Reference
Adversarial Assurance Glossary
Each term is defined twice: once in plain language, once technically. Where the two differ in emphasis, the plain-language version is the one to use with a board.
Showing 14 of 14 terms.
- Adversarial digital twin Modelling
Plain: A living model of the organisation that an attack can be run against instead of the real thing.
Technical: A continuously updated graph of assets, identities, relationships, access, controls, trust and business consequence, updated by events rather than nightly snapshots.
- Attack path Modelling
Plain: A chain of weaknesses that together lead somewhere that matters.
Technical: An ordered sequence such as external exposure, identity weakness, privilege expansion, internal reachability, sensitive asset — reasoned over as a whole rather than as isolated findings.
- Authorization level Governance
Plain: How far the work is allowed to go, agreed before it starts.
Technical: Levels 0 to 4, from passive analysis of supplied artifacts through simulation, safe non-invasive validation and controlled authorized testing, to preauthorized reversible containment.
- Blast radius Modelling
Plain: How much a single compromise can reach.
Technical: The set of assets, data and actions reachable from a compromised component, used as an architectural budget rather than only a post-incident description.
- Choke point Modelling
Plain: One place that appears on many different attack routes.
Technical: A high-centrality node or edge whose removal disconnects a disproportionate share of paths between untrusted sources and crown jewels.
- Crown jewel Modelling
Plain: The systems and data whose loss would actually hurt.
Technical: A business-critical asset used as the terminal node in path analysis and as the anchor for consequence modelling.
- Defense independence Modelling
Plain: Whether two safeguards can fail for the same reason.
Technical: A calculated measure of shared dependency between controls; two controls resting on the same identity provider offer redundancy that may be illusory.
- Evidence ledger Evidence
Plain: A record of exactly what the conclusion rests on.
Technical: A provenance record linking finding to evidence, inference, assumptions, confidence, affected paths, recommended mitigation, expected risk reduction, and validation status.
- Fail-closed execution Governance
Plain: When permission is unclear, nothing happens.
Technical: ActionAllowed = Authorized ∧ Scoped ∧ PolicyCompliant ∧ RiskBounded. Any unmet condition denies the action; authority is never inferred.
- Irreversibility score Governance
Plain: How hard an action is to undo, which decides whether a human must approve it.
Technical: Impact weighted by irreversibility; high scores such as moving money, deleting data or modifying production require stronger approval.
- Minimal cut set Modelling
Plain: The smallest combination of failures that produces a catastrophe.
Technical: In a cyber fault tree, a minimal combination of control failures sufficient to cause the defined material breach.
- Out-of-distribution Output
Plain: This situation is too unlike anything the method has been validated against.
Technical: A release state issued when the system, data or use case materially differs from validated conditions.
- Release state Output
Plain: The verdict attached to a piece of work before anyone acts on it.
Technical: One of released, released with qualifications, insufficient evidence, out-of-distribution, or blocked.
- Security invariant Governance
Plain: A rule that must always hold, checked continuously.
Technical: A machine-evaluable graph constraint such as forbidding contractor reachability to production administration, verified against the twin on every change.