Three intertwined bodies of work developed and tested together this session.
Full test suite green: python (edge) + phpunit (cloud).
## Edge/Pi install path
- setup_online.sh: apt-lock retry (apt-daily contention), Type=exec unit +
StartLimitIntervalSec in [Unit], journalctl-on-failure, crash-loop detection.
- onnxruntime is now best-effort (no sdist / no 32-bit-ARM wheel); [6/8] probes the
real backend and hard-fails if none works.
- Hailo HEF auto-fetch: edge/tools/fetch_hailo_hef.py (pinned sha256, refuses to guess
8 vs 8L, HAILO_ARCH override for serial-less boards); hailo_pose.py run() timeout fix.
- install_offline.sh: mirrors the backend gate; onnxruntime optional.
## Cafe24 cloud package (cloud/deploy/)
- build_cafe24_package.sh emits move_detect_app/ (private, one folder) + www/move_detect/
(docroot subdir); no-dev classmap autoloader; preflight.php; hardened .htaccess.
- Subdirectory mount support: Request::baseFrom/pathFor strip the mount prefix; index.php
walks up to find the app root; views/JS prefix every URL with APP_BASE; session cookie
scoped to the mount.
## Delivery audit — silent-failure & correctness (docs/delivery-audit.md)
All 4 blockers + all Phase-1 silent-failure + Phase-2 HIGH correctness:
- Edge cloud clients send a custom User-Agent (cafe24 WAF blocks python-requests) and
allow_redirects=False.
- Daemon fails loud on blank CAFE24_URL/DEVICE_SECRET; run_loop logs swallowed exceptions;
heartbeat carries real fps + per-camera health; UvcSource reopens a dead camera;
offline queue is size-capped and drops poison rows; TARGET_FPS/QUEUE_DB validated.
- Zone coordinates normalized 0..1 on both sides (were canvas-px vs camera-px — every zone
misplaced); pinch zones emit PINCH/HIGH (were downgraded to WARNING).
- Fall detection uses bbox cues when keypoints are low-confidence (occluded fallen person);
ghost incident not cancelled by a bystander; displacement normalized per-second (FPS-invariant).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019Xmn8KZWqiQUguEMRtn6rW
ConfigController::get and HeartbeatController::post verified HMAC but never
checked nonce replay or timestamp freshness, so a captured signed request
(e.g. a replayed heartbeat) could be replayed forever, masking a dead
device. Inject NonceRepository (+Config) into both and add the same
timestamp-window + nonce-remember checks IngestController already applies,
after HMAC verification and before the response/side effect.
Only SameSite=Lax protected the acknowledge and zone save/delete endpoints,
even though a Csrf helper and mintable token already existed. Add
GET /api/csrf (CsrfController::token) to mint/return a per-session token,
check it (auth -> rbac -> csrf -> mutate) in EventDetailController::acknowledge
and ZoneConfigController::save/delete, and wire app.js/zone-draw.js to fetch
the token on load and send it as X-Csrf-Token on every POST.
Every login attempt, including unknown users (via DUMMY_HASH), forces a full
Argon2id verify with no rate limit, letting a flood take down the login
endpoint / shared host via hashing-cost amplification. Add a pre-auth
per-IP limiter (RateLimiterRepository, default 20/min) as the first check
in AuthController::login, before any user lookup or password hashing.
An unauthenticated attacker could guess a real device_id and 429-lock that
camera by hammering /api/events before HMAC verification ran, dropping its
accident events. Add a cheap pre-auth per-IP limiter (Config::ingestIpRateLimitPerMin,
default 300/min) that fires before any device-secret lookup, and move the
existing per-device limiter to run only after HMAC + timestamp + nonce checks
succeed, so only authenticated requests consume a device's own quota.
