Quality Requirement Tests¶

Quality Requirements Covered¶

QR-001: Recognition Response Time
QR-002: Resistance to Static Photo Spoofing
QR-003: Recognition Model Modularity
QR-004: Temporary Access Window Enforcement

QRT-001: Recognition Pipeline Response Time¶

Linked quality requirement: QR-001

Verification method: Automated performance CI check.

Test data, setup, or environment: Standard CI/Docker test environment. The test uses a controlled fake frame, fake face app, and fake face result instead of live hardware.

Automated command or CI check: pytest tests/quality/test_recognition_performance.py

Expected measurable result: The test verifies two levels of the pipeline: 1. Embedding extraction from a detected face completes in <= 3.0 seconds. 2. The full access-decision pipeline (capture -> extract -> compare against the database -> decision), as described in the QR-001 scenario, also completes in <= 3.0 seconds and produces a valid access decision.

The test fails if either step exceeds 3.0 seconds, crashes, or fails to produce a decision.

Evidence link: Latest protected default-branch CI run after CI integration.

Limitation: Uses controlled fake inputs instead of a live USB camera; the fake face-detection app returns instantly, so this measures pipeline/business-logic overhead rather than actual ML model inference latency.

CI Job: pytest tests/quality/ runs automatically on every PR and push to main

Status: Implemented

Known production gap: See QR-001 "Current status", end-to-end response time exceeds 3.0s on real hardware (Raspberry Pi 5). Tracked in #171.

QRT-002: Static Photo Spoofing Rejection¶

Linked quality requirement: QR-002

Verification method: Automated security integration CI check.

Test data, setup, or environment: Standard CI/Docker test environment. The test uses mock recognizers and controlled embedding vectors to simulate spoof, no-face, and low-similarity attempts.

Automated command or CI check: pytest tests/quality/test_photo_spoofing_rejection.py

Expected measurable result: The test passes only if the validation pipeline correctly rejects spoof or invalid recognition attempts, including at least 9 out of 10 low-similarity attempts. It fails if fewer than 9 attempts are rejected or if access is erroneously granted.

Evidence link: Latest protected default-branch CI run after CI integration.

Limitation: Uses controlled mock recognizers and embeddings instead of physical printed-photo attacks with a live camera.

CI Job: pytest tests/quality/ runs automatically on every PR and push to main

Status: Implemented

QRT-003: Inference Provider Contract Modularity¶

Linked quality requirement: QR-003

Verification method: Automated architecture integration CI check.

Test data, setup, or environment: Standard CI/Docker test environment. A fake recognition provider is injected into the standard application flow instead of the real ML library.

Automated command or CI check: pytest tests/quality/test_inference_provider_contract.py

Expected measurable result: The test passes if the application successfully receives a valid access decision from the mock provider without raising dependency errors, proving the pipeline does not directly depend on a specific recognition library.

Evidence link: Latest protected default-branch CI run after CI integration.

Limitation: Uses a mock provider instead of executing the real InsightFace model, ONNX liveness detector, camera input, or GPIO hardware.

CI Job: pytest tests/quality/ runs automatically on every PR and push to main

Status: Implemented

QRT-004: Temporary Access Window Enforcement¶

Linked quality requirement: QR-004

Verification method: Automated unit test on the pure business-rule functions, plus a mocked-DB test on the data-access layer.

Test data, setup, or environment: Standard CI test environment. No real database is used — connect_to_db() is mocked; test fixtures provide rows representing permanent, active-temporary, expired-temporary, and not-yet-started-temporary employees with a fixed now timestamp for determinism. Legacy date-typed (pre-migration) values are also covered.

Automated command or CI check: pytest tests/unit/test_temporary_access.py

Expected measurable result: _temporary_access_is_active() returns True only when the current time is within [start_date, expiration_date] inclusive (or for non-Temporary status), correctly handling both datetime and legacy date inputs. _normalize_access_window() converts date inputs to full-day datetime bounds before writes. get_all_embeddings() excludes any temporary employee outside the active window from the results used for recognition matching. add_employees() returns False and skips the INSERT when a duplicate name exists.

Evidence link: Latest protected default-branch CI run.

Limitation: Uses a mocked cursor rather than a real PostgreSQL instance, so it verifies the Python-level filtering and normalization logic but not PostgreSQL-specific behavior (e.g. the DATE→TIMESTAMP migration SQL is checked for the expected substrings, not executed against a real database).

CI Job: pytest tests/ runs automatically on every PR and push to main

Status: Implemented