Practical reference

Digital intelligence glossary

430 concepts explained through practical use, with a guide for taking each topic further.

Compare concepts

Distinctions for choosing the right check. Each pair connects definitions, a concrete situation and a method.

Authentication / Authorisation
Authentication
Verify the identity of a person or service.
Authorisation
Decide which actions are permitted on a resource.

Fictional example In this example, a person can read a record without permission to delete it. A successful login does not settle that second permission.

Explore the method — Identity and account recovery: govern the complete lifecycle

Reference : OWASP — Authorization [EN]

Reference : NIST — SP 800-63B-4, Authentication [EN]

Backup / Restoration
Backup
Keep a copy with a defined scope and frequency.
Restoration
Recover data and verify that the service works.

Fictional example A copy exists, but a required key is missing. A successful copy therefore does not demonstrate service recovery.

Explore the method — Test a backup restoration from copy to working service

Reference : NIST — SP 800-34 Rev. 1 (2010) [EN]

Latency / Throughput
Latency
Time to process a request, depending on the measurement point.
Throughput
Number of operations processed per unit of time.

Fictional example An API can respond in 200 ms and process 100 requests per second through concurrent work. The two values are not simple reciprocals.

Explore the method — Make API integrations resilient to outages

Reference : Google SRE — Monitoring Distributed Systems [EN]

Source / Repost
Source
An identified origin of information or an observation.
Repost
A publication relaying previously published information.

Fictional example Ten articles repeat one press release: ten reposts do not demonstrate ten independent confirmations. Trace the origin before judging confidence.

Explore the method — OSINT guide: produce reliable decision intelligence

Reference : Berkeley / OHCHR — Berkeley Protocol [EN]

Indexing / AI citation
Indexing
A page being present in a search engine index; display is not guaranteed.
AI citation
A link or reference observed in a generated answer for a particular question and context.

Fictional example An indexed page may never be cited in the answers you observe. Keep the question, tool, language, date and passage to describe the observation.

Explore the method — Measure visibility in AI-assisted search

Reference : Google Search Central — AI features and your website [EN]

Lab measurement / Field measurement
Lab
A controlled scenario measurement useful for reproducing a cause.
Field
Measurement of real experiences with a defined scope, device mix and period.

Fictional example A fast desktop test need not contradict slow mobile LCP. Compare documented conditions and populations before drawing a conclusion.

Explore the method — Diagnose Core Web Vitals in field data

Reference : web.dev — Web Vitals [EN]

Hash / Digital signature
Cryptographic hash
A computed value for comparing bytes. A hash alone does not identify the file’s author.
Digital signature
Verification of signed data using a public key whose association with the signer must be established.

Fictional example On a compromised website, a file and its hash are replaced together: they match without establishing origin. Verify a signature using a key obtained through a trusted channel; it does not establish that the content is true.

Explore the method — Information integrity: verify provenance, context and transformations

Reference : NIST — FIPS 186-5, Digital Signature Standard [EN]

Encoding / Encryption
Encoding
A representation change, such as Base64, reversible without a secret key and providing no confidentiality guarantee.
Encryption
A cryptographic transformation intended to make data unreadable without the appropriate decryption key.

Fictional example A Base64 attachment remains readable after decoding. Protect a confidential document with suitable encryption and controlled keys and recipients; encryption alone does not decide who is authorised to access the document.

Explore the method — Secure, high-performance web: a journey-led audit guide

Reference : IETF — RFC 4648, §12: Security Considerations [EN]

CVSS / EPSS
CVSS
Describes a vulnerability’s technical severity using a specific version and metric vector.
EPSS
Estimates the probability that a published CVE will be exploited in the next 30 days, using a dated model.

Fictional example CVSS 9.8 and EPSS 0.02 describe two dimensions: severity and an estimated 2% probability. Neither replaces local exposure and impact.

Explore the method — Prioritize vulnerabilities and verify remediation

Reference : FIRST — CVSS v4.0 [EN]

Reference : FIRST — EPSS [EN]

TLP / Encryption
TLP
Indicates with whom information may be shared under the source’s rules.
Encryption
Protects data confidentiality through cryptographic mechanisms and controlled keys.

Fictional example Encrypting a TLP:AMBER document does not authorise sharing it with anyone. Conversely, writing TLP:AMBER on a file does not encrypt its bytes.

Explore the method — OSINT guide: produce reliable decision intelligence

Reference : FIRST — TLP 2.0 [EN]

SBOM / VEX
SBOM
Inventory of software components for a defined product, version and scope.
VEX
Statement of vulnerability status for an identified product, to examine with its context and supporting evidence.

Fictional example A library appears in the SBOM. That alone does not demonstrate exploitability; a VEX may clarify status without replacing verification of its scope.

Explore the method — Software supply chain assurance: components to release

Reference : OpenVEX — Specification [EN]

Idempotence / Atomicity
Idempotence
Repeating the same operation preserves the intended effect of one execution under the contract’s conditions.
Atomicity
A set of steps covered by a transaction applies entirely or not at all.

Fictional example Avoiding a duplicate order on retry concerns idempotence. Committing two writes together in a local transaction concerns atomicity. Test these guarantees separately.

Explore the method — Make API integrations resilient to outages

Reference : IETF — RFC 9110, Idempotent Methods [EN]

Reference : PostgreSQL — Transactions [EN]

SLI / SLO
SLI
Observed quantitative service-level measure with a defined method, population and window.
SLO
Target value or range for a service-level indicator under explicit conditions.

Fictional example For the same scope, an SLI of 99.7% misses an SLO of 99.9%. The measurement, target and any contractual commitment are separate objects.

Explore the method — Make API integrations resilient to outages

Reference : Google SRE — Service Level Objectives [EN]

RPO / RTO
RPO
Objective describing how far back in time data must be recoverable after a disruption.
RTO
Target duration for restoring the service after a disruption within the selected scope.

Fictional example An RPO of 15 minutes and an RTO of two hours require different checks. Test recovered-data freshness and restoration duration; objectives alone prove neither.

Explore the method — Test a backup restoration from copy to working service

Reference : NIST — SP 800-34 Rev. 1 (2010) [EN]

Log / Distributed trace
Log
Record of events with context according to defined fields and collection.
Distributed trace
Representation of an operation’s path through correlated spans across components.

Fictional example Errors appear in three services’ logs. Without context propagation, they do not automatically form a complete trace; also check sampling.

Explore the method — Make security event logs ready for detection

Reference : OpenTelemetry — Signals [EN]

Mean / Percentile
Mean
Sum of observed values divided by their count; it can hide a minority of very high values.
Percentile
Position threshold in ordered values, requiring an explicit calculation method and scope.

Fictional example 99 requests at 100 ms and one at 10,100 ms have a mean of 200 ms. Nearest-rank p95 is 100 ms here; it does not describe the slowest request.

Explore the method — Diagnose Core Web Vitals in field data

Reference : Google SRE — Service Level Objectives [EN]

Correlation / Causation
Correlation
Observed association between variables, dependent on the data and scope analysed.
Causation
Relationship where intervening on a cause produces an effect under assumptions a study design must support.

Fictional example Sales rise during a campaign, but the price also falls. Association with the campaign does not isolate its effect; an appropriate causal design is needed.

Explore the method — Run an A/B experiment that supports a sound decision

Reference : NIST — Scatter Plot, Causality Is Not Proved By Association [EN]

Pseudonymisation / Anonymisation
Pseudonymisation
Replace direct identifiers to limit attribution without additional information; reidentification may remain possible.
Anonymisation
Make identification impossible in practice irreversibly, after assessing relevant means of reidentification.

Fictional example An email replaced by its hash may still be found by comparing candidate addresses. Removing a name or computing a hash does not prove anonymity.

Explore the method — Synthetic data: verify privacy and task utility

Reference : CNIL — L’anonymisation de données personnelles [FR]

Logical deletion / Media sanitization
Logical deletion
Mark an item as removed for the application without guaranteeing stored data disappears.
Media sanitization
Make access to targeted data on a medium infeasible for a defined level of effort using a suitable method.

Fictional example A record disappears from the interface but remains in a backup. That removal does not prove media sanitization; define the copies and scope checked.

Explore the method — Data lifecycle governance: retain, archive and delete with evidence

Reference : NIST — SP 800-88 Rev. 2 (2025) [EN]

Timestamp / Elapsed duration
Timestamp
An event’s position in a time reference, retained with the necessary offset or zone.
Elapsed duration
Interval measured between two instants using a clock suitable for elapsed-time measurement.

Fictional example A wall-clock correction can move timestamps backwards. Time a process using a suitable monotonic clock; keep the time useful for chronology separately.

Explore the method — Reliable reports: time zones, cutoffs and late events

Reference : Python — time.monotonic [EN]

OSINT

Research and analysis of lawfully public information to answer a defined question. A screenshot alone is insufficient: origin, date and confidence level need to remain traceable.

OSINT guide: produce reliable decision intelligence

Evidence chain

A recorded sequence of sources, copies, dates and transformations that shows how a conclusion was reached. It makes corrections possible when a source changes or disappears.

Information integrity: verify provenance, context and transformations

Exposure surface

The domains, applications, accounts, services and suppliers accessible from or dependent on the outside. An inventory should assign an owner and priority to each asset.

Cyber exposure assessment: a prioritised checklist

Third-party risk

Risk introduced by reliance on a supplier, software product or platform. Assess it through entrusted data, granted access, service criticality and exit options.

Third-party digital risk: assess a critical supplier

DNS resilience

The ability to maintain or restore resolution for essential domains after error, outage or lost access. It calls for an inventory, access protections and a recovery exercise.

Govern domains and rehearse DNS recovery

Browser extension

A browser add-on that performs a specific task. Installation should be tied to a business need, an owner and a review of site and data access.

Review browser extensions before enterprise deployment

Host permission

An extension right to interact with one or more websites. The scope of allowed addresses should match the functions actually needed. A host permission declares site access through patterns, separately from API permissions. Effective rights also depend on settings and granted permissions.

Review browser extensions before enterprise deployment

Reference : Google Chrome · Declare permissions [EN]

Optional permission

Access requested by an extension only when a feature needs it. It can reduce standing rights where the product and browser support it.

Review browser extensions before enterprise deployment

Approved extension list

An explicit set of add-ons permitted in a managed browser, with owner, purpose and reassessment conditions. It helps removal after a change or incident.

Review browser extensions before enterprise deployment

Idempotency

An operation is idempotent when repeating identical requests has the same intended effect on the server as one request. The response may vary between calls. An API may also document an idempotency key to deduplicate specific operations.

Make API integrations resilient to outages

Reference : IETF · RFC 9110, Idempotent Methods [EN]

SBOM

An inventory of software components and versions for a specific release. It becomes useful when it matches the deployed artifact and helps locate affected components after an alert.

Software supply chain assurance: components to release

RAG

A method that supplies a model with documents retrieved while answering. Quality depends on source selection, freshness and citation checks as well as the model itself.

RAG quality: evaluate a knowledge assistant with evidence

AI agent

A system that can chain decisions and call tools to reach a goal. Its scope, permissions, audit trail and human approval points should be defined before sensitive use.

AI agent governance: a production readiness guide

Hreflang

An annotation connecting equivalent language versions of a page. Each accessible version should name itself and its counterparts so search engines can show the appropriate language.

Multilingual SEO: architecture, hreflang and editorial quality

Reference : Google Search Central · Localized Versions of Your Pages [EN]

AI search visibility

The measurable presence of a page as a source or destination in AI-assisted search journeys. Useful measurement separates impressions, clicks, observed citations and conversions.

Measure visibility in AI-assisted search

Information provenance

A description of content origin, date and known transformations. It helps separate a primary document, a repost and an interpretation that still carries uncertainty.

Information integrity: verify provenance, context and transformations

Phishing-resistant MFA

Multifactor authentication using a protocol that prevents an impostor service from obtaining reusable authentication proof. WebAuthn binds the proof to the legitimate service’s domain. Account recovery also needs appropriate protection.

Deploy phishing-resistant authentication with safe recovery

Reference : NIST · SP 800-63B-4 [EN]

Cyber incident exercise

A controlled simulation that tests roles, decisions, communication and service restoration. Its value comes from the gaps corrected after the session.

Cyber incident exercise: test decisions before the crisis

Digital due diligence

A documented review of an asset, organisation or supplier before a decision. It compares claims with evidence, dependencies, risks and exit conditions.

Digital due diligence: a trust checklist before commitment

Structured data

Information added to page code to describe its content and entities explicitly. It should match visible material and cannot replace clear text or eligibility for indexing.

SEO, AEO and GEO: an SXO-led roadmap

Canonical URL

The preferred address of a page when several similar URLs exist. It helps consolidate search signals and should point to an accessible page carrying the relevant content.

Multilingual SEO: architecture, hreflang and editorial quality

Reference : Google Search Central · Canonical URLs [EN]

Web performance

The loading and responsiveness quality of a site under real conditions. Assess it across important journeys, devices and networks alongside security.

Secure, high-performance web: a journey-led audit guide

Data lifecycle

The stages of collection, use, retention, archiving and deletion. Each stage benefits from a purpose, owner, time limit and evidence of completion.

Data lifecycle governance: retain, archive and delete with evidence

AI vendor evaluation

Assessment of an AI service through permitted uses, data handling, tests, operating limits and reversibility. A sales demonstration does not prove production quality.

AI vendor evaluation: evidence, dependencies and exit conditions

Human approval point

A control where a person can inspect, change or reject an automated action before a sensitive effect. Specify what they see, their deadline and their real decision authority.

AI agent governance: a production readiness guide

RPO and RTO

RPO defines the maximum acceptable data loss after an incident; RTO defines the target time to restore service. Both need to be tied to services and tested in exercises.

Cyber incident exercise: test decisions before the crisis

Graceful degradation

Limited but intentional operation when a dependency fails. It defines retained functions, potentially delayed data and the information shown to users.

Make API integrations resilient to outages

Confidence assessment

An explicit judgement of how certain a conclusion is, based on evidence quality and convergence. Keep it separate from the importance of the decision.

OSINT guide: produce reliable decision intelligence

Primary source

A document or record created close to the event being investigated. Retain its origin and date; a repost or summary does not replace checking the source.

OSINT guide: produce reliable decision intelligence

Attack surface

The points through which a system can be reached or influenced. Inventorying assets, access and dependencies helps reduce it where it matters.

Cyber exposure assessment: a prioritised checklist

Least privilege

A principle that limits a person’s or process’s rights to what their task requires. Review those rights when the role or use changes.

AI agent governance: a production readiness guide

Audit trail

A chronological record that helps reconstruct an action, its actor, context and outcome. AI agent tool calls and approvals should remain attributable.

AI agent governance: a production readiness guide

Search intent

The need a person is trying to meet behind a query. Content should answer that need before targeting keyword variations.

SEO, AEO and GEO: an SXO-led roadmap

Supplier exit plan

Planned steps for recovering data, transferring a service and removing access when a supplier relationship ends.

Third-party digital risk: assess a critical supplier

LCP

Largest Contentful Paint. The time from navigation until the largest eligible content element visible in the viewport, such as an image or block of text, has rendered.

Diagnose Core Web Vitals in field data

Reference : web.dev · Largest Contentful Paint [EN]

Critical dependency

A service, component or supplier whose failure interrupts an important journey. Connect its criticality to degraded operation and recovery plans.

Make API integrations resilient to outages

Incident log

A timeline of facts, decisions, owners and times during an incident. It supports coordination, handover and the after-action review.

Cyber incident exercise: test decisions before the crisis

Evaluation set

Questions, documents and expected answers used to test a search or RAG system. It should represent difficult cases from real use.

RAG quality: evaluate a knowledge assistant with evidence

Prompt injection

An attempt to make an AI system follow untrusted instructions found in an input or document. Tool permissions and approval of sensitive effects limit the impact.

AI agent governance: a production readiness guide

Cryptographic hash

A value calculated from a file to check whether its bytes changed. It establishes a technical match, not the trustworthiness of the source on its own.

Information integrity: verify provenance, context and transformations

Emergency account

Access reserved for recovery when ordinary administration is unavailable. Its use should be protected, logged and tested.

Identity and account recovery: govern the complete lifecycle

x-default

An hreflang value naming a fallback page when none of the declared languages matches the user. It does not establish that two pages are translations.

Multilingual SEO: architecture, hreflang and editorial quality

Retention period

The time for which data remains useful for a defined purpose. An operational rule identifies its start, end and exceptions.

Data lifecycle governance: retain, archive and delete with evidence

AI service portability

The ability to retrieve useful data, configurations and procedures and leave a provider without losing a critical function.

AI vendor evaluation: evidence, dependencies and exit conditions

Observed citation

A verified mention of a source in an AI-assisted search response. Measure it for defined queries and dates, separately from clicks and conversions.

Measure visibility in AI-assisted search

Software artifact

A build output intended for deployment, such as a package or container image. A component inventory should match that exact artifact.

Software supply chain assurance: components to release

Domain registrar

The organisation through which a registrant registers and manages a domain name. Its access, renewals and transfer process affect digital continuity.

Govern domains and rehearse DNS recovery

Webhook

A notification between services. Handle authenticity, repetition and event order according to the provider.

Uncertain payments: recover checkout without duplicate orders

Extension manifest

A file describing an extension’s capabilities and requested permissions. Review it again after a significant change.

Review browser extensions before enterprise deployment

Restore point

The state of data to which a service can return after an incident. Check its date, coherence and actual loss against expected activity.

Test a backup restoration from copy to working service

Isolated backup copy

A copy protected from ordinary production access. Isolation reduces the chance that an outage or attack makes both the live service and its copy unusable.

Test a backup restoration from copy to working service

Functional restoration

A rehearsal that validates a rebuilt service through a real task, its data, identities and dependencies. Merely opening an archive does not establish this result.

Test a backup restoration from copy to working service

Recovery order

The sequence for restoring services and dependencies chosen by business impact. It prevents bringing a component online before resources it needs.

Test a backup restoration from copy to working service

Incident triage

The initial assessment of a signal to establish facts, scope, impact and proportionate response. It remains provisional and should be revised as evidence develops.

Triage a cyber incident and preserve early evidence

Volatile evidence

Information that may disappear quickly, such as some logs or memory states. Its collection should preserve context, time and access history.

Triage a cyber incident and preserve early evidence

Incident hypothesis

A possible explanation for an event, stated separately from confirmed observations and linked to checks that could support or reject it.

Triage a cyber incident and preserve early evidence

Incident escalation

Passing a situation to decision makers or specialists according to impact, urgency and necessary action. Record the channel, decision owner and time.

Triage a cyber incident and preserve early evidence

Representative sample

A reasoned selection of pages, components, languages and interactive states covering the variety of a product in an accessibility evaluation.

Evaluate web accessibility through journeys and evidence

Focus order

The sequence followed when moving through interactive elements with a keyboard. It should make content understandable and usable without confusing jumps.

Evaluate web accessibility through journeys and evidence

Accessible name

The label that assistive technology associates with a control, link or image. It should explain its function in the context of the journey.

Evaluate web accessibility through journeys and evidence

Accessibility retest

A renewed check after a barrier is fixed, repeating the reproduction steps and nearby journeys to identify regressions.

Evaluate web accessibility through journeys and evidence

CISA KEV

CISA’s catalog of vulnerabilities known to be exploited in the wild. A listing is a priority signal to match against the product and version actually deployed.

Prioritize vulnerabilities and verify remediation

Verified fix

An update or mitigation checked on the affected asset, with its resulting version, suitable vulnerability check and test of the working service.

Prioritize vulnerabilities and verify remediation

Patch exception

A time-bound decision to delay an update, recording the risk, compensating measure, owner and review date.

Prioritize vulnerabilities and verify remediation

Compensating control

A temporary protection that reduces exposure when direct remediation is unavailable or deferred. Its effectiveness and duration must be checked.

Prioritize vulnerabilities and verify remediation

Verifier binding

A cryptographic property linking an authentication response to the genuine service, for example its domain name, so it cannot be reused at an impostor site.

Deploy phishing-resistant authentication with safe recovery

WebAuthn

A web standard for public-key authentication bound to a particular service. Deployment also needs safe enrollment and account recovery. A web API to create and verify service-bound public-key credentials. The server checks the ceremony and required flags.

Deploy phishing-resistant authentication with safe recovery

Reference : W3C · WebAuthn Level 3 [EN]

Authenticator enrollment

The controlled binding of a new sign-in method to an account, with checks, an event record and user notification.

Deploy phishing-resistant authentication with safe recovery

Authenticator revocation

Removing a lost or compromised sign-in method and recovering the account through a verified procedure.

Deploy phishing-resistant authentication with safe recovery

Event dictionary

Versioned reference for collected event triggers, parameters, types and owners connected to a decision.

Build an event measurement plan and verify collection

Event trigger

Observable condition that starts collection and distinguishes an attempt from a confirmed business success.

Build an event measurement plan and verify collection

Event parameter

Value attached to an event to describe context, with a documented name, type and need.

Build an event measurement plan and verify collection

Measurement deduplication

A rule that counts the same action once despite reloads, retries or multiple collection channels.

Build an event measurement plan and verify collection

Randomization unit

Entity assigned consistently to an experimental variant, such as an eligible user rather than one page load. Entity assigned to a variant by the protocol, such as a user, device or organisation. Keep it stable and distinguish it from the measured event.

Run an A/B experiment that supports a sound decision

Reference : Microsoft Research · Experimentation [EN]

Primary metric

Measure chosen before an experiment to support its decision, distinct from exploratory analysis and guardrails.

Run an A/B experiment that supports a sound decision

Sample ratio mismatch

Difference between planned and observed variant allocation that should be investigated before reading effects.

Run an A/B experiment that supports a sound decision

Experimental guardrail

Potential-harm measure, such as errors or abandonment, watched even when the primary metric improves.

Run an A/B experiment that supports a sound decision

Measurement purpose

Specific reason audience data is collected and used, guiding fields, recipients and retention.

Measure audiences while respecting privacy choices

Consent exemption

A situation governed by applicable conditions for an actual configuration; a marketing label alone does not establish it.

Measure audiences while respecting privacy choices

Consent withdrawal

A changed choice that relevant collection must actually honor, including after an earlier visit.

Measure audiences while respecting privacy choices

Analytics minimization

Collecting only data needed for stated decisions and purposes, then reviewing access and retention.

Measure audiences while respecting privacy choices

Offer identifier

Stable reference linking a sold variant to its price, stock, page and distribution feeds.

Improve multichannel product catalog data quality

Product variant

The actual option selected by a customer, whose attributes, images and availability should match the offer.

Improve multichannel product catalog data quality

Price consistency

Timestamped agreement among visible price, feed and structured data for one offer and currency.

Improve multichannel product catalog data quality

Catalog source of truth

System or procedure that resolves each product attribute before it is propagated to pages and external channels.

Improve multichannel product catalog data quality

Security event

Record of an action or decision useful for reconstructing a scenario, with time, origin, outcome and necessary context.

Make security event logs ready for detection

Correlation identifier

Value connecting records from one request or operation across systems without exposing unnecessary sensitive data.

Make security event logs ready for detection

Log integrity

Ability to detect unauthorized alteration or disappearance of records needed for an investigation.

Make security event logs ready for detection

Detection rule

Tested condition turning events into a signal assigned to a person and response procedure.

Make security event logs ready for detection

Declared HTML language

The lang value of a document or passage that identifies its actual language for browsers and assistive technology.

QA multilingual content and complete user journeys

Equivalent localized page

Localized page answering the same intent as another version and eligible to be linked as a real alternative.

QA multilingual content and complete user journeys

RTL reading direction

Right-to-left organization needed for some languages and tested across text, components and interactions.

QA multilingual content and complete user journeys

Multilingual journey QA

Complete task test in one language, covering navigation, forms, errors, confirmations and local formats.

QA multilingual content and complete user journeys

URL discovery

The stage at which a search engine learns an address exists from a link, sitemap or other signal without a crawl guarantee.

Diagnose crawling and indexing of useful pages

Web crawling

A crawler request for a resource, distinct from rendering, canonical selection and indexing.

Diagnose crawling and indexing of useful pages

Noindex directive

A signal asking that a page not be indexed; the crawler must be able to read it for the instruction to be seen.

Diagnose crawling and indexing of useful pages

Orphan page

A page with no known navigable internal link, making discovery and placement in a site journey harder.

Diagnose crawling and indexing of useful pages

Form label

Persistent name associated with a field to explain what is expected, including while a value is being entered.

Design accessible forms and recoverable errors

Error summary

List of validation problems that helps locate each affected field and understand a possible correction.

Design accessible forms and recoverable errors

Recoverable error

Entry failure with a cause, useful guidance and a way to continue without losing valid values. A failure explained with a possible correction, without needlessly losing the request, valid inputs or the route to a person.

Design accessible forms and recoverable errors Accessible conversational assistants: multiple modes, recovery and human handoff

Form state

Distinct situation such as entry, waiting, refusal, outage or confirmation that should be tested in the whole journey.

Design accessible forms and recoverable errors

API contract

Description of requests, responses, schemas and behavior on which interface clients rely.

Evolve API contracts without losing clients

API problem type

Stable identifier for an error category so a client can distinguish the general cause from occurrence details.

Evolve API contracts without losing clients

Consumer compatibility

Ability of an existing client to continue its use after an interface change, verified with representative cases.

Evolve API contracts without losing clients

Idempotent processing

Handling that avoids an extra business effect when the same request or notification arrives again.

Evolve API contracts without losing clients

SPF

DNS mechanism indicating servers authorized to send for a domain; its result should be checked on real messages. DNS mechanism authorising servers to use a domain in the SMTP envelope. An SPF pass alone does not establish alignment with the From domain.

Improve email sender trust and deliverability

Reference : SPF · RFC 7208 [EN]

DKIM

Cryptographic email signature verifiable with a key published by the signing domain. Signature over message parts associated with the signer’s d= domain. A valid signature neither encrypts the message nor guarantees benign content.

Improve email sender trust and deliverability

Reference : DKIM · RFC 6376 [EN]

DMARC

Policy and check for alignment between the visible sender domain and SPF or DKIM authentication. From-domain validation through at least one authenticated and aligned SPF or DKIM result. Published failure policy informs receivers, which retain local discretion.

Improve email sender trust and deliverability

Reference : DMARC · RFC 9989 [EN]

One-click unsubscribe

An unsubscribe mechanism that an email client can request through an HTTPS POST under RFC 8058.

Email marketing: preferences and unsubscribe choices that take effect

Content owner

Person or team accountable for a page’s accuracy, update decision and review date.

Govern and maintain a useful content library

Review trigger

An observable change warranting reconsideration of a decision, such as a removed constraint, new use or different outcome.

Architecture decisions: retain the rationale and know when to reconsider

Editorial merge

Combining pages addressing the same need into a more useful resource while handling links and retired URLs.

Govern and maintain a useful content library

Content retirement

Documented decision to unpublish a page with checks on its destination, internal links and language versions.

Govern and maintain a useful content library

Community rule

Published expected behavior and possible action that make moderation understandable and actionable.

Moderate a community and escalate sensitive cases

Community triage

Classification of messages as requests, complaints, abuse or urgency to direct response and ownership.

Moderate a community and escalate sensitive cases

Social escalation

Transfer of a sensitive case to the person and channel able to decide and act within the needed time.

Moderate a community and escalate sensitive cases

Moderation review

Reassessment of a contested removal or block to confirm or correct it and improve the applied rule.

Moderate a community and escalate sensitive cases

INP

Interaction to Next Paint. This responsiveness metric reports a representative high latency among clicks, taps and key presses during a visit. Each measurement runs from the start of the interaction to the next painted frame.

Diagnose Core Web Vitals in field data

Reference : web.dev · Interaction to Next Paint [EN]

CLS

Cumulative Layout Shift. A unitless score that sums the scores of unexpected layout shifts in the worst cluster. Reserving space for media can reduce these shifts.

Diagnose Core Web Vitals in field data

Reference : web.dev · Cumulative Layout Shift [EN]

TTFB

Time to First Byte is the delay from requesting a resource to receiving its first byte. It helps separate network, server and cache waiting from later rendering.

Secure, high-performance web: a journey-led audit guide

Cache invalidation

An operation making a stored response unusable after a change, verified across cache layers.

Shared caching: test privacy, variants and invalidation

Object-level authorization

Checking that a person or service may act on the requested object even when its identifier is known. Each read or write operation must enforce the decision.

Evolve API contracts without losing clients

API version sunset

The announced date and process for retiring an old interface after consumer inventory, migration time and testing of remaining use.

Evolve API contracts without losing clients

Retry budget

A limit on attempts, time and load when a call fails. It prevents an unavailable dependency from triggering cascading requests.

Make API integrations resilient to outages

Replay window

The time during which a signed message may be accepted after checking its timestamp and identity. It reduces the chance of applying an old webhook again.

Evolve API contracts without losing clients

Retrieval recall

The share of relevant documents retrieved on an evaluated set. It depends on the set’s relevance judgments and does not by itself assess the final answer.

RAG quality: evaluate a knowledge assistant with evidence

Answer abstention

An assistant indicating that available sources do not support a reliable conclusion instead of producing an unsupported answer.

RAG quality: evaluate a knowledge assistant with evidence

Evaluation drift

A change in measured behavior after modifying the corpus, model or test method. Versioned cases help identify the cause.

RAG quality: evaluate a knowledge assistant with evidence

Tool allowlist

The limited actions an AI agent may call in a given context. Permissions and approvals are defined per action, not by prompt quality alone.

AI agent governance: a production readiness guide

Inventory freshness

Delay between a stock change at the source and its appearance on a page, in a cart or in an external feed. Measure it by channel.

Improve multichannel product catalog data quality

Variant consistency

Agreement among the selected option, image, attributes, price and availability of the actual offer, checked when the variant changes.

Improve multichannel product catalog data quality

Focus return

Restoring navigation after closure to the trigger or an available logical step.

Dialogs and keyboards: test the focus journey

Input error prevention

Design that exposes constraints before submission and helps users correct a problem without losing valid entries.

Design accessible forms and recoverable errors

Redirect map

A mapping from old to new URLs during a migration, recording destination, reason, owner and post-release verification.

Diagnose crawling and indexing of useful pages

Internal link depth

The number of navigation steps needed to reach a page from a given entry point. It helps reveal useful content that is hard to discover.

Diagnose crawling and indexing of useful pages

Hard bounce

A delivery failure indicating an address or destination is persistently unavailable. Check its cause before removal or another attempt.

Improve email sender trust and deliverability

Soft bounce

A potentially temporary delivery failure, such as throttling or unavailability. A bounded retry policy prevents endless sends.

Improve email sender trust and deliverability

GraphQL resolver

Function that supplies a GraphQL field value. Its access to objects and sensitive fields must enforce the requester’s permissions.

Secure a GraphQL API: authorization, query cost and evidence

GraphQL query cost

Estimate of server work for an operation considering fields, lists, nesting and requested volumes.

Secure a GraphQL API: authorization, query cost and evidence

GraphQL alias

Name assigned to a field result in a query. Multiple aliases can call the same field and increase executed work.

Secure a GraphQL API: authorization, query cost and evidence

Field-level authorization

Decision to return a field based on role, object and context, even when its parent object is accessible.

Secure a GraphQL API: authorization, query cost and evidence

Product group

A set of variants sharing common characteristics while each sold combination retains its own identity.

Keep product variants consistent from URL to cart

Preselected variant URL

Direct link opening a product with the intended option, image, offer and availability without relying on a prior session.

Keep product variants consistent from URL to cart

Variant identifier

Stable reference distinguishing a sold combination across catalog, page, cart and feed.

Keep product variants consistent from URL to cart

Offer discrepancy

Difference in price, stock or attributes between the displayed variant and the one actually added or bought.

Keep product variants consistent from URL to cart

Field data

Measurements collected during real visits, interpreted with their period, device, coverage and population.

Diagnose Core Web Vitals in field data

LCP element

Largest visible element used in the Largest Contentful Paint measure. Identifying it helps find the actual delay.

Diagnose Core Web Vitals in field data

Performance segment

Comparable subset of pages, devices or journeys where a metric can point to a specific fix.

Diagnose Core Web Vitals in field data

Lab data

Measurements from a controlled scenario useful for reproducing a cause but not by themselves representing all users.

Diagnose Core Web Vitals in field data

Social listening scope

Definition of sources, languages, terms, dates and limits for collecting public conversations.

Triage social listening signals before acting

Repost cluster

Set of posts derived from one publication or event, distinct from independent cases.

Triage social listening signals before acting

Signal record

Dated note linking a post to provenance, verified facts, uncertainty and a handling decision.

Triage social listening signals before acting

Confidence level

Explicit assessment of a conclusion’s strength based on available sources, context and contrary evidence.

Triage social listening signals before acting

Source independence

Two publications do not corroborate each other if both repeat the same origin. Trace their primary source and distribution chain before counting confirmations.

OSINT guide: produce reliable decision intelligence

Collection stopping rule

A pre-agreed condition for ending a search when the decision evidence is sufficient or the deadline arrives, while recording unresolved questions.

OSINT guide: produce reliable decision intelligence

Asset owner

The person or function accountable for confirming the purpose and risk of a domain, service or account and whether to keep or retire it.

Cyber exposure assessment: a prioritised checklist

Closure evidence

Verifiable evidence that a gap was actually fixed, such as a new configuration, denied-access test, service removal or dated external check.

Cyber exposure assessment: a prioritised checklist

Indirect prompt injection

An adversarial instruction in a source the agent reads, attempting to turn data into authority.

Prompt injection: test an AI agent’s trust boundaries

Human action approval

A step in which an authorized person reviews context and approves an agent action with an external or hard-to-reverse effect.

AI agent governance: a production readiness guide

Editorial equivalence

Genuine correspondence of useful information across language versions: promise, conditions, limits and action, even when words or examples differ.

Multilingual SEO: architecture, hreflang and editorial quality

Hreflang reciprocity

A relationship in which language versions mutually declare their counterparts. It should connect accessible pages that genuinely correspond.

Multilingual SEO: architecture, hreflang and editorial quality

Retention start event

An observable event that starts the retention period for a data category, such as contract closure, instead of a date chosen without context.

Data lifecycle governance: retain, archive and delete with evidence

Deletion propagation

Applying a purge decision to copies, exports and secondary systems, with a deadline, authorized exceptions and processing evidence.

Data lifecycle governance: retain, archive and delete with evidence

Tested exit clause

A portability commitment checked by actually trying an export, restore or transfer before a provider becomes critical.

AI vendor evaluation: evidence, dependencies and exit conditions

Certification scope

The products, locations, activities and period actually covered by an attestation. Its name alone does not prove that a purchased service is included.

AI vendor evaluation: evidence, dependencies and exit conditions

Provenance manifest

Assertions and signatures attached to media to trace some stages of its origin or editing. Its presence alone does not establish that the depicted scene is true.

Information integrity: verify provenance, context and transformations

Exercise inject

Information introduced at a set point in a crisis simulation to test an observable decision, communication or handoff.

Cyber incident exercise: test decisions before the crisis

Vulnerability disclosure policy

A public page defining which systems may be tested, what conduct to avoid and how to report a security flaw to its owner.

Coordinated vulnerability disclosure: receive, fix and inform

Vulnerability coordinator

An intermediary helping multiple finders, vendors and users share facts and prepare a coherent fix or advisory.

Coordinated vulnerability disclosure: receive, fix and inform

Security advisory

A dated notice identifying affected products and versions, available mitigations, user actions and known limits; update it when scope changes.

Coordinated vulnerability disclosure: receive, fix and inform

Human handoff

An explicit transfer from an automated assistant to a responsible person, with a known channel, availability and only the context needed for the request.

Accessible conversational assistants: multiple modes, recovery and human handoff

Multimodal conversational interface

A dialogue service offering more than one input or output mode, such as text and speech, without making an essential task depend on one mode.

Accessible conversational assistants: multiple modes, recovery and human handoff

Handoff summary

A summary shown to the user before transfer, covering the request, steps tried and uncertainties so the person can correct it or remove details.

Accessible conversational assistants: multiple modes, recovery and human handoff

OAuth 2.0

A framework for delegating resource access. It does not by itself establish identity or permissions on each business object.

OAuth and OpenID Connect: verify an integration end to end

OpenID Connect

An identity layer on OAuth 2.0; the client validates an identity token distinct from the access token intended for an API.

OAuth and OpenID Connect: verify an integration end to end

PKCE

A mechanism binding authorisation-code exchange to a transaction verifier. It does not replace business permissions.

OAuth and OpenID Connect: verify an integration end to end

Token audience

The recipient a token is intended for. A valid signature does not remove the need to check the audience.

OAuth and OpenID Connect: verify an integration end to end

Token issuer

The authority that issued a token. It must match the expected provider and be checked with the other properties.

OAuth and OpenID Connect: verify an integration end to end

Trust boundary

A separation between origins and permissions. Crossing it requires validation instead of trusting received text.

Prompt injection: test an AI agent’s trust boundaries

Tool effect

An actual reading, modification or sending action. Its observation may differ from the agent’s message.

Prompt injection: test an AI agent’s trust boundaries

False refusal

Rejection of a legitimate task. Measure it separately from forbidden actions that were allowed.

Prompt injection: test an AI agent’s trust boundaries

Third-party script

Code supplied by another entity and loaded in a journey. Check dependencies, data access and removal.

Third-party scripts and tags: govern access, cost and removal

SRI

Subresource Integrity: checking certain files against an expected hash. It checks bytes, not code safety.

Third-party scripts and tags: govern access, cost and removal

Loading chain

Scripts and requests initiated by an initial resource. Inventory downstream loading too.

Third-party scripts and tags: govern access, cost and removal

Data contract

Definition of fields, types, units, rules, timing and responsibilities for a data flow or metric, with change management.

Data reliability: diagnose a dashboard before making a decision

Data lineage

Traceability of origin and transformations between source and output, helping locate failures or inconsistent rules.

Data reliability: diagnose a dashboard before making a decision

Data freshness

The gap between availability and an expected event. Recent data is not necessarily correct or complete.

Data reliability: diagnose a dashboard before making a decision

Data completeness

Presence of expected items within a scope and period. It requires a known denominator and does not prove correctness.

Data reliability: diagnose a dashboard before making a decision

Uncertain payment

An operation not yet confirmed by the authoritative source. A timeout does not establish failure.

Uncertain payments: recover checkout without duplicate orders

Payment reconciliation

Comparison of provider state with orders and fulfilment to detect missing items, duplicates and discrepancies.

Uncertain payments: recover checkout without duplicate orders

Synthetic illustration

A generated visual representing an idea or fictional scene. Disclose its status; it is not a documentary photograph.

Editorial media: provenance, alternatives and useful integration

Functional image

An image serving an action or destination. Its alternative should convey that function in context.

Editorial media: provenance, alternatives and useful integration

Complex image

A chart, map or diagram conveying several relationships. A description or table should complement brief alt text.

Editorial media: provenance, alternatives and useful integration

Not assessed

No maturity conclusion has been reached. This status should not automatically become score 0.

Data reliability: diagnose a dashboard before making a decision

Not applicable

A justified exclusion within a defined scope. Rationale is still required without a score.

Data reliability: diagnose a dashboard before making a decision

Paired version test

Comparison of configurations on identical inputs with shared criteria and case-level review of losses.

Evaluate AI regressions before changing versions

Independent reference set

Cases kept out of optimisation to evaluate a version without using the test as training.

Evaluate AI regressions before changing versions

Wilson interval

An interval for a binomial proportion, requiring counts, confidence level and review of assumptions.

Small samples: communicate measurement uncertainty

Inconclusive result

A result insufficient for a decision under the protocol. It does not establish that the effect is zero.

Small samples: communicate measurement uncertainty

UTM convention

Campaign-parameter naming rules preventing fragmented values caused by case or wording.

Campaign links and UTMs: keep measurement coherent

Durable receipt

A notification retained for processing, distinct from a completed business effect.

Webhooks: handle repeats, delays and recovery

Unique business effect

An outcome protected against repeats and concurrency even when multiple notifications arrive.

Webhooks: handle repeats, delays and recovery

Portability acceptance

An export and task-reconstruction test outside a supplier with losses and adaptations recorded.

Portability: verify an exit from a digital supplier

Faceted navigation

Narrowing a catalog by attributes, requiring understandable filter state and coherent URL policies.

Search and facets: make a catalog usable

URL migration map

A justified mapping between old addresses, destinations and removals used to test cutover.

URL migration: prepare SEO and user acceptance

Account impersonation

A separate account claiming another identity, distinct from unauthorised control of the real account.

Social accounts: prepare impersonation, access loss and recovery

RAG access revocation

Withdrawal of document access checked across retrieval, chunks, citations and caches of existing conversations.

RAG: verify document access rights and isolation

Observable cohort window

Follow-up period completed for a cohort at the cutoff. An unreached horizon is not zero retention.

Retention cohorts: compare observable windows

Attributed sale

A sale associated with a channel under a rule and measurement window. Association alone does not prove causation.

Influencer campaigns: connect deliverables, evidence and measurement

Acceptance exception

A documented delivery gap with impact, owner and decision. An accepted exception remains distinct from a passing test.

Project acceptance: turn needs into criteria and evidence

Sellable inventory

Units available for sale under location, commitments and allocation rules, distinct from physical and incoming stock.

Inventory: verify availability, reservations and concurrency

HTTP revalidation

Checking with the origin before reusing a stored representation, including when required by a no-cache policy.

HTTP caching: protect personalized responses

Rich-result eligibility

Meeting requirements of a feature currently supported by a search engine. This guarantees neither display nor ranking.

Structured data: verify meaning, evidence and supported features

Descriptive transcript

Text conveying speech, useful sounds and visual information needed to understand media.

Social video: prepare captions, transcripts and alternatives

Reduced-motion preference

A system choice requesting reduced non-essential motion, checked across styles, scripts and media.

Animation: verify reduced motion, pause and understanding

Action-bound approval

Approval of specified operation parameters with scope and lifetime rather than a general intention.

AI agents: authorize an action and verify its effect

Compensating action

A separate operation addressing effects of an executed action; it does not guarantee a perfect return to the previous state.

AI agents: authorize an action and verify its effect

Table grain

The business meaning of a row, such as an order or order line, which determines appropriate keys and aggregation.

Data joins: avoid multiplying business metrics

Join fanout

Repetition of a value when one row matches multiple rows in another table, potentially distorting totals.

Data joins: avoid multiplying business metrics

Incremental conversion

An additional conversion estimated against an unexposed comparison under an explicit protocol.

Advertising: distinguish attribution from incrementality

Control contamination

Exposure of a control group to the studied treatment, potentially affecting interpretation of the comparison.

Advertising: distinguish attribution from incrementality

Fully loaded service cost

Included cost components over an explicit horizon: setup, usage, operation and exit. Scope should accompany the figure.

Digital services: compare total cost and cost per outcome

Cost per conforming outcome

Included costs divided by outcomes meeting defined business criteria, with period and denominator documented.

Digital services: compare total cost and cost per outcome

Partial refund

Return of part of a collected amount, reconciled with previous refunds and currency.

Returns and refunds: connect customer cases to payment state

Refund state

A stage confirmed by a payment system, distinct from return status and actual receipt in a bank account.

Returns and refunds: connect customer cases to payment state

SSRF

Abuse of a server feature to request an unauthorized destination; URL imports are among the surfaces to review. Server-Side Request Forgery: inducing an application to send a server request to an unauthorized destination. URL checks alone are insufficient: resolution, the actual connection and every redirect must obey egress policy.

URL imports: prevent unauthorized server requests

Reference : OWASP · SSRF Prevention Cheat Sheet [EN]

Network egress policy

Rules limiting destinations, protocols and ports reachable by a component, complementing application checks.

URL imports: prevent unauthorized server requests

Initial HTML

The document received in the HTTP response before script execution, compared with rendered content.

JavaScript SEO: check received and rendered content

Initial/rendered parity

Comparison of essential information, links and directives between the received response and executed document.

JavaScript SEO: check received and rendered content

Usable social archive

Necessary content and metadata readable off-platform with losses and retention rules documented.

Social content: prepare a usable and proportionate archive

Export loss

A field, medium, relationship or feature missing from an export, identified through a reconstruction test.

Social content: prepare a usable and proportionate archive

Uncertain submission outcome

A state in which an interface lacks reliable confirmation even though the server may have executed the request.

Network failures: preserve input and resume a task

Draft retention policy

Rules governing fields retained to resume input, their location, duration, access and deletion.

Network failures: preserve input and resume a task

Relative change

Change divided by its initial value, distinct from a difference in percentage points.

Published numbers: verify sources, calculations and interpretation

Denominator

Reference population or quantity used to calculate a ratio; exclusions should accompany the result.

Published numbers: verify sources, calculations and interpretation

Quota

Usage allowance over a period, distinct from a short-term rate or concurrency limit.

API quotas: bound load and recover without a retry storm

Concurrency limit

Maximum number of operations simultaneously in progress within a defined scope.

API quotas: bound load and recover without a retry storm

Table header

A cell naming a row or column whose association with values should be available to assistive technologies.

Data tables: preserve meaning on mobile and with screen readers

Table caption

A title or short description identifying a data table’s subject.

Data tables: preserve meaning on mobile and with screen readers

Functional unit

Amount of service to which an impact measurement is related, such as a completed transaction.

Software carbon impact: define boundaries before comparing

Measurement boundary

Components, stages and impacts included or excluded in an assessment, determining valid comparisons.

Software carbon impact: define boundaries before comparing

Preview control

A directive or attribute limiting snippets in a supporting search service; its scope depends on that service.

AI search: verify visibility and publication controls

Editorial availability

An explicit list of published, maintained content editions, rather than inferring translation from an existing URL.

AI search: verify visibility and publication controls

Structural validation

Checking result shape, including types, required fields and constraints. It does not establish factual correctness.

AI extraction: validate JSON before integration

Missing versus null value

Distinct states: an absent field supplies no value; a field containing null expresses a null value under the contract.

AI extraction: validate JSON before integration

Event time and ingestion time

When an event occurred versus when the system received it; their difference explains some report revisions.

Reliable reports: time zones, cutoffs and late events

Closed reporting period

An interval with an explicit amendment policy: frozen, revisable until a deadline or accompanied by versioned corrections.

Reliable reports: time zones, cutoffs and late events

Marketing suppression list

A limited record used to prevent re-enrollment or marketing contact after a person opts out.

Email marketing: preferences and unsubscribe choices that take effect

ADR

A record retaining context, options, an architectural decision and consequences so its rationale remains understandable.

Architecture decisions: retain the rationale and know when to reconsider

Review provenance

Information explaining where and how a review was collected and what the verification actually establishes.

Customer reviews: provenance, moderation and consistent ratings

Rating aggregate

A summary computed over a defined set of ratings with consistent counts, exclusions and item association.

Customer reviews: provenance, moderation and consistent ratings

Pagination cursor

A continuation marker for a collection; it establishes neither access authorization nor collection stability.

API pagination: complete exports without missing or duplicated records

Snapshot export

An extraction tied to a defined collection state to limit changes while reading it.

API pagination: complete exports without missing or duplicated records

Crawlable pagination

A collection split into pages reachable through links and distinct URLs, allowing items beyond the initial view to be discovered.

Paginated lists: make catalogs accessible and crawlable

Shareable list state

A URL restoring an understandable collection position or filter without depending solely on a local session.

Paginated lists: make catalogs accessible and crawlable

Social metric population

The set actually covered by an indicator, such as estimated people, impressions, plays or followers under its definition.

Social platforms: compare metrics without false equivalence

Metric comparability

Whether two measurements can be compared after checking definition, population, period and collection method.

Social platforms: compare metrics without false equivalence

Status message

Information about an action’s outcome, progress or error that may need announcement without moving focus. Information about an action’s results, waiting, progress or errors without a change of context, as defined by WCAG. Relevant messages must be presentable to assistive technology without receiving focus.

Accessible status messages: confirm actions without losing context Design accessible forms and recoverable errors

Reference : W3C · WCAG 2.2 [EN]

Live region

A region whose updates can be announced to assistive technology according to its semantics and settings.

Accessible status messages: confirm actions without losing context

Quasi-identifier

An attribute that can help identify a person when combined with other information, without being their name.

Synthetic data: verify privacy and task utility

Synthetic dataset utility

The ability of generated data to support a defined task, evaluated separately from disclosure risk.

Synthetic data: verify privacy and task utility

Weighted mean

A mean calculated with group counts rather than giving every group equal weight.

Means and percentiles: aggregate without hiding slow tasks

Aggregatable distribution

A representation that allows compatible groups to be combined before calculating a statistic.

Means and percentiles: aggregate without hiding slow tasks

Qualification criterion

An observable condition used to move a contact to a defined sales-process stage.

CRM leads: qualify with criteria and evidence

Qualification history

A record of stage changes, reasons, dates and responsible owners.

CRM leads: qualify with criteria and evidence

End of support

A date or condition after which the supplier stops the defined maintenance services for a version.

Software inventory: prepare for end of support

Retention exception

An owned, limited decision to temporarily retain a component despite a documented deviation.

Software inventory: prepare for end of support

Currency minor unit

The unit used by a payment contract to represent an amount, checked for the currency and operation.

Commerce: verify currency amounts and rounding

Rounding residual

The difference between a reference total and the sum of separately rounded amounts.

Commerce: verify currency amounts and rounding

File quarantine

A state where a received file stays isolated and unavailable for normal use until a processing decision. An isolated state for a received file while checks and the release decision remain incomplete. Normal reading stays blocked; originals, previews and derivatives must respect the same access decision.

File uploads: validate before processing or publishing

Reference : OWASP · File Upload Cheat Sheet [EN]

Declared file type

A format indication supplied with a file, distinct from its validated content.

File uploads: validate before processing or publishing

Soft 404

A page whose content resembles a missing resource or error while its response reports success.

HTTP status and soft 404: diagnose before fixing

Final HTTP status

The code returned by the destination after redirects, distinct from the initial response code.

HTTP status and soft 404: diagnose before fixing

Contextual image description

Text communicating useful information in a visual for its message and task.

Social images: write alternatives that preserve meaning

Long text alternative

A complete explanation of a complex image available beyond a short description.

Social images: write alternatives that preserve meaning

Image slot size

The width allocated to an image in the layout, which can differ from viewport width.

Responsive images: test srcset, sizes and cropping

Responsive art direction

Selection of different crops or compositions to preserve useful information in available space.

Responsive images: test srcset, sizes and cropping

Cache key

Identifier used by a cache to select a stored response; relevant request variations must be accounted for.

Shared caching: test privacy, variants and invalidation

URL mapping

A list associating each old address with a relevant destination or an explicit removal decision.

URL migration: prepare mappings and test redirects

Redirect chain

A sequence of redirect responses before the final destination, checked for unnecessary steps and loops.

URL migration: prepare mappings and test redirects

Canary release

Limited exposure of a new version to observe its effects before expansion, under defined criteria.

Progressive releases: stop criteria and rollback

Stop criterion

A predefined condition that pauses expansion or an operation and assigns a decision.

Progressive releases: stop criteria and rollback

Recognition error

A difference between visible source text and automatic transcription, distinct from interpretation error.

OCR: validate document reading before automation

Missing value

An absent observation or usable value, distinct from a measured zero or a not-applicable value.

Missing data: distinguish absence, zero and estimation

QR destination

An address or action encoded in a symbol, checked separately from the printed material’s appearance and identity.

Campaign QR codes: verify the printed material and destination

Service retirement

Organised cessation of a service, handling remaining needs, dependencies, data and entry points.

Retiring a digital service: remaining needs, data and handover

Effective cancellation date

When cancellation actually changes access or billing, distinct from the time of the request.

Subscriptions: connect cancellation, billing and access rights

Asynchronous job state

Confirmed progress of deferred processing, distinguishing receipt, execution and final outcome.

Asynchronous API jobs: track status, resume and cancel

Sitemap shard

A file grouping a coherent subset of site URLs, referenced by an index when several files are needed.

Sitemaps: verify URLs, shards and lastmod dates

Self-selection

Participation chosen by respondents, potentially yielding a group different from the population about which conclusions are intended.

Community polls: interpret who responds and who is missing

Localized input contract

Rules connecting accepted formats, locale, unit, interpretation and confirmation of entered values.

Localized input: resolve number and date ambiguity

RP ID transition

Changing a service’s WebAuthn scope, distinct from simply renaming an account.

Passkeys: prepare a domain and RP ID change

Expected signer identity

An identity authorised by a prior policy to sign a particular artifact, distinct from any technically valid signature.

Software updates: verify a signer identity change

Unreconciled event

An expected event whose identifier has no usable matching entry within a defined scope and cutoff.

Collection outages: reconcile logs and missing events

Versioned citation anchor

A citation tied to a passage and identified document version; it does not by itself establish the claim’s truth.

AI citations: check an answer when its source changes

Dangling DNS mapping

A DNS record referencing a deleted resource or one whose control is no longer assured; reclaimability depends on the provider.

Dangling subdomains: retire a CNAME without leaving a claimable dependency

Cross-window message contract

Explicit rules for origin, sending window, type, payload and state of a postMessage exchange.

postMessage: verify origin, window and requested command

Aggregate extraction budget

A shared limit on a batch’s decompression work, including its files and permitted nested archives.

ZIP and TAR archives: extract within a directory and resource budget

Actually served certificate

Certificate presented by a particular TLS endpoint during a connection; it can differ from the recently issued or stored certificate.

TLS renewal: verify the certificate actually being served

CVSS

A measure of technical vulnerability severity. It does not replace assessment of your service’s exposure and impact. Retain the version, vector and metric group with the score so the assessment remains interpretable and reproducible.

Prioritize vulnerabilities and verify remediation

Reference : FIRST · CVSS v4.0 [EN]

EPSS

Model-based estimate of vulnerability exploitation probability. Distinct from CVSS severity and local risk. The percentile expresses rank among scores, not probability. Also record the date and model version.

Prioritize vulnerabilities and verify remediation

Reference : FIRST · EPSS [EN]

TLP

Designation of sharing boundaries for security information. Check the source’s rule before redistribution. In TLP 2.0, AMBER+STRICT restricts sharing to the recipient organisation. Marking does not replace technical access control.

OSINT guide: produce reliable decision intelligence

Reference : FIRST · Traffic Light Protocol [EN]

IOC

Technical trace suggesting an attack or compromise. A match requires contextual analysis. Record type, value, source and observation period; a shared or reassigned address can produce a misleading match.

Triage a cyber incident and preserve early evidence

Reference : NIST · Indicator of Compromise [EN]

STIX 2.1

A structured language for threat-intelligence objects and relationships. A valid STIX object can still be stale or insufficiently supported. Distinguish observations in Observed Data from an Indicator’s detection patterns, then check object dates and references.

Make security event logs ready for detection

Reference : OASIS · STIX 2.1 [EN]

TAXII 2.1

A threat-intelligence exchange protocol, distinct from the STIX content model. Access to a collection does not guarantee object reliability. Check collection permissions and follow pagination; one successful response does not establish collection completeness.

Make API integrations resilient to outages

Reference : OASIS · TAXII 2.1 [EN]

SPDX

A standard for describing software components and associated information. Specify document version; a bill of materials does not establish vulnerability absence. Match component scope and identifiers against the artifact actually delivered, rather than relying on the project name alone.

Software supply chain assurance: components to release

Reference : SPDX · Specifications [EN]

CycloneDX

A bill-of-materials format covering components and dependencies. Having a CycloneDX document guarantees neither completeness nor a match with the deployed product. Inspect declared dependencies and missing information; an unlisted component is not thereby absent from the software.

Software supply chain assurance: components to release

Reference : OWASP · CycloneDX [EN]

PQC

Cryptography designed to resist known attacks using classical or quantum computers. A post-quantum algorithm still needs correct protocol integration and testing.

Software supply chain assurance: components to release

Reference : NIST · Post-quantum cryptography [EN]

KEM

Key encapsulation mechanism: establishes a shared secret using a public key and a ciphertext. It is not by itself a signature or peer authentication mechanism.

Software supply chain assurance: components to release

Reference : NIST · FIPS 203 [EN]

ML-KEM

Module-lattice-based key encapsulation specified by FIPS 203. Its three parameter sets establish a shared secret rather than sign a document.

Software supply chain assurance: components to release

Reference : NIST · FIPS 203 [EN]

ML-DSA

Module-lattice-based digital signature specified by FIPS 204. The private key signs and the public key supports signature verification.

Software supply chain assurance: components to release

Reference : NIST · FIPS 204 [EN]

SLH-DSA

Stateless hash-based digital signature specified by FIPS 205. Its SHA2 and SHAKE variants distinguish smaller signatures, labelled s, and faster signing, labelled f.

Software supply chain assurance: components to release

Reference : NIST · FIPS 205 [EN]

MTA-STS

Recipient-domain policy for compliant SMTP TLS connections and expected MX hosts. It concerns transport between supporting servers rather than end-to-end encryption.

Improve email sender trust and deliverability

Reference : MTA-STS · RFC 8461 [EN]

TLS-RPT

Reports on SMTP TLS transport successes and failures. They cover a different scope from DMARC domain-identity reports.

Improve email sender trust and deliverability

Reference : SMTP TLS Reporting · RFC 8460 [EN]

DANE

SMTP TLS transport authentication using DNSSEC-protected TLSA records. Requires DNSSEC validation and does not validate the message’s From domain.

Improve email sender trust and deliverability

Reference : DANE for SMTP · RFC 7672 [EN]

ARC

Authentication assessments carried through intermediaries. The experimental ARC protocol is not by itself a trust framework or delivery guarantee.

Improve email sender trust and deliverability

Reference : ARC · RFC 8617 [EN]

Passkey

A cryptographic credential that may be syncable or device-bound. The label alone proves neither MFA, non-exportability nor an AAL.

Deploy phishing-resistant authentication with safe recovery

Reference : W3C · WebAuthn Level 3 [EN]

AAL

NIST authentication assurance level. AAL1, AAL2 and AAL3 describe factors and protections; they are not product certifications.

Deploy phishing-resistant authentication with safe recovery

Reference : NIST SP 800-63B-4 [EN]

UP

User Presence: a WebAuthn flag for user presence. Touching a key does not prove user verification (UV).

Deploy phishing-resistant authentication with safe recovery

Reference : W3C · WebAuthn Level 3 [EN]

UV

User Verification: a WebAuthn flag for verification by the authenticator, for example using a PIN or local biometrics. The service checks it when required.

Deploy phishing-resistant authentication with safe recovery

Reference : W3C · WebAuthn Level 3 [EN]

BOLA

Broken Object Level Authorization: knowing or changing an identifier permits access to an object outside one’s rights. The server must check each access.

RAG: verify document access rights and isolation

Reference : OWASP · API1:2023 Broken Object Level Authorization [EN]

ACL

Access Control List: permissions associated with a resource, identifying which subjects may perform which operations. The service must still enforce the decision.

RAG: verify document access rights and isolation

Reference : NIST CSRC · Access Control List [EN]

RBAC

Role-Based Access Control: permissions are assigned to roles and roles to subjects. The service checks effective roles and the target resource.

RAG: verify document access rights and isolation

Reference : OWASP · Authorization Cheat Sheet [EN]

C2PA

The coalition publishing Content Credentials specifications for media provenance. Validated provenance does not establish the truth of a scene.

Information integrity: verify provenance, context and transformations

Reference : C2PA · Content Credentials Explainer 2.4 [EN]

crJSON

A JSON-LD view derived from C2PA data for purposes including reporting and interoperability. It is neither independently verifiable nor a validation input format.

Information integrity: verify provenance, context and transformations

Reference : C2PA · Technical Specification 2.4 [EN]

Content binding

Information associating digital content with an asset’s particular C2PA manifest. Inspect binding type and scope before concluding which data are covered.

Information integrity: verify provenance, context and transformations

Reference : C2PA · Content Credentials Explainer 2.4 [EN]

ETag

An opaque HTTP validator for the selected representation. A strong ETag can protect a write with If-Match; an ETag grants no access rights.

HTTP caching: protect personalized responses

Reference : IETF · RFC 9110 [EN]

If-Match

An HTTP precondition using strong ETag comparison. A stale version can produce 412. The * wildcard checks existence, not a specific version.

HTTP caching: protect personalized responses

Reference : IETF · RFC 9110 [EN]

If-None-Match

An HTTP precondition using weak ETag comparison. On a match, GET or HEAD can receive 304; another method receives 412 within the standard’s defined scope.

HTTP caching: protect personalized responses

Reference : IETF · RFC 9110 [EN]

Robots Exclusion Protocol (REP)

A protocol standardised by RFC 9309 for communicating crawl rules through robots.txt. It neither protects data access nor decides indexing.

Diagnose crawling and indexing of useful pages

Reference : IETF · RFC 9309 [EN]

X-Robots-Tag

An HTTP response header carrying indexing or preview directives for supporting search engines, including non-HTML resources. The crawler must be able to observe the response.

Measure visibility in AI-assisted search

Reference : Google Search Central · Robots meta and preview controls [EN]

data-nosnippet

A boolean HTML attribute supported by Google on span, div and section to exclude a passage from snippets. Presence applies even with the value false; it does not protect access to the text.

Measure visibility in AI-assisted search

Reference : Google Search Central · Robots meta and preview controls [EN]

Recovery initiation criteria

Local conditions applied to decide when incident recovery starts, considering facts, assumptions and risks. CSF 2.0 RS.MA-05 sets no universal deadline.

Triage a cyber incident and preserve early evidence

Reference : NIST · CSF 2.0, RS.MA-05 / RC.RP-03 / RC.RP-06 [EN]

Restoration asset integrity

Checks on backups and other assets before using them for restoration under RC.RP-03. This differs from validation of already restored assets and services under RC.RP-05.

Triage a cyber incident and preserve early evidence

Reference : NIST · CSF 2.0, RS.MA-05 / RC.RP-03 / RC.RP-06 [EN]

Recovery closure criteria

Conditions for declaring the end of recovery and completing documentation under RC.RP-06. Service return does not automatically close the investigation or all corrective actions.

Triage a cyber incident and preserve early evidence

Reference : NIST · CSF 2.0, RS.MA-05 / RC.RP-03 / RC.RP-06 [EN]

Retry-After

An HTTP response field giving a wait as non-negative integer seconds or an HTTP date. It does not by itself establish whether an operation may be repeated.

API quotas: bound load and recover without a retry storm

Reference : IETF · RFC 9110 §10.2.3 [EN]

Idempotency key

An application identifier tying replay of one operation to an API contract, context and parameters. Scope and lifetime depend on the provider; it does not guarantee permanent uniqueness.

Make API integrations resilient to outages

Reference : Stripe · Idempotent requests [EN]

STIX confidence

A STIX creator’s assessment of data correctness, expressed as an integer from 0 to 100. An absent value is unspecified, not zero, and does not measure the probability of compromise.

Make security event logs ready for detection

Reference : OASIS · STIX 2.1 · §3.2, §5.2 [EN]

STIX Sighting

A relationship object expressing a reporter’s belief that an intelligence entity was seen. It may reference Observed Data; its count is distinct from the number of data observations.

Make security event logs ready for detection

Reference : OASIS · STIX 2.1 · §3.2, §5.2 [EN]

security.txt

A file for discovering vulnerability reporting channels and disclosure practices, described by RFC 9116. On the web it lives under /.well-known/ over HTTPS; Contact and exactly one Expires are mandatory. It does not by itself grant testing permission.

Coordinated vulnerability disclosure: receive, fix and inform

Reference : IETF · RFC 9116 [EN]

VEX

An assertion of vulnerability impact on identified products, interpreted with its issuer, time and scope. OpenVEX distinguishes not_affected, affected, fixed and under_investigation. An SBOM describes components; VEX supplies declared applicability analysis. For not_affected, OpenVEX requires a justification or impact_statement; fixed identifies corrected versions rather than their actual deployment.

Software supply chain assurance: components to release

Reference : OpenVEX · Specification v0.2.0 [EN]

Build provenance

Declared information about producing a software artefact: build platform, process and inputs. Provenance needs binding to the relevant bytes and verification against roots of trust and expectations; presence alone does not establish authenticity.

Progressive releases: stop criteria and rollback

Reference : SLSA 1.2 · Build: Provenance [EN]

OAuth audience

In OAuth, the audience identifies the resource an access token is intended for. For an OIDC ID token, aud contains the recipient client’s client_id. A valid signature does not replace this check.

OAuth and OpenID Connect: verify an integration end to end

Reference : IETF · RFC 9068 [EN]

OAuth scope

An OAuth scope expresses access scope defined by the authorisation server. Its meaning depends on the resource contract; a read scope does not replace checks on the tenant, owner and requested object.

OAuth and OpenID Connect: verify an integration end to end

Reference : IETF · RFC 6749 [EN]

Webhook signature

A webhook signature authenticates message bytes under the expected key and contract. It establishes neither freshness nor uniqueness nor business success; these remain separate checks.

Webhooks: handle repeats, delays and recovery

Reference : GitHub · Validating webhook deliveries [EN]

Webhook deduplication

Webhook deduplication recognises an event already received within its provider, account and environment scope. It does not by itself prevent distinct events from producing the same effect; business-operation identity and concurrency also need controls.

Webhooks: handle repeats, delays and recovery

Reference : Stripe · Handle duplicate events [EN]

Chain of custody

Chain of custody records who held or handled an item, when, for which action and with which handover or receipt. It complements integrity checks without by itself establishing content truthfulness.

OSINT guide: produce reliable decision intelligence

Reference : NIST IR 8387 · Digital Evidence Preservation [EN]

Reference digest

A reference digest is a retained hash value for later comparison of an exact byte scope. Record its algorithm, calculation time and protection against replacement; it does not establish origin by itself.

OSINT guide: produce reliable decision intelligence

Reference : NIST · Secure Hash Standard [EN]

PDF redaction

PDF redaction removes targeted text or images from the release copy once applied and saved. Distinguish it from visual masking and from hidden-information sanitization.

OSINT guide: produce reliable decision intelligence

Reference : Adobe · Redacting and sanitizing PDFs [EN]

PDF sanitization

PDF sanitization finds and removes relevant hidden information, such as metadata or embedded content, using suitable functions. Verify its scope separately from redaction of visible text and images.

OSINT guide: produce reliable decision intelligence

Reference : Adobe · Redacting and sanitizing PDFs [EN]

DNS rebinding

A DNS answer change that can cause a different address to be reached after an initial validation. SSRF protection must enforce the validated address at the connection component and retain the policy on retries and fallback.

URL imports: prevent unauthorized server requests

Reference : OWASP · SSRF Prevention Cheat Sheet [EN]

JSON Schema

A language for describing constraints on JSON data. The dialect, required fields and validator options must be explicit. Object conformity establishes neither fidelity to a document nor permission to perform an action.

AI extraction: validate JSON before integration

Reference : JSON Schema · Object reference [EN]

Duplicate JSON keys

Repeated occurrence of the same name in a JSON object. RFC 8259 recommends unique names; parsers may handle duplicates differently. An integration profile can reject them before one value overwrites another.

AI extraction: validate JSON before integration

Reference : IETF · RFC 8259 [EN]

Ambiguous local time

A local time that corresponds to two instants when clocks move back. A named zone alone cannot select the occurrence; additional information is needed, such as a consistent offset or an explicit fold choice.

Reliable reports: time zones, cutoffs and late events

Reference : Python · PEP 495 [EN]

UTC offset

The numeric difference between local time and UTC at a given instant. An offset such as +02:00 supports conversion but does not by itself identify a geographic time zone or its clock-change rules.

Reliable reports: time zones, cutoffs and late events

Reference : IETF · RFC 3339 [EN]

Fail closed

Behavior that refuses an action when its security conditions are unmet or cannot be verified. Scanner or authorization-check failure must not become permission; an explicit recovery step may remain necessary.

File uploads: validate before processing or publishing

Reference : OWASP · Authorization Cheat Sheet [EN]

Continuation token

An API-supplied token for continuing a paginated read. Under AIP-158 it is opaque and grants no access rights; permissions still need checks on every request under the service context and contract.

API pagination: complete exports without missing or duplicated records

Reference : Google · AIP-158 Pagination [EN]

Keyset pagination

Pagination continuing after an ordered key, often a sort value plus a tie-breaking identifier. A total order avoids some ambiguities but does not alone guarantee a frozen view when keys or the collection change.

API pagination: complete exports without missing or duplicated records

Reference : PostgreSQL 18 · Row constructor comparison [EN]

KEV

CISA catalogue of vulnerabilities known to be exploited. Absence from KEV does not establish no exploitation or product safety.

Prioritize vulnerabilities and verify remediation

Reference : CISA · KEV data schema [EN]

dateAdded (KEV)

Date an entry was added to CISA KEV; distinct from the CVE identifier year, discovery or an article publication date.

Prioritize vulnerabilities and verify remediation

Reference : CISA · KEV data schema [EN]

dueDate (KEV)

Deadline published by CISA for the KEV entry’s action. It is neither a universal SLA nor evidence that remediation was installed.

Prioritize vulnerabilities and verify remediation

Reference : CISA · KEV data schema [EN]

forensicTriage (KEV)

Optional KEV field concerning forensic triage as described by CISA. Distinguish Yes, No and an absent value; No does not prove no compromise.

Prioritize vulnerabilities and verify remediation

Reference : CISA · KEV data schema [EN]

CSF Core

The NIST CSF 2.0 taxonomy of cybersecurity outcomes, organised into six functions, categories and subcategories. The Core does not prescribe one implementation method.

Digital due diligence: a trust checklist before commitment

Reference : NIST · CSWP 29 · §2–3 / Appendix C [EN]

Current Profile (CSF)

The part of a CSF organisational profile describing outcomes currently achieved or attempted, and how or to what extent they are achieved. Link it to scope and observations.

Digital due diligence: a trust checklist before commitment

Reference : NIST · CSWP 29 · §2–3 / Appendix C [EN]

Target Profile (CSF)

The part of a CSF organisational profile specifying desired outcomes selected and prioritised for risk management objectives. A target does not prove achievement.

Digital due diligence: a trust checklist before commitment

Reference : NIST · CSWP 29 · §2–3 / Appendix C [EN]

CSF Tiers

Four tiers describing the rigor of cybersecurity risk governance and management practices: Partial, Risk Informed, Repeatable, Adaptive. They are not NIST certification.

Digital due diligence: a trust checklist before commitment

Reference : NIST · CSWP 29 · §2–3 / Appendix C [EN]

DNSSEC

DNS extensions providing data-origin authentication and integrity through a chain of trust. DNSSEC does not encrypt queries or establish a website operator’s legal identity.

Govern domains and rehearse DNS recovery

Reference : IETF · RFC 4033 [EN]

DNSKEY

A DNS record carrying a public key, flags, a protocol field and an algorithm identifier. Its presence does not prove that the key is authenticated.

Govern domains and rehearse DNS recovery

Reference : IETF · RFC 4034 [EN]

DS (DNSSEC)

A delegation record carrying a Key Tag, key algorithm, digest selector and digest used to match a DNSKEY. It contains no private key.

Govern domains and rehearse DNS recovery

Reference : IETF · RFC 4034 [EN]

RRSIG

A signature record covering DNS records sharing name, class and type. It includes a covered type, algorithm and time bounds distinct from TTL.

Govern domains and rehearse DNS recovery

Reference : IETF · RFC 4034 [EN]

Certificate Transparency (CT)

Public-log ecosystem for observing and auditing the logging of TLS certificates and precertificates. A CT entry does not prove a service is live.

OSINT guide: produce reliable decision intelligence

Reference : IETF · RFC 6962 [EN]

SCT

Signed Certificate Timestamp: a log’s signed commitment to incorporate an entry within its MMD. It is not an inclusion proof by itself.

OSINT guide: produce reliable decision intelligence

Reference : IETF · RFC 6962 [EN]

STH

Signed Tree Head: a signed log-tree state including its size and root. Inclusion and consistency proofs answer different questions.

OSINT guide: produce reliable decision intelligence

Reference : IETF · RFC 6962 [EN]

MMD

Maximum Merge Delay: the maximum incorporation delay promised by a CT log after an SCT, expressed in seconds in Chrome’s list. It is not an uptime measurement.

OSINT guide: produce reliable decision intelligence

Reference : IETF · RFC 6962 [EN]

RDAP

Registration Data Access Protocol: access to registration data for domains and Internet resources. A response is a dated observation, not complete proof of a publisher’s identity.

Govern domains and rehearse DNS recovery

Reference : IETF · RFC 9083 [EN]

Registrar (RDAP)

Role of an entity representing the authority responsible for registration in the registry. Distinct from registrant, reseller and hosting provider.

Govern domains and rehearse DNS recovery

Reference : IETF · RFC 9083 [EN]

Registrant (RDAP)

Role of an entity designated as the registration holder. An absent published contact does not demonstrate that no registrant exists.

Govern domains and rehearse DNS recovery

Reference : IETF · RFC 9083 [EN]

RDAP redaction

Signalling of redacted RDAP fields under RFC 9537: removal, empty value, partial value or replacement. The signal does not reveal the original value.

Govern domains and rehearse DNS recovery

Reference : IETF · RFC 9537 [EN]

WCAG 2.2

Web Content Accessibility Guidelines version 2.2: a W3C Recommendation with Level A, AA and AAA success criteria. A criterion selection or calculation does not establish complete page or process conformance.

Design accessible forms and recoverable errors

Reference : W3C · WCAG 2.2 [EN]

Contrast ratio

The lighter relative luminance plus 0.05 divided by the darker plus 0.05. It ranges from 1:1 to 21:1; the relevant threshold depends on use and criterion.

Design accessible forms and recoverable errors

Reference : W3C · WCAG 2.2 [EN]

Relative luminance

A normalised measure of a colour’s relative brightness. For sRGB, WCAG defines channel linearisation followed by a weighted sum; simply averaging RGB numbers is unsuitable.

Design accessible forms and recoverable errors

Reference : W3C · WCAG 2.2 [EN]

activeTab

A Chrome permission granting temporary access to an invoked tab after a user gesture; navigation to a different origin or closing the tab revokes it.

Review browser extensions before enterprise deployment

Reference : Google Chrome · activeTab [EN]

Content script

Extension script running in web pages under its injection conditions. Review static declarations separately from programmatic injection.

Review browser extensions before enterprise deployment

Reference : Google Chrome · Content scripts [EN]

Native messaging

Communication between an extension and a native application registered on the endpoint. Also review native-component installation, allowed origins and removal.

Review browser extensions before enterprise deployment

Reference : Google Chrome · Native messaging [EN]

Media sanitization

A process making access to target data on media infeasible for a given effort level. Scope, method and evidence matter; file deletion alone does not establish media sanitization.

Data lifecycle governance: retain, archive and delete with evidence

Reference : NIST · SP 800-88 Rev. 2 [EN]

Cryptographic erase

A Purge technique based on sanitizing relevant cryptographic keys, subject to encryption, implementation and key-copy-management conditions. Removing an API credential does not establish CE.

Data lifecycle governance: retain, archive and delete with evidence

Reference : NIST · SP 800-88 Rev. 2 [EN]

Sanitization verification

Inspection of a sanitization technique’s immediate outcome to determine successful completion, including errors and anomalies. Distinguish it from the validation decision.

Data lifecycle governance: retain, archive and delete with evidence

Reference : NIST · SP 800-88 Rev. 2 [EN]

Sanitization validation

Acceptance or rejection of sanitization based on verification outcomes, scope and data sensitivity. A tool’s completed status alone is insufficient.

Data lifecycle governance: retain, archive and delete with evidence

Reference : NIST · SP 800-88 Rev. 2 [EN]

Syslog facility

Syslog message-family code from 0 to 23. Local families need a documented convention; they do not universally identify a product.

Make security event logs ready for detection

Reference : IETF · RFC 5424 [EN]

Syslog severity

Declared level 0–7: 0 is Emergency and 7 is Debug. This code alone does not determine business incident priority.

Make security event logs ready for detection

Reference : IETF · RFC 5424 [EN]

Syslog PRI

Integer combining facility and severity: facility × 8 + severity, spanning 0–191. PRI uses angle brackets, with no leading zero except <0>.

Make security event logs ready for detection

Reference : IETF · RFC 5424 [EN]

Syslog NILVALUE

Hyphen representing unavailable information in syslog fields that permit it. This marker does not mean zero.

Make security event logs ready for detection

Reference : IETF · RFC 5424 [EN]

Problem type URI

URI reference identifying an HTTP problem category. Title text does not replace this identity; about:blank is the default when type is absent.

Evolve API contracts without losing clients

Reference : IETF · RFC 9457 [EN]

Problem instance

URI reference identifying a particular occurrence through instance. It differs from the identity of the problem type.

Evolve API contracts without losing clients

Reference : IETF · RFC 9457 [EN]

Advisory problem status

The status member of Problem Details: an advisory restatement of the occurrence’s code. It does not replace the received HTTP response status.

Evolve API contracts without losing clients

Reference : IETF · RFC 9457 [EN]

application/problem+json

JSON media type specified for RFC 9457 Problem Details. Recognising its name validates neither the business contract nor all message properties.

Evolve API contracts without losing clients

Reference : IETF · RFC 9457 [EN]

CSP fallback

CSP directive lookup order when a more specific directive is absent. First presence wins even with an empty list; source lists are not merged.

Secure, high-performance web: a journey-led audit guide

Reference : W3C · CSP3 · Working Draft 2026-09-16 [EN]

CSP Report-Only

Header-delivered CSP policy monitoring violations without enforcing its blocking. A meta element cannot supply this mode.

Secure, high-performance web: a journey-led audit guide

Reference : W3C · CSP3 · Working Draft 2026-09-16 [EN]

CSP frame-ancestors

CSP directive restricting who can embed the resource. Distinct from frame-src; no default-src fallback and ignored in meta.

Secure, high-performance web: a journey-led audit guide

Reference : W3C · CSP3 · Working Draft 2026-09-16 [EN]

CSP form-action

CSP directive restricting form submission destinations. No default-src fallback; it does not replace server-side access checks.

Secure, high-performance web: a journey-led audit guide

Reference : W3C · CSP3 · Working Draft 2026-09-16 [EN]

GraphQL request error

GraphQL error preventing execution. The result has a non-empty errors list and no data entry.

Secure a GraphQL API: authorization, query cost and evidence

Reference : GraphQL · September 2025 [EN]

GraphQL execution error

Error at a response position during GraphQL execution. data remains present, partial or null according to Non-Null constraints.

Secure a GraphQL API: authorization, query cost and evidence

Reference : GraphQL · September 2025 [EN]

GraphQL response path

Path to a GraphQL response position using response names, aliases and zero-based list indices. It keeps the failing position after propagation.

Secure a GraphQL API: authorization, query cost and evidence

Reference : GraphQL · September 2025 [EN]

GraphQL Non-Null propagation

GraphQL error propagation through Non-Null positions to a nullable position; it may reach root data.

Secure a GraphQL API: authorization, query cost and evidence

Reference : GraphQL · September 2025 [EN]

Core Web Vitals p75

75th-percentile value for a Core Web Vitals metric in a defined scope. Distinct from a mean and a Lighthouse score.

Diagnose Core Web Vitals in field data

Reference : Google · Core Web Vitals / CrUX [EN]

CrUX origin scope

CrUX scope aggregating experiences for an origin. An origin result does not describe only the inspected URL.

Diagnose Core Web Vitals in field data

Reference : Google · Core Web Vitals / CrUX [EN]

CrUX rolling window

Rolling 28-day period used by the CrUX API. Adjacent daily windows share days and are not independent.

Diagnose Core Web Vitals in field data

Reference : Google · Core Web Vitals / CrUX [EN]

PSI missing-INP exception

PageSpeed Insights rule: when INP is unavailable, good LCP and CLS can pass the assessment. This does not demonstrate responsiveness.

Diagnose Core Web Vitals in field data

Reference : Google · Core Web Vitals / CrUX [EN]

GTIN check digit

Final GTIN digit computed with alternating 3 and 1 weights and modulo 10. Coherence proves neither GS1 allocation nor authenticity.

Keep product variants consistent from URL to cart

Reference : GS1 / Google · Product identity [EN]

GTIN zero padding

Leading zeros added to represent a shorter GTIN in 14 positions when required by a format. This creates no new GTIN-14.

Keep product variants consistent from URL to cart

Reference : GS1 / Google · Product identity [EN]

ProductGroup ID

Variant group identifier in ProductGroup. If variants repeat it using inProductGroupWithID, the values must match.

Keep product variants consistent from URL to cart

Reference : GS1 / Google · Product identity [EN]

variesBy

ProductGroup property describing variation axes. Google documents six supported Schema.org properties; describe only actual axes.

Keep product variants consistent from URL to cart

Reference : GS1 / Google · Product identity [EN]

Documentary dependency

Declared link between a record and the documents it cites. Several reposts of one record do not establish several independent confirmations.

Triage social listening signals before acting

Reference : CIE · Documentary reading workshop [EN]

Citation cycle

Loop in which records cite each other directly or indirectly. It does not by itself resolve a claim’s origin or truth.

Triage social listening signals before acting

Reference : CIE · Documentary reading workshop [EN]

Sample ratio mismatch (SRM)

Statistical discrepancy between observed unit allocation and the planned allocation across variants. A data quality signal, not a treatment effect measure.

Run an A/B experiment that supports a sound decision

Reference : Microsoft Research · Experimentation [EN]

Deprecation HTTP field

HTTP response field defined by RFC 9745. Its Structured Fields Date announces past or future deprecation; it does not by itself change resource behaviour.

Make API integrations resilient to outages

Reference : IETF · API lifecycle [EN]

Sunset HTTP field

HTTP response field defined by RFC 8594. Its HTTP-date conveys retirement intent, not an availability guarantee or proof of actual shutdown.

Make API integrations resilient to outages

Reference : IETF · API lifecycle [EN]

Precision@k

Proportion of relevant records among the first k results. Requires known judgments in that window.

RAG quality: evaluate a knowledge assistant with evidence

Reference : Stanford CS276 · Retrieval evaluation [EN]

Recall@k

Share of relevant corpus records found within k. State the corpus and zero-denominator convention.

RAG quality: evaluate a knowledge assistant with evidence

Reference : Stanford CS276 · Retrieval evaluation [EN]

RR@k

Inverse rank of the first relevant result within k; 0 if none, when necessary judgments are known.

RAG quality: evaluate a knowledge assistant with evidence

Reference : Stanford CS276 · Retrieval evaluation [EN]

nDCG@k

Discounted cumulative gain through k, normalised by the ideal ranking. Document gain, discount, missing judgments and zero ideal.

RAG quality: evaluate a knowledge assistant with evidence

Reference : Stanford CS276 · Retrieval evaluation [EN]

Cardinalité de jointure

Number of matches between rows of two sources for a condition; repeated keys can multiply pairs and measurements.

Data joins: avoid multiplying business metrics

Reference : PostgreSQL 18 · Table expressions and comparison functions [EN]

Complétion par NULL

Insertion of null values on an unmatched side of an outer join. Distinguish it from an actual record containing a null value.

Data joins: avoid multiplying business metrics

Reference : PostgreSQL 18 · Table expressions and comparison functions [EN]

IS NOT DISTINCT FROM

SQL predicate comparing values like ordinary equality while treating two nulls as not distinct; define its business meaning explicitly.

Data joins: avoid multiplying business metrics

Reference : PostgreSQL 18 · Table expressions and comparison functions [EN]

Échelle décimale d’API

Decimal count e used by an API contract to represent one major unit as 10ᵉ integer units. This count alone does not define the permitted step.

Commerce: verify currency amounts and rounding

Reference : Stripe · Special currency cases [EN]

Pas entier de montant

Required integer multiple in an amount contract. Two-decimal representation may require integers divisible by 100.

Commerce: verify currency amounts and rounding

Reference : Stripe · Special currency cases [EN]

Conversion exacte d’unité

A representation change preserving value without rounding; distinct from currency exchange and payment authorisation.

Commerce: verify currency amounts and rounding

Reference : Stripe · Special currency cases [EN]

SameSite

A cookie attribute controlling some cross-site sending. A site can include multiple origins: HTTPS subdomains sharing a registrable domain can be same-site. Does not replace CSRF protection or server-side authorisation.

Secure, high-performance web: a journey-led audit guide

Reference : MDN · Set-Cookie [EN]

CSIRT / constituency

A CSIRT serves a defined constituency: an organisation, sector or community. Its country and FIRST membership do not establish your eligibility, its availability or the services it offers. Check all three on its official website.

Triage a cyber incident and preserve early evidence

Reference : FIRST · CSIRT Services Framework 2.1 [EN]

PSIRT

A team coordinating the identification, assessment and handling of vulnerabilities in products or services supplied by its organisation. A PSIRT focuses on product security; it does not necessarily replace the CSIRT handling incidents in your own infrastructure.

Coordinated vulnerability disclosure: receive, fix and inform

Reference : FIRST · PSIRT Services Framework 1.1 [EN]