (JWT) that an OpenID Connect provider issues to tell an app who signed in. Required claims include the issuer, subject, audience, expiry (exp) and issued-at time.

Identity Assertion Authorization GrantIdentity Assertion Authorization Grant · ID-JAG draft · draft-ietf-oauth-identity-assertion-authz-grant · IDJAG · Identity Assertion JWT Authorization Grant

An OAuth draft from the Internet Engineering Task Force (IETF) letting an app trade a corporate identity provider's assertion for an access token to a third-party API, keeping cross-app access under enterprise control.

Identity assurance level (IAL)IAL · IAL1 · IAL2 · IAL3 · Identity Assurance Level · Identity Assurance Levels

NIST SP 800-63 scale (IAL1 to IAL3) for how strongly an applicant's real-world identity was proofed before a credential is issued; sits beside the registered federation assurance level (FAL).

Identity fabricidentity fabrics

KuppingerCole's term for a set of connected identity services that work across all of an organization's applications.

Identity launderingidentity laundering

Passing an untrusted caller's request through a trusted internal service so that downstream systems treat it as more trustworthy than its origin warrants.

Identity resilienceidentity resilience

The ability to keep or restore identity infrastructure (IdPs, directories, app connections) after an incident or drift.

IDORinsecure direct object reference · insecure direct object references

Insecure direct object reference: a bug where an app trusts the record id in a request and skips checking that the caller may see that record, so changing the id exposes someone else's data.

IdPidentity provider · identity providers

The system that authenticates users and issues the tokens other applications trust. Okta, Microsoft Entra, and PingOne are identity providers.

IDVidentity verification · identity proofing

Checking that a person is who they claim to be, usually with a government document and a live selfie, before an account is created or recovered.

IESGInternet Engineering Steering Group · RFC Ed Queue · RFC Editor queue

Internet Engineering Steering Group: the IETF body that reviews and approves a working group's draft for publication. Once it approves, the draft waits in the RFC Editor's queue for its number; that queue is the last step before a standard is published.

IETFInternet Engineering Task Force

The body that writes the internet's protocol standards, including OAuth, JWT, and the workload-identity work known as WIMSE.

IETF working groupWG · OAuth WG · WIMSE WG

A group inside the Internet Engineering Task Force (IETF) that works on one topic, such as OAuth or workload identity, and reaches consensus on its drafts before the Internet Engineering Steering Group (IESG) approves them.

IGAIdentity Governance and Administration · IAG · Identity and Access Governance

Identity Governance and Administration — deciding who should have access to what and proving it: joiner/mover/leaver provisioning, access certification, least-privilege cleanup.

Impersonationimpersonation semantics

One party acting fully as another, indistinguishable to any system it reaches, rather than acting as itself with borrowed authority (delegation). RFC 8693's OAuth token exchange names both as the two semantics a security token can carry.

Implicit grantimplicit grant · OAuth implicit flow

The original OAuth 2.0 flow that returns an access token straight in the browser redirect instead of exchanging a code. RFC 9700 says clients SHOULD NOT use it, because tokens leak through URLs and cannot be sender-constrained.

in-totoin-toto Attestation Framework · in-toto Statement

A framework for signed supply-chain attestations: a statement names its subjects by digest and carries a typed body.

Incident response (IR)incident response

Incident response: the team and process that contain, investigate and recover from a security breach.

Information flow controlIFC · information-flow control · data flow tracking

Tracking where each piece of data came from and who may see it, then blocking any step that would send it somewhere it should not go.

Instance Metadata ServiceIMDS · instance metadata endpoint

A local HTTP endpoint on a cloud machine that tells a workload about itself and hands it short-lived credentials; attackers reach it through server-side request forgery.

Institute of Internal AuditorsIIA · The IIA

The global body that sets the standards internal audit departments follow; when it names a control, corporate audit committees start asking about it.

International Telecommunication UnionITU

The United Nations agency for information and communication technologies, which publishes telecom and security standards such as X.509.

Internet Security AllianceISA

Trade association of CISOs across critical-infrastructure sectors; with NACD it writes and chief-authors the Director's Handbook on Cyber-Risk Oversight (Larry Clinton, President and CEO).

IOCIOCs · indicator of compromise · indicators of compromise

Indicator of compromise: a piece of evidence that a system has been breached, such as a malicious file hash, IP address, domain or log pattern. Defenders share IOCs so others can detect the same attack.

IPSIEInteroperability Profiling for Secure Identity in the Enterprise

An OpenID Foundation working group defining enterprise deployment profiles — how identity providers and workforce apps must interoperate for SSO, provisioning, and shared-signal features — so enterprises get a consistent security baseline instead of ad hoc integrations.

ISACAInformation Systems Audit and Control Association

The professional association for IT auditors and governance specialists; it publishes COBIT and the audit programmes that decide what evidence an access control has to produce.

ISO/IEC 18013-5 (mobile driving licence)ISO/IEC 18013-5 · ISO 18013-5 · mdoc · ISO mdoc

ISO standard defining the mobile driving licence (mDL) data model and in-person presentation, the basis of mdoc credentials.

ISO/IEC 29115ISO 29115

An international standard that defines levels of assurance for proving who someone is when they sign up and sign in.

ISPM / ITDRISPM · ITDR

Identity Security Posture Management (ISPM) continuously audits identity configuration for weaknesses; Identity Threat Detection and Response (ITDR) watches live identity activity for attacks in progress.

ITPidentity threat protection

Identity threat protection: tools that watch sign-in and account signals to spot attacks on identities as they happen. It is often paired with ISPM.

ITU Telecommunication Standardization SectorITU-T

The International Telecommunication Union sector that develops international telecommunications and ICT standards called ITU-T Recommendations.

JA3/JA4 TLS fingerprintingJA3 · JA4

Hashes of Transport Layer Security (TLS) ClientHello parameters that identify a client's TLS stack, so the same client can be recognized across connections.

JITjust-in-time access · JEA · Just-Enough-Administration

Granting a permission only for the moment it is needed and withdrawing it afterwards, so no account sits with standing power an attacker could inherit.

JMLjoiner-mover-leaver · joiner mover leaver

The joiner-mover-leaver lifecycle: provisioning access when someone joins, adjusting it as their role changes, and revoking it when they leave.

JOSEJSON Object Signing and Encryption

The IETF family of JSON formats for signed and encrypted data (JWS, JWE, JWK, JWA) that JWTs are built on.

JSON Web EncryptionJWE · RFC 7516

The JOSE (JSON Object Signing and Encryption) format for encrypting a token's payload so only the key holder can read it; an encrypted access token cannot be validated without that key.

JSON Web Key (JWK)JWK · RFC 7517

The IETF format for writing a cryptographic key as JSON, so a token's signing key can be published and looked up.

JSON Web SignatureJWS · RFC 7515

The IETF format for signing a JSON payload so a receiver can check who produced it and that it was not changed.

JSON Web Token Best Current PracticesRFC 8725 · RFC8725 · BCP 225

IETF best practices for JWT validation: algorithm checks, key handling, explicit typing

JWKSJSON Web Key Set

JSON Web Key Set: a published list of public keys that lets anyone check signatures made by the matching private keys.

JWT Bearer GrantRFC 7523 · JWT Profile for OAuth 2.0 Client Authentication and Authorization Grants · JWT-bearer assertion · RFC7523

The OAuth mechanism (RFC 7523) for presenting a signed JSON Web Token (JWT) as an authorization grant or client credential, redeemed at an authorization server that trusts the token's issuer.

JWT explicit typingexplicit typing · typ header

Using the typ header parameter (e.g. secevent+jwt) in a JSON Web Token (JWT) so each kind of token can be told apart and wrong-kind tokens rejected.

JWT Profile for OAuth 2.0 Access TokensRFC 9068 · RFC9068 · JWT access token profile · at+jwt

RFC 9068: a profile for OAuth access tokens issued as a JSON Web Token (JWT), with a fixed typ of at+jwt and required claims such as issuer, audience, subject and client id.

JWT-SVIDJWT SVID

SPIFFE Verifiable Identity Document (SVID) issued as a signed JSON Web Token, used when a workload must prove who it is through a proxy or where mutual TLS is not available.

Kerberos

The ticket-based network authentication protocol (MIT, 1980s) that let a client prove its identity to a service without repeatedly sending a password; the direct ancestor of today's machine-identity and service-account credentialing schemes.

KEVKnown Exploited Vulnerabilities · CISA KEV · Known Exploited Vulnerabilities catalog · KEV catalog

The US Cybersecurity and Infrastructure Security Agency's (CISA) public catalogue of vulnerabilities with confirmed exploitation in the wild; each entry carries a deadline by which US federal agencies must remediate.

Key Binding JWTKB-JWT · key binding JWT · KB-SD-JWT

The signed token that a holder of an SD-JWT (Selective Disclosure JSON Web Token) attaches to a presentation to prove it holds the key named in the credential's cnf claim.

KMSkey management service · key management system

Key management service: a system that creates, stores and controls cryptographic keys, usually in the cloud. Customer-managed keys let a company keep control over who can decrypt its data.

KYBKnow Your Business

Know Your Business: checking that a business customer is real and finding who owns and controls it, the business version of Know Your Customer (KYC) checks.

KYCKnow Your Customer · know-your-customer

Know Your Customer: the identity checks a bank or regulated business must run on a customer before serving them; KYB is the same for a business customer.

LAMPSIETF LAMPS

The IETF working group on Limited Additional Mechanisms for PKIX and SMIME. It maintains certificate and signing formats, including post-quantum ones.

LDAPLightweight Directory Access Protocol

A protocol for looking up and managing entries in a directory, spoken by on-premises directories such as Microsoft Active Directory.

Least privilegeleast-privilege

Giving every account, human or machine, only the access its job needs and nothing more, so a compromise stays small.

Legal Entity Identifier (LEI)LEI · ISO 17442

A 20-character code identifying a legal entity in financial transactions, issued under a global system that GLEIF oversees.

Macaroonsmacaroon

Bearer credentials that can be attenuated by chained HMAC caveats without contacting the issuer.

MACed

Protected with a message authentication code (such as HMAC) instead of a digital signature: a shared secret key lets the receiver check that the data is unchanged and came from someone holding that key.

Machine-to-machine applicationM2M · M2M app · machine-to-machine app

An application registered with an identity provider that signs in as itself, with a client ID and secret, rather than on behalf of a person.

Man-in-the-middle attackman-in-the-middle · MITM

An attacker sits between two parties, reading and possibly changing their traffic.

Managed identitymanaged identities

An Azure identity for a workload whose credentials the platform issues and rotates, so no secret is stored in code or config.

mDL / mobile IDsmobile driver's license · mobile IDs · mDL

Mobile driver's license — the ISO/IEC 18013-5 standard for government-issued IDs held in phone wallets and presented digitally.

MFAmulti-factor authentication · two-factor authentication · 2FA

Proving who you are with more than a password, such as a one-time code, a hardware key, or a fingerprint.

Micro-segmentationmicrosegmentation · micro-segmentation

Dividing a network into small zones, each guarded by its own gateway or host firewall, so a compromise in one zone cannot spread freely to others.

MITRE

A US federally funded, not-for-profit R&D organization that operates government-sponsored research centers; maintains CVE and ATT&CK and contributes engineering staff to IETF and NIST standards work.

MITRE ATT&CKATT&CK

MITRE's catalog (Adversarial Tactics, Techniques, and Common Knowledge) of real-world attacker tactics and techniques, used to label what malware does and to write detections.

Mix-up attackmix-up attack · mix-up attacks

An attack on an OAuth client that talks to several authorization servers, one of them hostile, tricking it into sending a code or token to the wrong one. The defence is to check the issuer on every response.

MTTRmean time to recover · mean time to respond · mean time to remediate

Mean time to recover (or respond): the average time from detecting an incident to restoring service or containing it, one of the few security numbers boards are asked to track.

Multi-tenant appmulti-tenant app · multitenant app · multi-tenant application

An Entra app registration that other organisations' tenants can consent to, creating a service principal in each; the vendor's credentials on the one registration then work in every consenting tenant.

National security systemsNSS · national security system

US systems used for intelligence, military or cryptographic purposes, governed separately from civilian federal IT.

NCCoENational Cybersecurity Center of Excellence

NIST's lab that builds reference implementations of security practices with industry partners and publishes them as practice guides.

NHInon-human identity · non-human identities · NHIs · machine identity · machine identities · NPE · non-person entity · non-person entities

A non-human identity: any account, key, certificate or token that software (service accounts, bots, workloads, AI agents) uses to authenticate, as opposed to a person's login.

NIS2NIS2 Directive · Network and Information Systems Directive 2 · EU NIS2 Directive

The EU cyber-risk directive that imposes direct board-level accountability for management bodies, with penalties reaching €10 million or 2% of global turnover for non-compliance.

NISTNational Institute of Standards and Technology

The US National Institute of Standards and Technology, whose publications (the Cybersecurity Framework, SP 800-63 digital identity guidelines, the AI Risk Management Framework) set the baseline many enterprises and regulators measure identity and AI controls against.

NIST IR 8547NIST IR 8547 · Transition to Post-Quantum Cryptography Standards

NIST's draft transition plan that deprecates today's public-key algorithms after 2030 and disallows them after 2035.

NIST SP 800-207Zero Trust Architecture · ZTA

The National Institute of Standards and Technology (NIST) Zero Trust Architecture publication: no implicit trust from network location or credential possession, and continuous verification of each request.

NIST SP 800-53SP 800-53 · NIST SP 800-53 Rev. 5 · 800-53r5

NIST's catalog of security and privacy controls that US federal systems are assessed against.

NIST SP 800-63NIST 800-63 · Digital Identity Guidelines

NIST's Digital Identity Guidelines: originally federal e-authentication rules that became the de facto benchmark for identity-proofing and authentication assurance levels well beyond government procurement.

NISTIR 8587NIST IR 8587

NIST guidance on protecting tokens and assertions from forgery, theft and misuse.

Non-repudiationnon repudiation

Evidence that a party performed an act strong enough that it cannot credibly deny it later, as a signature gives.

NSANational Security Agency · NSA-CCSS · NSA Central Security Service

The US National Security Agency, whose Central Security Service (CCSS) mission includes contributing to open cryptographic and identity standards used across US government and critical-infrastructure systems.

OAuthOAuth 2.0 · OAuth 2.1 · RFC 6749

The web's standard framework for delegated authorization — letting one application act at another on a user's (or machine's) behalf without sharing passwords.

OAuth 2.0 Attestation-Based Client Authentication

IETF draft letting a native app present a key-bound attestation to authenticate to an authorization or resource server.

OAuth 2.0 Authorization Server Issuer IdentificationIssuer Identification · iss parameter · RFC 9207 · RFC9207

An OAuth mechanism (RFC 9207) in which the authorization server adds its identity (iss) to each authorization response, so the client can check it came from the server it started with, to counter mix-up attacks.

OAuth 2.0 Authorization Server MetadataRFC 8414 · RFC8414 · Authorization Server Metadata

An IETF standard (RFC 8414) defining a well-known JSON document an OAuth client fetches to discover an authorization server's endpoints and capabilities, generalizing OpenID Connect Discovery to OAuth 2.0.

OAuth 2.0 Device Authorization GrantRFC 8628 · RFC8628 · device authorization grant

RFC 8628: lets an input-limited device, such as a TV, get an OAuth token while the user approves the request on a second device.

OAuth 2.0 for First-Party Applications (FiPA)FiPA · OAuth First-Party Apps

IETF OAuth WG draft defining an authorization challenge endpoint and auth_session for native first-party clients.

OAuth 2.0 for Native Apps (RFC 8252)RFC 8252 · RFC8252

A Best Current Practice from the Internet Engineering Task Force (RFC 8252) for OAuth 2.0 clients that run as native apps, using loopback redirects and private-use URI schemes.

OAuth 2.0 Protected Resource MetadataProtected Resource Metadata · PRM · RFC 9728 · RFC9728

A discovery document (RFC 9728) a resource server publishes at a well-known URL, naming the authorization server that protects it and the scopes it supports.

OAuth 2.0 Security Best Current PracticeRFC 9700 · RFC9700 · BCP 240

RFC 9700 (January 2025), the IETF's consolidated security guidance for OAuth 2.0: exact-match redirect URIs, no implicit grant, and sender-constrained or short-lived tokens. It is a best current practice, so deployers are expected to follow it.

OAuth 2.0 Step Up Authentication ChallengeRFC 9470 · RFC9470 · step-up authentication challenge

RFC 9470: lets an API tell a client that the user must log in again, more strongly or more recently, before a call succeeds.

OAuth 2.0 Threat Model and Security Considerations (RFC 6819)RFC 6819 · RFC6819

The 2013 IETF document that first catalogued OAuth 2.0 threats. RFC 9700 extends and updates it and makes its attacker model explicit.

OAuth 2.0 Token RevocationRFC 7009 · RFC7009 · token revocation endpoint

RFC 7009: lets a client tell the authorization server to cancel one access or refresh token.

OAuth Attestation-Based Client Authenticationdraft-ietf-oauth-attestation-based-client-auth · Client Attestation PoP JWT

An IETF OAuth draft where a client proves its identity to the token endpoint with an attestation issued by a trusted party plus a fresh proof-of-possession token signed by the client's own key.

OAuth Deferred Token ResponseDTR · draft-ietf-oauth-deferred-token-response · draft-gerber-oauth-deferred-token-response · Deferred Token Response

An OAuth working-group draft that lets an authorization server answer a token request with 'not yet': the client gets a deferral code and polls until a person or process approves.

OAuth Identity and Authorization Chaining Across DomainsIdentity Chaining · draft-ietf-oauth-identity-chaining · OAuth Identity Chaining · identity and authorization chaining

An IETF OAuth working group draft (draft-ietf-oauth-identity-chaining) that defines how to preserve a user's identity and granted authorization as a request crosses OAuth trust-domain boundaries, by chaining OAuth 2.0 Token Exchange (RFC 8693) and the JWT bearer assertion grant (RFC 7523) at every hop.

OAuth SPIFFE Client Authenticationdraft-ietf-oauth-spiffe-client-auth · SPIFFE client authentication

An IETF OAuth working-group draft that lets a workload log in to an OAuth authorization server with its SPIFFE identity (a JWT, WIT or X.509 certificate) instead of a client secret.

OAuth Token Exchangetoken exchange · RFC 8693 · RFC 8693 OAuth Token Exchange · RFC 8693 OAuth 2.0 Token Exchange

The OAuth mechanism (RFC 8693) for trading one token for another with narrower scope or a different audience, so each hop in a chain gets a credential for just its next step.

OAuth Token IntrospectionRFC 7662 · RFC7662 · token introspection

An OAuth extension (RFC 7662) letting a resource server ask the authorization server, at call time, whether a token is still valid and what it is bound to — instead of trusting the token's own claims at face value.

OAuth Transaction Authorization Challengedraft-rosomakho-oauth-txn-challenge

An Internet-Draft (Rosomakho, Campbell, McGuinness, Kasselman, June 2026) in which a protected resource returns a signed challenge carrying authorization_details that the authorization server validates and derives granted details from.

Oblivious HTTP (OHTTP)OHTTP · Oblivious HTTP · RFC 9458

IETF standard (RFC 9458) that sends HTTP requests through a relay so the server cannot see the client IP, with HPKE hiding content from the relay.

ODRL Information Model 2.2ODRL

Open Digital Rights Language (ODRL): a World Wide Web Consortium (W3C) Recommendation defining a policy model and vocabulary for permissions, prohibitions and constraints.

Offline verificationoffline-verifiable · offline verification

Checking a credential's validity and scope at the point of use without calling the issuer or an authorization server.

OMB M-26-15M-26-15 · Execution of the Migration to Post-Quantum Cryptography

The June 2026 White House budget office memo telling federal agencies how and by when to move to post-quantum cryptography.

Online Certificate Status ProtocolOCSP

A protocol for asking a certificate authority, one certificate at a time, whether a certificate is still valid; the authority learns which certificate is being checked.

Open bankingopen banking standards

Regulation-driven standards (such as the Financial-grade API (FAPI) profiles of OAuth) that let a customer grant a third party scoped, time-limited access to bank data and payments.

Open Policy AgentOPA · Rego

An open-source, general-purpose policy engine for evaluating structured authorization decisions against a declarative policy language.

OpenFGAFGA · Open FGA

An open-source fine-grained authorization (FGA) engine, modelled on Google's Zanzibar, that stores relationship tuples and answers yes/no access checks.

OpenID ConnectOIDC

OpenID Connect — an identity layer on top of OAuth 2.0 that lets an app verify who the user is through an identity provider; the protocol behind most “Sign in with…” buttons.

OpenID Connect Discovery 1.0OpenID Connect Discovery · OpenID.Discovery

OpenID Connect Discovery — how a client finds an OpenID Provider's endpoints and keys from a published metadata document.

OpenID FederationOpenID Federation 1.0

An OpenID specification for building trust between organizations through signed statements about who vouches for whom, so parties with no shared identity provider can still verify each other.

OpenID for Verifiable Credential IssuanceOpenID4VCI · OID4VCI

The OpenID Foundation protocol for an issuer to put a digital credential into a person's wallet; the issuance half of the pair whose presentation half is OpenID4VP.

OpenID for Verifiable PresentationsOpenID4VP · OID4VP · OIDC4VP

The OpenID Foundation protocol for presenting a digital credential from a wallet to a verifier, with a transaction_data field that lets the holder sign extra information, such as a payment mandate, along with the presentation.

OpenID FoundationOIDF

The standards body behind OpenID Connect, CIBA, and AuthZEN — where much of the web's login plumbing gets written.

OpenID Provider (OP)OpenID Provider · OP

The server in OpenID Connect that signs a user in and issues an ID token saying who they are. Relying parties, the apps that accept the login, trust it for that answer.

OpenID4VCOpenID4VC · OpenID to Verifiable Credentials

OpenID Foundation family bridging OAuth/OpenID to verifiable credentials (issuance, presentation).

OToperational technology

Operational technology: the industrial control and sensor systems (SCADA, PLCs) that run physical plants and grids, historically kept isolated from IT networks and now increasingly connected to them.

OVorganization validation

Organization validation, a certificate level where the authority checks that a real organization owns the name, above domain validation and below extended validation (EV).

Over-provisioningoverprovisioning · over-provisioned access

Handing out more access than a person or workload needs. Over-provisioned accounts widen the damage when they are stolen.

OWASPOWASP Foundation · Open Web Application Security Project · OWASP GenAI Security Project

Open Web Application Security Project: the nonprofit that maintains the OWASP Top 10 list of web-application risks and the OWASP Top 10 for Large Language Model Applications.

OWASP Application Security Verification Standard (ASVS)ASVS · Application Security Verification Standard

An OWASP list of web-application security requirements in three verification levels that auditors and developers check software against.

OWASP Non-Human Identities Top 10 (2025)OWASP NHI Top 10 · NHI Top 10

OWASP's 2025 list of the ten main risks of non-human identities: improper offboarding, secret leakage, overprivilege, long-lived secrets and reuse.

P-256P-256 · secp256r1

An elliptic curve standardized by the National Institute of Standards and Technology (NIST), also called secp256r1, used for digital signatures and key agreement on the web.

Pairwise pseudonymous identifierPPID · pairwise identifier · pairwise subject identifier · pairwise pseudonymous identifier

Pairwise pseudonymous identifier (PPID): a user identifier that differs for every app or app group, so two apps cannot compare identifiers to tell they serve the same person.

PAMprivileged access management

Controls for the most powerful accounts: vaulting administrator credentials, granting elevated access only when needed, and recording what was done with it.

Passkey syncsynced passkeys

Copying passkeys between a user's devices through a cloud account, so a lost phone does not lock them out.

Path traversaldirectory traversal

Flaw where attacker-supplied file paths (for example with ../ sequences) make software read or write files outside the directory it meant to confine access to (Common Weakness Enumeration entry CWE-22).

Paved roadpaved roads

A security-team-built default path (templates, modules, patterns) that makes the secure way the easy way for engineers.

PCI DSSPCI-DSS · Payment Card Industry Data Security Standard · PCI · DSS

The card industry's security standard for any system that stores, processes or transmits card numbers; it is why a full card number or security code must never appear in a log.

Penetration testingpentest · pen test · pentesting · pen testing · penetration test

Authorized, scoped attempt to break into a system to find exploitable weaknesses before an attacker does, usually followed by a report of findings.

PEP / PDPPolicy Enforcement Point · Policy Decision Point · PEP · PDP

The split between the component that intercepts a request and enforces a decision (the policy enforcement point, such as an API gateway) and the service that makes the permit-or-deny decision (the policy decision point).

Phishing-resistant authenticationphishing-resistant MFA · phishing-resistant sign-in · phishing-resistant

Sign-in methods that cannot be talked out of the user, such as passkeys and hardware keys, because the secret is bound to the real site and never typed into a fake one.

PIIpersonally identifiable information

Personally identifiable information (PII): data that can identify a specific person, such as a name, address or ID number, and that laws often regulate differently by country.

PINpersonal identification number

A short numeric code a person types to unlock a card, phone or security key; it proves something the person knows.

PIVPersonal Identity Verification

The US government smart-card standard for holding a person's keys on a physical token; also used loosely for security keys like YubiKeys that implement it.

PKCEProof Key for Code Exchange · RFC 7636

An OAuth 2.0 security extension, mandatory in OAuth 2.1, that stops an intercepted authorization code from being redeemed: the client hashes a random code verifier into a code challenge sent up front, then must present the original verifier to exchange the code for a token.

PKIpublic key infrastructure

The certificates, certificate authorities and key-management processes that bind public keys to identities, underpinning TLS, code signing and machine/workload identity.

PKIX

Public-Key Infrastructure using X.509: the Internet Engineering Task Force's (IETF) family of standards for X.509 certificates on the internet, covering how certificates, revocation lists and certificate requests are formatted and checked.

Post-compromise securityPost-Compromise Security

Property that a connection recovers its confidentiality after an attacker has obtained its secrets, because later traffic uses fresh keys independent of the compromised ones.

Post-quantum cryptographyPQC · post-quantum

Cryptography designed to stay secure against quantum computers; ML-DSA (Module-Lattice-Based Digital Signature Algorithm) is a post-quantum signature algorithm standardized by the US National Institute of Standards and Technology (NIST).

Presentation attackpresentation attacks

Holding a fake, such as a photo, mask or recording, in front of a biometric sensor to pass as someone else.

Presentation attack detectionpresentation attack detection

Presentation attack detection (PAD): methods that detect fake biometric samples, such as masks, photos or screen replays, shown to a sensor during capture. Tested under ISO/IEC 30107.

Preventative controlspreventative control · preventive controls

Controls that block a disallowed action before it happens (for example cloud guardrail policies), as opposed to detecting it afterwards.

Private Key JWT client authenticationPrivate Key JWT · private_key_jwt

A client authentication method where the client proves who it is by signing a short-lived JWT with its private key instead of sending a shared secret. RFC 9700 recommends asymmetric methods like this one or mutual TLS.

Privilege escalationprivilege escalation

Gaining access or actions beyond what you were granted; for AI agents, an attacker's text makes the agent use tools the attacker could not call.

Proof of possessionPoP · proof-of-possession · PoP JWT

Proving you hold a private key by signing something fresh with it, so a token or attestation bound to that key cannot be replayed by whoever copies it.

Proof-of-Possession Key Semantics for JWTs (RFC 7800)RFC 7800 · RFC7800

Defines the cnf (confirmation) claim that lets a JWT carry or point to the key its holder must prove possession of.

ProVerifPROVERIF · ProVerif 2.00

Automatic protocol verifier that checks secrecy and authentication of a protocol model against a network attacker, symbolically.

Public clientpublic clients · public client

An OAuth client that cannot keep a secret, such as a single-page app or mobile app, so it cannot prove who it is with a client secret.

Pushed Authorization RequestsRFC 9126 · RFC9126

An OAuth extension (RFC 9126) letting a client submit its authorization request parameters directly to the authorization server over a back channel, with integrity protection, before redirecting the user — instead of exposing them in a browser URL.

RATSRemote ATtestation procedureS · RFC 9334 · RATS Architecture

Remote ATtestation procedureS: the IETF architecture (RFC 9334) for one party to produce evidence about its software and hardware state and another to verify it before trusting it.

RATS Conceptual Messages Wrapper (CMW)CMW · draft-ietf-rats-msg-wrap

A wrapper, defined by the Internet Engineering Task Force's Remote ATtestation procedureS (RATS) working group, that labels attestation Evidence or Results with a type so a verifier knows the format without prior agreement.

RATS Evidence

Remote ATtestation procedureS (RATS) claims produced by an Attester for appraisal by a Verifier; not an Attestation Result or authorization decision.

RATS passport modelPassport Model · passport topology

In Remote ATtestation procedureS (RATS), the topology where the Attester obtains signed Attestation Results from a Verifier and presents them to the Relying Party.

RBACrole-based access control

Access control in which each user holds one or more roles, and each role carries a fixed set of permissions that grant or deny actions.

ReBACrelationship-based access control · Relationship-Based Access Control

Access control that decides by the relationships between a subject and an object (owner, member, parent folder) rather than by fixed roles; Google's Zanzibar and OpenFGA are built on it.

Red teamingred team · red-team · red-teaming

Authorized adversarial testing in which a team plays the attacker to find weaknesses before real attackers do.

Refresh tokenrefresh tokens

An OAuth 2.0 credential issued with an access token that the client presents only to the authorization server to get a new access token of identical or narrower scope.

Registration authorityRA

A role in public key infrastructure (PKI) to which a certification authority delegates registration and certificate-management functions.

Relying partyRP

The website or service that accepts a sign-in or credential from someone else's identity system and relies on it to decide who the user is.

Remote code executionRCE · remote code execution

A flaw that lets an attacker run their own commands on someone else's machine over the network.

Remote identity proofing (RIDP)RIDP · remote ID proofing · remote identity proofing

Remote identity proofing (RIDP): verifying a person's real-world identity online by capturing their face and an identity document, checking liveness and authenticity, and matching the two, without meeting in person.

Resource exhaustionresource-exhaustion · CPU exhaustion · memory exhaustion · disk exhaustion

Using up a machine's processor time, memory or disk so that it slows or stops serving.

Resource Indicators for OAuth 2.0Resource Indicators · RFC 8707 · RFC8707

An OAuth extension (RFC 8707) letting a client name the resource server a token is for, via a resource parameter, so the token works at one audience only.

Resource owner password credentials grantresource owner password credentials grant · ROPC · password grant

The OAuth 2.0 grant in which the client collects the user's username and password and sends them to the authorization server. RFC 9700 says it MUST NOT be used.

Resource serverresource servers

In OAuth, the API that holds the protected data and accepts access tokens.

RFC 3161 (Time-Stamp Protocol)RFC 3161 · RFC3161 · Time-Stamp Protocol · TSA token

The Internet Engineering Task Force (IETF) protocol for asking a trusted time-stamping authority to sign that a given hash existed at a given time.

RFC 6750 (OAuth 2.0 Bearer Token Usage)RFC 6750

Defines how bearer tokens are sent in HTTP requests and how resource servers signal errors with WWW-Authenticate.

RFC 9942 COSE ReceiptsRFC 9942 · RFC9942

Receipt format in COSE (CBOR Object Signing and Encryption; CBOR is Concise Binary Object Representation), carrying inclusion proofs from a verifiable data structure.

RFC 9943 SCITT architectureRFC 9943 · RFC9943

Supply Chain Integrity, Transparency and Trust (SCITT) architecture for trustworthy and transparent digital supply chains, built on Issuers, Transparency Services, Signed Statements and Receipts.

RFC 9967 (SCIM Profile for Security Event Tokens)RFC 9967 · RFC9967 · draft-ietf-scim-events

IETF standard that carries SCIM (System for Cross-domain Identity Management) events as Security Event Tokens, so service providers and receivers exchange messages asynchronously. It also adds an optional asynchronous SCIM request mode.

Rich Authorization RequestsRAR · RFC 9396 · authorization_details

An OAuth extension that lets a client describe exactly what it wants authorized as structured data (e.g. which tool, which amount) instead of a single scope string, so an approver sees the specific action rather than a broad permission.

RISCRisk Incident Sharing and Coordination · Risk and Incident Sharing and Coordination Profile

An OpenID Shared Signals Framework profile for sharing account-risk events (account compromised, disabled, or recovered) between systems, so a relying party can act on a risk change it did not itself detect.

Root of Trustroots of trust

The smallest set of hardware, firmware and data a platform must trust without being able to check it, because nothing deeper can verify it; secure boot and attestation chains start from it.

RS256RS256

The JSON Web Token (JWT) signature algorithm that uses RSA (Rivest-Shamir-Adleman) keys with the SHA-256 (Secure Hash Algorithm, 256-bit) hash function.

RSA-PSS

The RSA signature scheme with probabilistic padding, preferred over older RSA v1.5 signatures. It appears in JOSE and COSE as a signing option.

Rug pull attackrug pull

A tool or MCP server is benign at install, then updated to carry malicious behavior or descriptions.

S256PKCE S256

The PKCE method in which an OAuth client sends a SHA-256 hash of a secret it keeps, so an intercepted authorization code cannot be redeemed by anyone else.

SAMLSAML 2.0

Security Assertion Markup Language (SAML): an XML-based (Extensible Markup Language) standard for single sign-on, in which an identity provider sends signed assertions about a user to an application.

Sandbox escapesandbox escape attack

Breaking out of an isolated runtime to run commands or reach files on the host.

Sandboxed executionsandboxed execution · sandboxed environment · sandboxing

Running agent-generated code inside an isolated, disposable environment so a malicious script cannot reach the host's files, network or processes.

SBOMsoftware bill of materials

A software bill of materials: a machine-readable list of every component and dependency inside a piece of software, so a team can tell quickly whether a vulnerable package is in it.

SCIMSystem for Cross-domain Identity Management

A standard application programming interface (API) for creating, updating and removing user accounts across applications, specified in RFC 7643 and RFC 7644.

SCITTSupply Chain Integrity, Transparency and Trust · SCITT working group

Supply Chain Integrity, Transparency and Trust: an IETF architecture for append-only transparency services that accept signed statements and return receipts proving a statement was registered.

SCITT ReceiptSCITT Receipt · SCITT Receipts

In SCITT, a signed proof from a Transparency Service that a Signed Statement was registered in its append-only log, usually carrying an inclusion proof.

Scope challengeinsufficient_scope challenge · scope challenges

An API's reply that the token lacks a needed permission, so the client can ask for just that permission and retry.

SD-JWTselective disclosure JWT · Selective Disclosure for JWTs · RFC 9901 · RFC9901

A signed JSON Web Token whose holder can reveal only some claims to a verifier (selective disclosure) and prove possession of the bound key. The Internet Engineering Task Force (IETF) format behind many verifiable credentials.

SEAT working groupSEAT WG

Internet Engineering Task Force (IETF) working group, Secure Evidence and Attestation Transport, chartered for attestation in Transport Layer Security (TLS).

SECSecurities and Exchange Commission · U.S. Securities and Exchange Commission

The US securities regulator whose 2023 cybersecurity disclosure rules require public companies to report material incidents within four business days and describe board oversight of cyber risk annually.

Secret rotationsecrets rotation · credential rotation · rotating secrets

Replacing a password, key or token with a new value on a schedule or after exposure, so a stolen copy stops working.

Secretlesssecretless authentication · going secretless

A way of authenticating machines that avoids stored secrets, using platform-issued identity such as SPIFFE or OIDC tokens instead.

Secrets managersecrets vault

A service that stores passwords, API keys and certificates and hands them to approved software on request, so they are not left in code or files.

Secure Patterns for Internet CrEdentials (SPICE) WGSPICE working group · SPICE

An IETF working group that develops profiles for digital credentials and presentations using existing Internet technologies, including selective disclosure, and coordinates related privacy and security work.

Security assertionbearer assertion

A signed package of identity and security information about a subject, with issuer, audience and expiry, that one security domain hands to another to prove who or what is asking.

security poverty linecybersecurity poverty line · security-poverty line

Wendy Nather's term for the budget level below which an organization cannot afford basic security.

Security Token ServiceSTS · Security Token Services

A service that verifies a caller and issues it a new, narrowly scoped, often short-lived token in place of the credential it presented. Examples include Amazon Web Services STS and the WS-Trust standard.

Selective disclosureselectively disclosable

Releasing a chosen subset of a credential's claims to a verifier while the verifier can still check authenticity.

Selective Disclosure CBOR Web Tokens (SD-CWT)SD-CWT · draft-ietf-spice-sd-cwt

IETF SPICE WG draft adapting SD-JWT selective disclosure to CBOR Web Tokens and COSE, with mandatory key binding.

Sender-constrainingsender-constraining · sender constraining · sender-constrained access tokens

Binding an access token to the key or certificate of the client it was issued to, so a stolen copy fails when presented by anyone else. DPoP and mTLS are the two OAuth mechanisms.

Separation of dutiesSoD · SOD · segregation of duties

A control that stops one person, or one agent, from holding two permissions that together allow fraud or error, such as creating and approving a payment.

Server-Side Request ForgerySSRF · Server Side Request Forgery

An attack that tricks a server into making web requests on the attacker's behalf, often to internal addresses the attacker cannot reach directly.

Service accountservice accounts · generic service account · shared service account

An account used by software rather than a person, such as an application talking to a database. Often long-lived, shared, and unowned, which is why it sits at the center of the non-human identity problem.

Service principalservice principals

An application's identity inside one Microsoft Entra tenant; it holds the app's permissions and credentials there.

Service provider (SCIM)SCIM service provider · SCIM SP

In SCIM (System for Cross-domain Identity Management), the system that receives provisioning requests and holds the user, group or agent records: the app or directory provisioned into, not the identity provider doing it.

Session fixationsession fixation

An attack in which the attacker plants a known session identifier in the victim's browser before login, so the attacker can reuse it once the victim authenticates. The usual fix is issuing a fresh session id at login.

Session hijackingsession hijack

An attack in which an unauthorised party obtains a valid session identifier and uses it to impersonate the original client and act on its behalf.

SHA-256SHA256

Secure Hash Algorithm 256 (SHA-256): a standard hash function that turns any data into a fixed 256-bit fingerprint, so changing the data changes the fingerprint.

Shadow ITshadow IT

Shadow IT: technology, such as unapproved SaaS tools or software, that employees adopt without the central IT or security team's review, approval or inventory.

Side-channel attackside channel · side-channel · side channels · timing side-channel

Learning a secret from indirect signals, such as timing or whether a step ran, instead of reading the secret itself.

SIEMsecurity information and event management

Security information and event management: the system a security team uses to collect logs from across the company and search them for signs of an attack.

Signed StatementSigned Statements · Transparent Statement · Transparent Statements

A COSE_Sign1 message by which an issuer makes a statement about an artifact; once a transparency service adds a receipt it becomes a Transparent Statement.

Signed Tree HeadSTH · STHs

A log's signed statement of its Merkle tree size and root hash at a time.

SIM cardSIM · SIM cards

The small chip card (Subscriber Identity Module, SIM) that ties a phone to a mobile number and carrier account. Whoever controls a SIM can receive that number's text-message codes.

SLSASupply chain Levels for Software Artifacts · Supply-chain Levels for Software Artifacts

A framework of graded build-integrity levels that shows software was built from the source it claims and not tampered with on the way to you.

SMS one-time codeSMS · SMS OTP · SMS text code · text-message code

A login code sent by text message. Still the most common second factor, and the weakest: it can be intercepted through SIM swapping or read off a phished page, which is why standards bodies now discourage it.

SOARsecurity orchestration, automation, and response · security orchestration automation and response

Security orchestration, automation and response: software that runs scripted playbooks so routine security alerts get handled automatically instead of by hand.

SOCsecurity operations center · security operations centre · SOC team · SOC teams

The security operations center: the team and tooling that watch alerts around the clock, decide what is an incident, and respond, including the credential resets and account blocks that identity systems carry out.

SOC 2SOC 2 Type II · SOC2

System and Organization Controls (SOC) 2: an audit report in which an outside accountant says whether a service company's security and privacy controls are designed and working as claimed.

Social engineering

Manipulating a target through persuasion rather than technical exploitation

Sockpuppetsockpuppets · sock puppet · sock puppets

A fake online account run by someone pretending to be a different person, often one of many run together.

Software statementsoftware statement · software statements · software_statement

A signed JSON Web Token in which a client developer or a third party vouches for facts about a piece of client software, such as its name and ID, which the client shows an authorization server when it registers.

SOXSarbanes-Oxley

The Sarbanes-Oxley Act, a US law that makes public companies prove who can change their financial systems and data. It is the reason many firms run access reviews.

SPIFFESPIFFE/SPIRE

Secure Production Identity Framework For Everyone: a set of standards, hosted as a graduated project of the Cloud Native Computing Foundation (CNCF), that issues short-lived cryptographic identities to workloads in place of shared secrets.

SPKI/SDSISPKI · SDSI · SPKI/SDSI

Simple Public Key Infrastructure (SPKI) and Simple Distributed Security Infrastructure (SDSI): a certificate model (RFC 2693) that binds authorization to keys rather than names.

SPRINDSPRIN-D · Federal Agency for Disruptive Innovation

Germany's Federal Agency for Breakthrough Innovation, which funds the open-source EUDI wallet work and staff who co-author identity standards.

SQL injectionSQLi

An attack that slips database commands into an input field so the application runs them. It is the classic example of untrusted input being treated as instructions, the same flaw behind prompt injection.

SSFShared Signals Framework · OpenID Shared Signals Framework

An OpenID framework for one system to push real-time security and identity events (e.g. a session was revoked, a device failed a check) to another over a lightweight, subscription-based channel. CAEP and RISC are event-set profiles built on top of it.

SSNSocial Security number

The US Social Security number, a nine-digit identifier often used as proof of identity in account opening and recovery.

SSOsingle sign-on · single sign on

Single sign-on: one login that grants access to many applications, with an identity provider vouching for the user to each of them instead of each keeping its own password.

Standing grantstanding grants · scoped standing grant

A permission issued ahead of time that covers a class of future actions, as opposed to approval of one specific action.

Step-up authenticationstep-up authentication · step up authentication

Asking a user to sign in again, more strongly or more recently, partway through a session because the request being made needs more assurance than the original login gave.

Step-up authorizationstep-up authorization flow · step-up auth · incremental authorization

Requesting additional OAuth scopes at runtime, on top of a token a client already holds, when an operation needs more permission than was originally granted — rather than restarting the whole authorization flow from scratch.

Storm-0501Storm 0501

Microsoft's tracking name for a ransomware group that now attacks cloud tenants through hybrid identity.

Subject Alternative NameSAN · URI SAN

A field in an X.509 certificate that lists the names the certificate stands for, such as a DNS name or a URI; workload-identity schemes put the workload's identifier there.

SVIDSPIFFE Verifiable Identity Document · SVIDs

SPIFFE Verifiable Identity Document: the short-lived credential, an X.509 certificate or a JWT, that a SPIFFE system issues to a workload to prove its identity.

Sybil attackSybil · Sybil attacks · Sybil resistance

One actor creating many fake identities to gain influence, dodge limits or escape a bad record.

Taint trackingtaint tracking mechanism

Marking data from untrusted sources and blocking it from reaching sensitive sinks, such as executable code, without sanitization.

Tamarin ProverTamarin · TAMARIN prover

A symbolic model checker that proves or finds attacks on security protocols under a Dolev-Yao attacker.

Tamper evidencetamper-evident

A property of a record whose alteration a verifier can detect from the record itself, without trusting whoever produced it.

TBACTask-Based Access Control

An access-control model that scopes an AI agent's permissions to the specific task it has been asked to perform, rather than to a persistent role, attribute set, or static list of OAuth scopes.

TDXIntel TDX · Trust Domain Extensions · Intel Trust Domain Extensions

Intel Trust Domain Extensions: a CPU feature that runs a virtual machine in hardware-isolated memory and can produce a signed report of what is running inside it.

The OAuth 2.1 Authorization Frameworkdraft-ietf-oauth-v2-1

IETF OAuth WG draft consolidating RFC 6749, 6750, PKCE, native-app guidance and the security BCP into one spec.

Threat modelingthreat model

Working out, before building, what could go wrong with a system and who might attack it.

TLSNotaryTLSNotary · TLS Notary

A protocol that produces cryptographic proof of a specific TLS session's request and response without exposing the whole session.

TOCTOUtime-of-check to time-of-use

Time-of-check to time-of-use: a race where something changes between the moment it is verified and the moment it is used. It can let a changed prompt or file slip past an approval.

TOFUtrust on first use

Trust on first use (TOFU): accepting the first key an entity presents and trusting it from then on, without verifying where it came from.

Token replaytoken theft · stolen token replay · refresh token theft

Using a stolen access, refresh or session token from the attacker's own machine to act as the user without re-authenticating.

Token Status ListOAuth Status List

A method from the Internet Engineering Task Force (IETF) for revoking tokens: the issuer publishes a compressed list of bits, and a verifier checks the bit for the token in front of it.

Toucan

A 1.5-million-sample synthetic tool-agentic dataset built from real-world MCP servers, used as a public comparison benchmark for tool-selection and task-to-scope matching research.

TPMTrusted Platform Module · trusted platform module

A tamper-resistant chip on a device's motherboard that stores keys and signs on the device's behalf, so an identity can be tied to one physical machine.

Transaction Token Chaining Profiledraft-fletcher-oauth-txn-token-chaining-profile · draft-fletcher-transaction-token-chaining-profile · Txn-Token chaining profile · Transaction Token Authorization Grant Profile · txn-chain+jwt

An individual IETF draft that lets a workload trade its in-domain transaction token at its own authorization server for a short-lived grant a partner domain will accept.

Transaction Token ServiceTTS

The service inside a trust domain that mints and signs transaction tokens when a request enters the domain.

Transaction Tokenstransaction token · txn-token · transactional tokens · draft-ietf-oauth-transaction-tokens · TXN

An Internet Engineering Task Force (IETF) OAuth draft for short-lived tokens that carry one transaction's context (who asked, for what, under which constraints) across internal services, so each hop enforces the original intent.

transparency logTransparency Log · append-only log

An append-only, publicly auditable record of events in which entries cannot be altered or removed without detection, proven by inclusion and consistency proofs.

Transparency ServiceTransparency Services

In SCITT, an entity that checks signed statements against a registration policy, appends them to a verifiable append-only log and returns signed receipts.

Trust AnchorTrust Anchors

Trust Anchor — the root authority in a federation whose signing keys a participant trusts directly; everything else is trusted by a signed chain back to it.

Trust ChainTrust Chains

Trust Chain — the ordered set of signed statements from one entity up to a Trust Anchor that proves it belongs to the federation.

Trust domain (workload identity)trust domains

The administrative boundary that issues and vouches for workload identities, named by an organization's domain; a verifier decides which trust domains it accepts.

Trust over IP FoundationToIP · Trust over IP

A Linux Foundation group that writes governance and technical models for digital trust ecosystems; with DIF it vets recommended DID methods.

Trust registry

Governed source of vetted issuers, state authorities and vocabularies used to decide whether an issuer is in good standing for a credential class.

Trusted Computing BaseTCB

The parts of a system (code and data) that must work correctly for its security guarantees to hold; the attacker is assumed unable to alter it.

Trusted execution environmentsecure enclave · trusted execution environments

A hardware-isolated area of a processor where code and keys run out of reach of the rest of the machine, able to sign evidence of what software it is running.

txn claimTXN claim · transaction identifier claim

Unique transaction ID claim in a Transaction Token, borrowed from Security Event Tokens, logged by every service for correlation.

UCANUser-Controlled Authorization Networks

A capability-token format in which each holder can delegate a narrower slice of its authority, verified offline by walking the chain of signed tokens.

UEBAuser entity behavior analytics · user and entity behavior analytics

User and entity behavior analytics: software that learns what normal activity looks like for each account or device and flags departures, originally built to catch insider threats.

UK National Cyber Security CentreNCSC · UK NCSC · National Cyber Security Centre

The UK government's cyber security authority, which publishes guidance to businesses and government.

Unlinkabilitypresentation unlinkability · verifier/verifier unlinkability

Property that adversaries cannot correlate separate presentations of one credential beyond the user's consent.

Updates to OAuth 2.0 Security Best Current Practicedraft-ietf-oauth-security-topics-update

An IETF draft updating the OAuth security best practices with newly found attacks.

Updates to OAuth JWT Client Authentication and Assertion-Based Grantsdraft-ietf-oauth-rfc7523bis · 7523bis · RFC 7523bis · RFC7523BIS

An OAuth working-group draft that tightens how a JWT used for client authentication names its audience, updating RFC 7523 and related RFCs.

Use of Remote Attestation with CSRsdraft-ietf-lamps-csr-attestation

An Internet Engineering Task Force Internet-Draft for encoding attestation Evidence inside certificate signing requests (CSRs) in the PKCS#10 (Certification Request Syntax) and CRMF (Certificate Request Message Format) forms.

VDS

Verifiable data structure: a log that can prove it only grows and shows everyone the same history. Transparency services build on it so past records cannot be rewritten.

Verifiable digital credential (VDC)VDC · VDCs · verifiable digital credential · verifiable digital credentials

A cryptographically verifiable digital version of identity attributes or a physical credential, held in a wallet and presented to a verifier; an umbrella term covering several credential formats.

verifiable LEI (vLEI)vLEI · verifiable Legal Entity Identifier

GLEIF's cryptographically verifiable version of the LEI: chained credentials that let a company, and people acting in named roles for it, prove that authority.

Verification methodverification methods

A public key or other material, listed in an identifier's document with its type and controller, that a verifier uses to check a proof made on behalf of the identifier's subject.

W3C

The web's standards body; in identity, home of WebAuthn, Decentralized Identifiers, and Verifiable Credentials.

W3C Verifiable Credentials Data Model v2.0VCDM2 · VC Data Model 2.0

W3C Recommendation defining the data model for verifiable credentials and presentations.

Web Infrastructure Model (WIM)WIM · Web Infrastructure Model

A formal, symbolic model of the web (browsers, servers, Domain Name System, a network attacker) in which web protocols are checked for security properties; used to analyse OAuth 2.0 and OpenID Connect.

Web of trust

Establishing that someone is real or trustworthy through vouching by people who already know them, instead of one central authority.

Web PKIWebPKI

The public certificate system that browsers use to trust websites, run by certificate authorities that browsers and operating systems approve.

WIMSEWorkload Identity in Multi System Environments

Workload Identity in Multi System Environments — an IETF working group standardizing how workload identities are expressed and trusted across clouds and trust domains.

WIMSE Architecturedraft-ietf-wimse-arch

The architecture document of the Internet Engineering Task Force (IETF) Workload Identity in Multi-System Environments (WIMSE) working group: how services prove who they are to each other across a multi-system environment.

WIMSE Workload Authentication Using Mutual TLSdraft-ietf-wimse-mutual-tls · WIMSE mTLS · Workload MTLS

A draft from the Workload Identity in Multi-System Environments (WIMSE) working group that profiles mutual Transport Layer Security (TLS) for workloads, with the workload identifier carried in the X.509 certificate.

WIMSE Workload Credentialsdraft-ietf-wimse-workload-creds · WIMSE-CREDS

A Workload Identity in Multi-System Environments (WIMSE) working-group draft defining the credentials a workload proves its identity with: a signed Workload Identity Token (WIT) and an X.509 Workload Identity Certificate (WIC).

WIMSE Workload Identifierdraft-ietf-wimse-identifier

A draft from the Internet Engineering Task Force (IETF) Workload Identity in Multi-System Environments (WIMSE) working group, defining the workload identifier format (wimse:// URIs) used to name a workload in a trust domain.

WIMSE Workload Proof TokenWPT · Workload Proof Token · draft-ietf-wimse-wpt · Authorization: WPT

An IETF draft proof-of-possession token for workloads: a short-lived signed JWT, carried in the Authorization header, that proves the caller holds the key behind its workload identity token and binds that token and any transaction token to one target URL.

WIMSE Workload-to-Workload Authentication with HTTP Signaturesdraft-ietf-wimse-http-signature · WIMSE HTTP signature · WIMSE HTTP Message Signatures

A Workload Identity in Multi-System Environments (WIMSE) working-group draft that has a workload sign each HTTP request with the key in its identity token, so the next hop can check the sender.

WITWorkload Identity Token

Workload Identity Token, the JWT form of a WIMSE workload identity credential, carrying the workload's identifier and public key so a peer can verify a signed request came from it.

Workload IdentifierWorkload Identifiers

The URI that names a workload, or a single instance of it, within a trust domain: its authority is the trust domain, and its path is set by that domain, opaque or structured.

Workload identityworkload identities

An identity for a running piece of software (a container, a service, a job) derived from what it is and where it runs, instead of a stored password. SPIFFE and WIMSE are the standards work behind it.

Workload Identity CertificateWIC

The WIMSE draft’s X.509 credential carrying one URI subject alternative name that identifies a workload within its trust domain.

Workload Identity Credential

Workload Identity Credential, in the IETF WIMSE framework: a signed credential (typically a JWT or X.509 certificate) that a workload presents to prove which workload it is.

Workload identity federationworkload federation

A workload proves who it is with a token from its own platform and trades it for an access token elsewhere, so no long-lived secret is stored.

WWW-AuthenticateWWW-Authenticate header · WWW-Authenticate response header

HTTP response header a server uses to challenge a client; RFC 9728 adds a resource_metadata parameter pointing at the resource's metadata.

X.509 PKI Certificate and CRL Profile (RFC 5280)RFC 5280 · RFC5280

IETF profile of X.509 certificates and CRLs for Internet PKI.

X.509-SVIDX509-SVID · SPIFFE-X509-SVID

The X.509 form of a Secure Production Identity Framework for Everyone (SPIFFE) Verifiable Identity Document (SVID): an X.509 certificate whose single URI Subject Alternative Name (SAN) carries the SPIFFE ID.

XACMLeXtensible Access Control Markup Language

An OASIS (Organization for the Advancement of Structured Information Standards) standard for writing fine-grained access-control policy as data and evaluating it centrally. The Attribute-Centric Authorization Language (ACAL) supersedes it.

YARAYARA rules · YARA rule

YARA: a pattern-matching rule language that malware analysts use to describe strings and conditions, then scan files or metadata to classify samples into families.

Zanzibar

Google's global, consistent authorization system for evaluating relationship-based access control (ReBAC) at scale, described in a 2019 USENIX ATC paper.

zCAP-LD (Authorization Capabilities for Linked Data)zCAP · ZCAP-LD · zcap

W3C CCG incubating spec for capability delegation with pluggable caveats.

Zero standing privilegesZSP · standing privileges · zero standing privilege

An access model where no account holds a usable permission at rest: every permission is granted just before it is used and revoked right after, so there is nothing sitting around for an attacker to steal or a reviewer to have to certify.

Zero trustzero-trust

A security model that verifies every request on its own merits instead of trusting anything because it sits inside the network.

Zero-knowledge proofZKP · zero-knowledge proofs · ZK

A way to prove a statement about a credential, such as being over 18, without revealing the credential itself.

zk-SNARKzk-SNARKs · SNARK

Succinct non-interactive zero-knowledge proof of a statement about hidden inputs

zk-STARKzk-STARKs

Zero-knowledge proof system built on hash functions (Scalable Transparent ARgument of Knowledge, STARK), believed to resist quantum attack rather than relying on elliptic-curve pairings.

AI

AI-native protocols and concepts that our research tracks.

A2AAgent2Agent · agent-to-agent protocol · A2A protocol

An open protocol for AI agents to discover and call one another. It is the peer of MCP, which connects an agent to tools.

Agent harness

The software around a model that turns it into an agent: the tools it may call, the context it is given, and the permissions and logging that bound it.

Agent Security BenchASB

Agent Security Bench (ASB): an academic benchmark, published at ICLR 2025, that measures how often attacks such as prompt injection and memory poisoning hijack LLM agents, and how well defenses hold.

Agent skillSKILL.md skill

A file-based package (SKILL.md with YAML frontmatter plus optional helper scripts) that extends a coding agent; runs locally with the user's privileges.

Agent Skillsagent skills · SKILL.md

Open format for packaging instructions, scripts and resources as a directory with a SKILL.md that agents load on demand.

AgentBenchAgentBench OS

A benchmark that scores LLMs as agents across environments; its Operating System suite has the agent do file and shell tasks.

AgentDojo

A benchmark environment for testing whether a large language model (LLM) agent completes its task while resisting prompt-injection attacks in tool outputs and external content.

AgentPoisonAgentPoison · AGENT POISON

A 2024 red-teaming attack that backdoors agents built on retrieval-augmented generation (RAG) and large language models (LLMs) by poisoning their memory with an optimized trigger.

AGNTCYAGNTCY

Open-source effort on infrastructure for an 'Internet of Agents': identity, discovery, messaging, observability.

AlpacaEval 2AlpacaEval2 · AlpacaEval 2 · AlpacaEval

A benchmark in which a GPT-4-based judge compares a model's answers with a reference model's answers.

AlpacaFarmAlpacaFarm benchmark

A set of 805 instructions, 208 of them with a data part, used to evaluate instruction-following language models.

Amazon Bedrock AgentCoreAWS Bedrock AgentCore

AWS managed platform for deploying and operating AI agents (runtime, memory, identity, gateway).

Anthropic InterviewerAnthropic Interviewer · Anthropic Interviewer dataset

Anthropic's AI-conducted interview tool and its released 125-scientist transcript dataset.

ARGUS (causal-provenance auditor)ARGUS auditor · ARGUS

Runtime auditor that releases a state-changing agent action only if benign context spans entail it and task invariants hold.

Attack success rateASR

The share of attack attempts in a test that achieve the attacker's goal.

AuthBench

A benchmark of 120 terminal tasks that tests whether coding agents can infer the task-level permissions they need, no more and no less.

AutoGen

Microsoft's open-source framework for building multi-agent conversational AI applications.

Berkeley Function Calling LeaderboardBFCL · BFCL V3

Benchmark of LLM function-calling accuracy (single-turn, multi-turn, relevance).

BIPIABenchmark for Indirect Prompt Injection Attacks

A benchmark and fine-tuning defense for indirect prompt injection, by Microsoft-affiliated researchers.

CAISICenter for AI Standards and Innovation · US AI Safety Institute · US AISI · U.S. AI Safety Institute

The NIST center for AI standards and testing, successor to the US AI Safety Institute.

Chain of thoughtCoT · chain-of-thought · raw CoT · reasoning trace

A model's written-out reasoning steps, produced before its answer or action; the text that chain-of-thought monitoring reads.

Chain-of-thought monitoringCoT monitoring · CoT monitors · chain-of-thought monitors

Automated review of a model's written-out reasoning while it works, used to flag misaligned or dangerous intentions and page humans before the actions land.

Claude 3.5 Sonnetupgraded Claude 3.5 Sonnet

Anthropic's Claude 3.5 Sonnet model, released in June 2024; an upgraded version followed in October 2024.

Claude 4 SonnetClaude-4-Sonnet

Anthropic's Claude 4 Sonnet large language model.

Claude Desktop

Anthropic desktop host app for MCP servers.

CLAUDE.mdCLAUDE.md

A project file that Claude Code reads for formatting rules, design principles and architectural constraints before it works on a repository.

ClawHubClawHub

OpenClaw's skill registry.

CrewAI

An open-source framework for orchestrating teams of role-based AI agents.

Data poisoningtraining data poisoning

Corrupting the data a model learns from so that it later behaves as an attacker wants.

Devin

Autonomous AI software-engineering agent that opens pull requests on GitHub.

Direct preference optimizationpreference optimization

A way to tune a model toward preferred answers and away from rejected ones using pairs of answers, without a separate reward model.

DPO

Direct preference optimization, a way to fine-tune a language model on pairs of preferred and rejected answers without a separate reward model. Prompt-injection defenses use it to train a model away from obeying instructions planted in its input.

ElizaOSEliza OS · ElizaOS framework

Open-source framework for building autonomous on-chain AI agents, often run in trusted execution environments (TEEs).

EU AI ActEuropean AI Act

The European Union's risk-tiered AI regulation, which for higher-risk automated decisions calls for provenance and bias-mitigation evidence an operator can produce on demand.

ExploitGym

A cybersecurity evaluation in which an AI agent must exploit a piece of software to retrieve a hidden flag.

False positive rateFPR

The fraction of actually negative cases incorrectly classified as positive; for a threat detector, benign cases incorrectly flagged as threats.

FIDES

Microsoft Research's information-flow control for AI agents: it labels data by how far it can be trusted and blocks tool calls that would let untrusted content steer a sensitive action.

Foundation modelfoundation models

A large pre-trained model, such as GPT-4, that is adapted to many tasks.

Gemini 2.5 FlashGemini-2.5-Flash

Google model.

Generative AIGenAI · Gen AI

AI that produces new text, images, audio or code from a prompt, as large language models do.

GPTgenerative pre-trained transformer

Generative pre-trained transformer: the family of large language models behind OpenAI's ChatGPT. Reports name versions such as GPT-4 and GPT-5.

GPT-3text-davinci-002

OpenAI's large language model family released in 2020, short for Generative Pre-trained Transformer 3.

GPT-3.5-TurboGPT-3.5 Turbo · gpt-3.5-turbo

OpenAI's chat model in the GPT-3.5 (Generative Pre-trained Transformer) family.

Greedy Coordinate Gradient (GCG)Greedy Coordinate Gradient · GCG

An attack that searches for a string of odd-looking tokens to add to a prompt so a model does what the attacker wants, first built for jailbreaks.

GUI agentGUI agents · computer-use agent

An AI agent that acts by reading screenshots of a graphical interface and issuing clicks, keystrokes and swipes.

Instruction hierarchyinstruction-hierarchy defense

A training technique that teaches a model to rank instructions by source, so text from a tool or web page loses to the system prompt.

Intent drift

An agent's actions gradually moving away from what the user asked, through its own reasoning and with no attacker involved.

ISO/IEC 42001ISO 42001 · ISO/IEC 42001:2023 · AI management system standard

The first international management-system standard for AI (December 2023): what an organisation must put in place to govern how it builds and uses AI, audited the way ISO 27001 audits security. Often cited for AI-agent governance; it is not a technical spec for agents.

IsolateGPTSecGPT · Isolate GPT

Hub-and-spoke architecture that runs each LLM app in its own process with its own LLM and routes cross-app messages through a trusted hub with user consent.

Jailbreakjailbreaking · universal jailbreak · non-universal jailbreak

A prompt or technique that gets an AI model to bypass its safety safeguards; a 'universal' jailbreak works broadly across topics, a narrow one only in specific circumstances.

LangGraph

A LangChain framework for building stateful, multi-agent orchestration graphs.

LangSmith

LangChain's hosted tracing and evaluation platform for LLM applications and agents.

LLMlarge language model · large language models

The text-predicting AI model behind chatbots and agents. It follows instructions statistically, which is why an instruction hidden in its input can redirect it.

LLM-as-a-judgeLLM as a judge · LLM-as-judge

Using a second language model to review another model's output or planned action and approve or block it; a probabilistic check, since the judge can itself be fooled.

Logic-layer Prompt Control InjectionLPCI

Covert attack class where payloads hidden in tools and memory steer an agent's logic layer

Machine learning

Software that learns patterns from data instead of following hand-written rules; large language models are one kind.

MCIPMCIP Guardian · Model Contextual Integrity Protocol

A defense for MCP agents that fine-tunes the model to output a full tool-call trace, which a guardian service then checks to say whether an invocation is poisoned.

MCP registryMCP registries

A searchable directory where Model Context Protocol (MCP) servers are published and found, such as Smithery or Glama.

MCP serverMCP Server · MCP servers

The tool-providing side of the Model Context Protocol (MCP): exposes tools, resources and prompts to an MCP client, often running locally with the user's privileges.

mcp-remote

Open-source package that lets local Model Context Protocol (MCP) clients connect to remote MCP servers.

MCPToxMCPTox benchmark

Research benchmark testing whether malicious instructions in MCP tool descriptions can redirect an agent through legitimate tools.

MELONMELON

Indirect prompt injection defense that compares masked re-executions of tool calls.

Meta SecAlignMeta SecAlign · Meta-SecAlign

Meta's open-weight language model with 70 billion parameters, trained with SecAlign, a defence against prompt injection.

MMLU

Massive Multitask Language Understanding (MMLU): a benchmark of multiple-choice questions across 57 subjects that measures a language model's general knowledge.

Model Context ProtocolMCP

Model Context Protocol: an open standard, originated by Anthropic, for connecting AI assistants to tools and data.

Model-level defensemodel-level defenses · model-level prompt injection defense

A prompt injection defense built into the model by training it to prioritise trusted instructions over instructions found in data, as opposed to a filter or policy around it.

Multi-agent systemmulti-agent system · multi-agent systems

A set of agents built on large language models (LLMs), coordinated by an orchestrator or fixed order to complete tasks for a user.

NIST AI 100-1AI RMF · AI Risk Management Framework · NIST AI RMF

The National Institute of Standards and Technology (NIST) AI Risk Management Framework (govern, map, measure, manage) for identifying and mitigating AI system risk.

OpenAI Atlas browserAtlas browser · OpenAI Atlas

OpenAI's agentic web browser.

OpenAI OperatorChatGPT Operator

OpenAI's agent that uses its own web browser to carry out tasks for a user.

OWASP AI Security Verification Standard (AISVS)AISVS · AI Security Verification Standard

An OWASP checklist of technically verifiable security requirements for AI systems, in levels 1 to 3, that sits beside ASVS.

OWASP AI Vulnerability Scoring SystemAIVSS · OWASP AIVSS

OWASP's scoring framework for AI and agentic vulnerabilities, adding agent-specific fields such as memory use and autonomy to severity scoring.

OWASP Top 10 for LLM Applications 2025OWASP LLM Top 10 · LLM01:2025 · LLM06:2025 Excessive Agency

OWASP's 2025 ranking of LLM application risks, including prompt injection (LLM01) and excessive agency (LLM06).

Project NANDANANDA

MIT-linked initiative exploring an 'Internet of Agents' (agent discovery and identity).

Qwen

Alibaba's family of large language models.

RAGretrieval-augmented generation

Retrieval-augmented generation: an AI model is handed relevant documents at question time so it answers from your data instead of memory.

RAS-Eval

A benchmark for the security of agents built on large language models (LLMs) in real-world environments.

ReAct agent loopReAct framework · ReAct agent

An agent design that alternates the model's written reasoning with tool actions and the observations they return.

Reinforcement learning from human feedbackRLHF · reinforcement learning from human feedback

Training a model with human ratings of its answers so it behaves the way people prefer.

Reward hackingreward hack · reward hacks · cheating on tasks

When an AI agent earns a high score by an unintended shortcut, such as looking up the answer or tampering with the grader, instead of doing the task as designed.

RLreinforcement learning · Reinforcement learning

Reinforcement learning, a way of training a model by rewarding the outcomes of its actions rather than showing it correct answers.

SecAlignSecAlign

Fine-tuning defense that uses preference optimization to make LLMs resist prompt injection.

SEP (MCP)Specification Enhancement Proposal · SEPs · Spec Enhancement Proposal

The Model Context Protocol project's change process: a numbered proposal (matching its GitHub pull request) that a sponsor shepherds to acceptance before a change enters the specification.

Skills.restskills.rest

Community registry of agent skills indexed from GitHub.

SkillsMPskillsmp.com

Community marketplace that aggregates agent skills from public GitHub repositories.

Spotlightingspotlighting_with_delimiting

A prompt-injection defense that marks untrusted text, for example with delimiters, so the model can tell data from instructions.

Streamable HTTPStreamable HTTP transport · HTTP+SSE · MCP Streamable HTTP

The Model Context Protocol's HTTP transport, which replaced the earlier HTTP plus server-sent-events transport; the 2026-07-28 revision made it stateless.

StruQstructured queries · structured instruction tuning

A prompt-injection defense that sends the prompt and the untrusted data to the model as two separate inputs, and trains the model to obey only the prompt.

Supervised fine-tuningsupervised fine-tuning · SFT

Further training of a pre-trained model on example inputs with the answers wanted, the usual step before preference tuning.

Vibe codingvibe coded · vibe-coded

Building software by describing what you want to an AI coding tool and accepting what it produces, often with little review.

Web agentweb automation agent · browser agent

LLM-driven agent that perceives and acts on web pages through a browser

τ²-benchtau2-bench · τ2-bench · tau-squared bench

Benchmark of conversational customer-service agents with tools and written policies (airline, retail, telecom).

Agents that pay

The payment protocols and card-network schemes that let an agent spend, and the terms they use for who allowed it.

Agentic commerce

Shopping and payment tasks that an AI agent researches, selects or completes on a user's behalf instead of the user working through a checkout.

Agentic Commerce ProtocolAgentic Checkout Specification

An open specification maintained by OpenAI and Stripe for an AI agent completing a purchase on a merchant's own checkout and payment rails: REST APIs for checkout sessions, delegated payment tokens and delegated authentication.

AP2Agent Payment Protocol

Google's open protocol for letting AI agents make payments on a person's behalf with signed proof of what they approved.

Chargebackchargebacks

A card-network dispute process in which the issuing bank reverses a payment after a cardholder complains, for example about non-delivery or fraud, and the merchant must contest it.

Decoupled AuthenticationDecoupled Auth · decoupled auth

An EMV 3-D Secure option where the cardholder confirms a payment outside the checkout, for example by approving it later in a banking app, instead of being challenged in the middle of the purchase. The payments world's version of a backchannel approval.

Delegated AuthenticationDelegated Auth · delegated auth

An EMV 3-D Secure option where the merchant or wallet, not the card issuer, checks the cardholder and passes the result to the issuer in the authentication request. It only works where the issuer has agreed to trust that party's check.

EMVEMV specifications · EMV standard

Europay, Mastercard and Visa (EMV) specifications: the family of card-payment specifications, from chip cards to online checkout, written by EMVCo for the card networks.

EMV 3-D Secure3-D Secure · 3D Secure · 3DS · 3DS2 · EMV 3DS · Three-D Secure · 3-D Secure 2

The card networks' protocol for checking that an online card payment really comes from the cardholder. Version 2 lets the bank score the risk quietly and only interrupt the shopper (a "challenge") when something looks off.

EMV Secure Remote CommerceSRC · Secure Remote Commerce · EMV SRC · SRC Core Specification

EMVCo's specification for a shared e-commerce checkout, the 'click to pay' flow, in which a consumer's stored cards are presented by a network-run system rather than typed into each merchant's form.

EMVCo

The body owned by the big card networks (Visa, Mastercard, American Express and others) that writes the chip-card and 3-D Secure specifications.

Facilitator (x402)x402 facilitator · facilitator service

In x402, the third-party or self-hosted service that checks a signed payment authorization and submits it to the blockchain on the resource server's behalf. The server trusts its answer at the moment it hands over the resource.

L402L402 protocol

A Lightning Labs protocol that uses HTTP 402 Payment Required with macaroons and Lightning payments so a client pays per request for a web resource.

Machine Payments ProtocolMPP

Stripe and Tempo open standard for a challenge-credential-receipt flow over HTTP 402

Mandate (AP2)Checkout Mandate · Payment Mandate · open mandate · closed mandate · Mandate Pair

In Google's Agent Payments Protocol (AP2), a signed statement of what a person authorized an agent to buy and pay, either naming the exact checkout (closed) or setting limits (open).

PSD2 Strong Customer AuthenticationPSD2 · Strong Customer Authentication · dynamic linking

Strong customer authentication under the EU's Payment Services Directive 2 (PSD2): an online payment must be approved with two separate factors and tied to the exact amount and payee.

Shared payment tokenshared payment token · shared payment tokens · SharedPaymentToken

Stripe's payment credential for agent purchases: a token issued from a customer's saved payment method, capped by seller, amount and expiry, that the agent can revoke until the seller uses it.

Spending budgetgranted budget · AAuth budget

A ceiling on what an agent may consume at one resource, in a resource-declared unit, carried in the auth token and enforced by the resource.

Stablecoinstablecoins · USDC

A crypto token pegged to a currency, usually the US dollar, so a payment settles on a blockchain without the price moving.

Trusted Agent ProtocolVisa TAP · Visa Trusted Agent Protocol

Visa's protocol for letting a merchant tell a trusted artificial intelligence (AI) shopping agent from a bot: the agent signs each web request with HTTP Message Signatures.

Universal Commerce ProtocolUCP

A Google-led open protocol letting an AI agent (e.g. inside Gemini) browse a merchant's catalog, authenticate the shopper, and complete checkout without the shopper visiting the merchant's site — the commerce-checkout equivalent of MCP.

Verifiable IntentMastercard verifiable intent · agent-intent/verifiable-intent · VI

A specification for recording, in a chain of signed credentials, what a person authorised an AI agent to buy, so a merchant or payment network can check a purchase against that record.

Verify-then-PayVerify-then-Pay

Settlement model in which funds sit in escrow and release only after evidence of task execution verifies.

x402HTTP 402 payment required

A payment protocol that revives the HTTP 402 'Payment Required' status code so an AI agent or other client can pay for a web resource in the request itself.

Foundations

The general web and computing standards that identity specs are built on.

ABNFaugmented Backus-Naur form

Augmented Backus-Naur Form, the notation IETF specifications use to define exactly which strings a protocol field may contain.

API gatewayAPI gateways

Application programming interface (API) gateway: a front door that routes, authenticates and rate-limits API traffic before it reaches the services behind it.

ASCIIRFC 20 · RFC0020

American Standard Code for Information Interchange (ASCII), the 7-bit character encoding defined in RFC 20.

ASN.1ASN.1 · Abstract Syntax Notation One

Abstract Syntax Notation One: the schema language X.509 certificates and requests are defined in.

base64urlbase64url

A way to write binary data as URL-safe text using letters, digits, hyphen and underscore; JSON Web Signature uses it so signed content can travel inside JSON and URLs.

BCP

Best Current Practice: an Internet Engineering Task Force document series for agreed operational guidance, such as BCP 201 on cryptographic algorithm agility.

CAIP-2Chain Agnostic Improvement Proposal 2 · CAIP-2 network identifier

A namespace:reference string, such as eip155:8453, that names a blockchain network the same way across tools; x402 v2 uses it for every network field.

Canonicalizationcanonicalization contract

Reducing data to one fixed byte form so that two parties who sign or hash the same content get the same result.

CBORConcise Binary Object Representation · RFC 8949

A compact binary data format, a binary cousin of JSON, used with COSE to carry signed data.

CDDL (RFC 8610)CDDL · Concise Data Definition Language

Concise Data Definition Language: notation to describe CBOR and JSON data structures.

CDNCDN · content delivery network

A content delivery network: servers spread across many locations that cache and serve a site close to its users.

CIcontinuous integration · CI pipeline

Continuous integration, the automated pipeline that builds and tests code on every change, usually paired with continuous delivery (CI/CD) to deploy it. Its runners and jobs hold credentials of their own, which makes them non-human identities.

CI/CDbuild pipeline · build pipelines

Continuous integration and continuous delivery (CI/CD): the automated pipelines that build, test and deploy software, often holding machine credentials of their own.

Constrained Application Protocol (RFC 7252)CoAP · Constrained Application Protocol · RFC 7252 · RFC7252

A lightweight request/response protocol for constrained devices, a binary cousin of HTTP.

CRMcustomer relationship management

Customer relationship management: the system a company keeps its customer records, sales deals and support history in, such as Salesforce. It holds data many AI agents are being asked to read and change.

Datalog

A small logic query language that Biscuit tokens use to write authorization policies.

DevOpsDevSecOps

The practice and teams that build and run software together, shipping changes fast; their pipelines hold many machine secrets.

Digest Fields (RFC 9530)RFC 9530 · RFC9530 · Content-Digest

HTTP header fields (Content-Digest, Repr-Digest) that carry a hash of a message body so a signature can cover the body.

Directed acyclic graph (DAG)DAG · directed acyclic graph

A set of linked items where every link points one way and no chain of links loops back to its start, used to model task steps or content-addressed data.

Distributed hash tableDHT

A distributed system that maps keys to values across participating nodes.

DNSDomain Name System

The Internet's naming system: it maps names to typed records, such as addresses and TXT text records, that clients can look up.

DNS-SDDNS-Based Service Discovery

A way to find named services on a network through Domain Name System (DNS) records.

DNSSECDNS Security Extensions

DNS Security Extensions: cryptographic signatures on DNS answers so a resolver can tell a genuine record from a forged one.

DockerDocker container · Docker containers

A container runtime that runs software in isolated environments that can be started, used and destroyed.

ES256ECDSA P-256 with SHA-256

A common JSON Web signature algorithm: ECDSA on the P-256 curve with SHA-256.

ES512ECDSA P-521 with SHA-512

A JSON Web Signature algorithm: Elliptic Curve Digital Signature Algorithm (ECDSA) on the P-521 curve with SHA-512.

Ethereum Virtual MachineEVM

The execution environment that runs smart-contract bytecode and applies its state changes on Ethereum and compatible chains.

Extensible Provisioning Protocol (EPP)RFC 5730 · EPP

The protocol domain-name registrars use to create, renew, transfer and delete names at a registry.

FQDNfully qualified domain name · fully-qualified domain name

A fully qualified domain name: the complete Domain Name System (DNS) name of a host or organization, such as example.com.

GraphQL

A query language and specification for APIs in which the caller asks for exactly the fields it needs.

HPE iLOHPE iLO 7 · iLO 7

HPE's server management controller with a hardware root of trust, named by the draft as an example Tier 1 root.

HTTPHypertext Transfer Protocol

Hypertext Transfer Protocol: the request-and-response protocol of the web, in which a client sends a method such as GET or POST to a URL and receives a status code and body.

ICANNInternet Corporation for Assigned Names and Numbers

The non-profit that coordinates internet domain names and accredits the registrars that sell them.

Idempotency keyIdempotency-Key · idempotency

A unique value a client attaches to a request so that a retry of the same request produces the same result once, not a second payment or a second token. For agents that retry on their own, it is the control that stops duplicates.

IoTInternet of Things

Internet of Things: network-connected devices such as sensors, appliances and medical equipment, often shipped with default credentials and long service lives.

IPFSInterPlanetary File System

Content-addressed peer-to-peer file storage, where files are retrieved by a hash of their contents.

Istio

Kubernetes service mesh that secures traffic between workloads with mutual Transport Layer Security (mTLS).

ITSMIT service management

IT service management: the tools and processes, such as ticketing systems, that a company uses to run and support its IT services.

JavaScript Object Notation (JSON)JSON · RFC 8259 · RFC8259

The standardized JSON data-interchange syntax, defined in RFC 8259.

JSON Canonicalization SchemeJCS · RFC 8785 · RFC8785

RFC 8785: one fixed way to serialize JSON, so two parties that hash or sign the same data get the same bytes.

JSON SchemaJSON Schemas

A vocabulary for describing and validating JavaScript Object Notation (JSON) documents.

JSON-LDJSON for Linked Data

JSON for Linked Data: a way of writing JSON whose field names point to shared vocabularies, so different systems agree on what a claim means; the W3C credential and DID specifications use it.

JSON-RPC 2.0JSON-RPC · JSON-RPC 2.0

A lightweight remote-procedure-call format that wraps each request and response in JavaScript Object Notation (JSON); the Model Context Protocol (MCP) and Agent2Agent (A2A) both use it for messages.

JWTJSON Web Token · JSON Web Tokens · RFC 7519

A compact, signed token that carries claims (who, for what, until when) between systems. It is the usual shape of an OAuth access token.

KubernetesK8s

An open-source system for running and scheduling containerized workloads. Widely used as the deployment substrate for services and, increasingly, AI agent workloads — which is why workload-identity standards like SPIFFE/SPIRE are usually described in Kubernetes terms.

Markdown

Markdown is a lightweight plain-text markup syntax that converts to HTML, used for documentation and agent instruction files.

Merkle treeMerkle inclusion proof · Merkle inclusion material

A hash tree whose root commits to every leaf, so a short path of hashes proves one entry is included without revealing the rest.

MIMEMultipurpose Internet Mail Extensions

Multipurpose Internet Mail Extensions: the labelling system that tells software what kind of data a message or file holds, such as text/plain or application/json.

mTLSmutual TLS · mutual-TLS · mutually-authenticated TLS

Mutual TLS: both sides of a connection present and verify a certificate during the TLS handshake, so the server authenticates the client as well as the client authenticating the server.

Nonce

Single-use random value that proves a response is fresh and not replayed.

Object identifierOID

A globally unique identifier, written as a dotted number sequence and defined in Abstract Syntax Notation One (ASN.1), used to name algorithms, certificate extensions and other registered objects.

OEMoriginal equipment manufacturer · original equipment manufacturers

Original equipment manufacturer: the company that builds a device or its chips. In attestation, the OEM often installs the device's first identity and keys at the factory, so later trust in the device starts with trust in that maker.

OpenAPI SpecificationOpenAPI · OpenAPI Specification

A machine-readable format for describing HTTP APIs.

OpenTelemetryOTel

Vendor-neutral, open-source framework and specification for collecting traces, metrics and logs from software.

P-384secp384r1

A 384-bit elliptic curve standardized by the National Institute of Standards and Technology (NIST), also called secp384r1, used for digital signatures and key agreement.

P9999th percentile · p99 latency

The 99th-percentile response time: the time within which 99 percent of requests finish. Engineers use it to judge how much delay a token exchange or policy check adds at scale.

PyPIPython Package Index

The official public registry for publishing and installing Python software packages; a frequent target for software-supply-chain attacks such as typosquatting or dependency confusion, since anything published there can be installed and run automatically by whoever names the dependency.

QR codeQR

A printed square barcode that a camera reads as text, such as a link or an instruction.

RDAPRegistration Data Access Protocol · RFC 9083 · RFC 9082

The JSON-over-HTTPS protocol for looking up who registered a domain name or IP block; it replaced WHOIS.

Remote procedure callRPC

A request that asks another process to run a procedure and return its result.

RFCRequest for Comments · RFCs

Request for Comments: the numbered document series in which the Internet Engineering Task Force publishes internet specifications, from standards such as RFC 8693 to informational notes. A draft is not an RFC.

Robots Exclusion Protocol (RFC 9309)robots.txt · RFC 9309 · Robots Exclusion Protocol

A plain-text file at a website's root, defined in RFC 9309, in which the site owner tells automated crawlers which paths they may fetch.

SDLCsoftware development lifecycle · secure software development lifecycle

The software development lifecycle: the stages from design through code, test and release that a software team follows.

Server-Sent Events (SSE)SSE · Server-Sent Events

A web feature where a server keeps one HTTP connection open and streams updates to the client. Agent protocols use it for live task updates, so whoever can open the stream can read the session.

Service meshservice meshes · sidecar proxy

Infrastructure that puts a proxy beside each service to handle its network security, usually issuing certificates and enforcing mutual TLS.

Sidecar

A helper process run beside an application, in the same pod or host, that handles cross-cutting work such as proxying traffic or fetching credentials, so the application code does not change.

smart contract

A program stored on a blockchain that carries out its agreed rules automatically when their conditions are met.

SMTPSimple Mail Transfer Protocol

Simple Mail Transfer Protocol, the standard way mail servers send email to each other.

SPAsingle-page application · single-page applications

Single-page application: a web app that runs in the browser and calls APIs directly. It cannot keep a secret, so it needs sign-in flows built for public clients.

SRV recordSRV · DNS SRV record

Service record: a Domain Name System (DNS) record that tells clients which host and port offer a named service over a given protocol.

SSHSecure Shell · SSH keys

The protocol administrators and automated jobs use to log in to servers; its long-lived keys are a common unmanaged machine credential.

SSLSecure Sockets Layer

Secure Sockets Layer, the encryption protocol that TLS replaced. Every SSL version is deprecated as insecure, but the name survives in everyday use for TLS certificates and connections.

Standard input/outputSTDIO · stdio

A process’s input and output streams; MCP’s stdio transport exchanges JSON-RPC messages with a server subprocess through these streams.

Structured Field Values for HTTPRFC 9651 · RFC9651 · Structured Fields · HTTP-SF

RFC 9651: a typed grammar for HTTP header values (lists, dictionaries, strings, tokens) so new headers can be parsed without a bespoke format each time.

SVCBHTTPS records

Service Binding, a DNS record type that tells a client how to reach a service, including its endpoints and settings.

TCPTransmission Control Protocol

Transmission Control Protocol: a transport protocol that provides ordered, reliable byte streams between network endpoints.

The I-JSON Message Format (RFC 7493)RFC 7493 · RFC7493 · I-JSON

I-JSON (Internet JSON), RFC 7493: a restricted profile of JSON (JavaScript Object Notation) that avoids interoperability traps, such as integers too large to survive a round trip.

TLA+TLA+ · TLC model checker

Temporal Logic of Actions (TLA+): a formal specification language and model checker for state machines.

TLS

Transport Layer Security, the protocol that encrypts traffic between two systems and lets each check the other's certificate. Mutual TLS (mTLS) makes both sides prove who they are, which is how many workload identity schemes authenticate services.

TLS terminationTLS-terminating proxy · TLS terminating

Ending a Transport Layer Security (TLS) connection at a proxy or load balancer, which forwards the request on a new connection; the caller's certificate does not travel past it.

TTLtime to live

Time to live: how long a credential, token or cached record stays valid before it must be renewed or thrown away.

TXT recordDNS TXT record

A Domain Name System (DNS) record that holds free-form text for a domain; protocols use it to publish keys, policies or pointers anyone can look up, such as for email authentication or domain-ownership checks.

URIUniform Resource Identifier · URIs

Uniform Resource Identifier, the general form of a web address. Identity specs use URIs to name things unambiguously: an issuer, an audience, a resource, or a workload's identity in SPIFFE.

URNUniform Resource Name

Uniform Resource Name: a permanent, location-independent name for something, such as a token type, written as urn:… .

UUIDuniversally unique identifier

Universally unique identifier: a 128-bit value that systems can generate on their own, with a very low chance of two ever matching.

Virtual private cloudVPC

A logically isolated virtual network for resources hosted in a public cloud.

VPNvirtual private network

Virtual private network: an encrypted tunnel that makes a remote device look like it sits inside an organisation's own network.

W3C PROVW3C PROV

The World Wide Web Consortium's Provenance (PROV) specifications for recording how data came to be: which entities, activities and agents were involved, as a graph that can be queried.

WebAssembly (WASM)WASM · WebAssembly

A portable, capability-restricted bytecode runtime often used to sandbox untrusted code.

Webhookwebhooks · webhook delivery · callback URL

A way for one service to notify another: the receiver gives the sender an HTTPS address, and the sender posts a signed message there when something happens, instead of the receiver polling.

X.509X509 · X.509 certificate · X.509 certificates

The standard format for public-key certificates: a signed document that binds a name to a public key, used in TLS and in workload identity as the certificate form of a credential.

YAMLYAML Ain't Markup Language

YAML (YAML Ain't Markup Language) is a human-readable data serialization format used for configuration files and document frontmatter.