Traffic Light & Cyber Range Simulator

Modernized traffic-light cyber range with live 2D/3D visuals, MQTT-driven control loop, attack orchestration, and an embedded red-light camera simulator. Everything runs locally (simulation only); no physical devices are touched.

Overview

• Frontend: React + Vite + Electron shell, with Spline 3D landing/dashboard backgrounds and switchable 2D/3D intersection renders.
• Backend: Flask API + MQTT client; optional DB state logging with tunable queue sizes; simulation-only guardrails.
• Attacks: DoS/replay/MITM/timing perturbations, camera tampering presets, and live SSE/mqtt streams for dashboards.
• Profiles: Ideal/legacy envelopes that change timings, mitigations, and fallback behaviors.

Initialize & Run

Prerequisites

• Python 3.10+ (with pip)
• Node 18+ (npm)
• Mosquitto (local broker) or set MQTT_BROKER_HOST/PORT

One-time setup

# Backend
cd backend
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt

# Frontend
cd ../frontend
npm install

Dev workflow (recommended)

From the repo root, start the stack:

# MQTT broker (local install or your own)
# Mosquitto default ports: 1885 (tcp), 9001 (ws)

# Backend
cd backend
venv\Scripts\activate
python src/main.py

# Frontend (separate shell)
cd frontend
npm run dev

# Electron shell (separate shell, after frontend dev server is up)
cd electron
npx electron ..

Packaged/static workflow

cd frontend
npm run build    # builds UI + syncs static assets to backend/src/static

cd ../backend
venv\Scripts\activate
python src/main.py   # serves the built UI at http://127.0.0.1:5000

# Optional: Electron against built UI
cd ../electron
npx electron ..

Quick launcher (Windows)

run.bat will start Mosquitto (if installed), backend, frontend dev server, auto-initialize the simulator, and launch Electron.

Getting Started

1. Start the Flask backend (python backend/src/main.py) and the frontend client.
2. Use /api/traffic/* and /api/attack/* endpoints or the dashboard panels to provision profiles, launch attacks, and watch events stream live.
3. Optional tooling:
   • Export the last 200 synthetic incidents via GET /api/sim/events/export.
   • Run unit tests (after installing pytest) with python -m pytest tests/sim_camera/test_state.py.

Camera Profiles

Profile	Detection	OCR	Health	Ticket Delay (min)	Baseline Deterrence
ideal	0.94	0.92	0.95	3	0.45
mid	0.82	0.78	0.75	7	0.70
legacy	0.58	0.55	0.50	18	1.05

Switch profiles via POST /api/sim/camera/profile with body {"profile": "mid"}. Each profile defines detection probabilities, OCR confidence ranges, baseline enforcement delays, and the deterrence multiplier that feeds the red-light runner model.

Attack Presets

Presets live in backend/src/sim_camera/presets.json and are loaded at runtime. Each entry exposes the base effect payload that is scaled by intensity (low, medium, high).

ID	Default Duration (s)	Key Effects
natural_state	--	Baseline operation; no simulated impact
cam_occlusion	600	Detection -45%, OCR -40%, health -50%
lens_glare	420	Detection -25%, OCR -50%
network_outage	900	Backend offline, enforcement paused, packet loss 65%
sim_ddos	480	Detection -35%, OCR -20%, latency +350 ms
time_drift	780	Clock offset +14 s, detection/OCR -5%
replay	540	Replay flag active, detection -20%
signage_tamper	720	Deterrence x1.5, detection/OCR -10%
insider_override	600	Enforcement paused, latency +150 ms

Deploy an attack via POST /api/sim/attack/deploy:

{
  "preset_id": "network_outage",
  "duration_sec": 300,
  "intensity": "high",
  "target_intersection": "main"
}

SSE events (/api/sim/events/stream) broadcast attack_deployed, attack_ended, and event frames so the UI can show countdowns and recent incidents.

Event Schema

Every stored incident follows this shape:

{
  "event_id": "string",
  "intersection_id": "main",
  "lane_id": "north",
  "vehicle_type": "car",
  "license_plate": "ABC1234",
  "plate_confidence": 0.91,
  "timestamp_frame": "2025-11-02T12:34:56+00:00",
  "timestamp_violation": "2025-11-02T12:34:55+00:00",
  "phase": "NS_GREEN",
  "time_into_red": 1.6,
  "speed_estimated": 58.2,
  "image_front_url": null,
  "detected_by_camera": true,
  "detected_by_backend": true,
  "detection_accuracy": 0.92,
  "ticket_generated": true,
  "ticket_id": "TKT-XXX",
  "enforcement_delay_minutes": 3.5,
  "camera_health": 0.88,
  "attack_flags": ["network_outage"],
  "notes": "",
  "false_positive": false,
  "needs_review": false,
  "operator_actions": []
}

Tickets are generated only when:

• backend link is healthy,
• central detection succeeded,
• plate_confidence >= ticket_threshold (profile-specific).

Results & Exports

• Traffic Throughput vs. Attack and Violation Rate vs. Profile charts show mitigation effects.
• Export incidents for analysis via /api/sim/events/export?format=csv|json.

Experiments & Metrics

• Baseline 1,000 violations per profile; compare detection/OCR rates.
• Apply cam_occlusion or network_outage and watch detection drop vs false positives.
• Combine time_drift + replay to inspect timestamp divergence.
• Metrics block shows rolling detection, OCR, violations, tickets, mean delay, false-positive ratio.

Performance / Tuning

• Light mode (default): SIM_LIGHT_MODE=1 skips DB state logging. Enable logging with SIM_STATE_LOGGING=1 and optional SIM_DB_QUEUE_SIZE (default 5000, min 500).
• MQTT pacing: MQTT_MIN_INTERVAL_MS (default 40ms for non-attack topics), MQTT_QUEUE_LIMIT (default 400). Attack topics bypass pacing.
• Broker host/port: MQTT_BROKER_HOST, MQTT_BROKER_PORT (default 1885).
• Frontend: 3D renderer trimmed (shadows off, moderate assets) for smoother attack runs.