Functional and performance tests use ISO/IEC/IEEE 29119 design techniques chosen for the item and cover at least functional suitability, completeness, correctness and appropriateness, classification accuracy, performance efficiency (response time, throughput, scalability), resistance to the declared attack vectors (presentation, injection, deepfake, replay and hill-climb) and the privacy, security and acceptability characteristics declared. Subject to the statistical framework in Clauses 7.11 to 7.14, samples are proportionate to the level claimed: Basic at least 30 subjects and 80 percent accuracy, Effective at least 300 subjects in one demographic group and 90 percent, Highly Effective at least 300 subjects per skin tone across no fewer than three skin tones and both genders and 95 percent, Strict at least 3,000 subjects spanning several demographics and environments and 99 percent. Test populations are varied by age, by gender, by skin tone, by accessibility need and by environment unless exclusions are justified and declared. Indicators of effectiveness are determined per component and for the whole system, results demonstrate the level claimed, and where components differ the schedule states each level.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.