PAI-SF / 1 / PAI-SF-SEN-003

Adversarial Physical-World Input Detection

Objective

Detect deliberate adversarial patterns designed to manipulate perception, distinct from passive degradation.

Control / requirement

Auxiliary out-of-distribution/adversarial-pattern detection triggering reduced confidence or alert on detection.

Applicability

Systems operating in environments where adversarial physical-world attack is a credible threat per the threat catalog.

Expected evidence

Adversarial test suite results; detection rate benchmarks. [T3]

Assurance expectation

Test evidence against a defined, current set of known adversarial techniques, refreshed as the threat catalog updates.

Dependencies

Section 5 threat catalog (adversarial physical-world inputs entry).

Exclusions

Not applicable to fully enclosed/access-controlled operating environments where adversarial physical access is not credible.

Maturity / conformance relevance

Expected at High-Assurance level; Operational level may address only known, published attack patterns.

Ecosystem relationship

Provides adversarial-perception incident categories for UAIF™; informs AI-IRF™ investigation of suspected deliberate manipulation.

Domain

2. Actuator and Kinetic Command Safety

Download full Control Catalogue ↓