BLE Crowd Radar

Real-time Bluetooth Low Energy crowd telemetry, proximity estimation, and edge Machine Learning.

✔ 100% Privacy by Design ⚡ Edge ONNX Neural Engine 🔒 Zero Cloud Telemetry

🛡 Zero-Data Transmission Pledge

Your privacy is absolute. Crowd Density Radar operates 100% on your device. We do not operate remote servers, tracking databases, analytics SDKs, or cloud backends.

No Data Ever Leaves Your Device

All Bluetooth Low Energy advertisement packet analysis, RSSI distance calculations, and ONNX crowd classification are computed locally in device RAM.

🔍 What We Sniff (And What We Don't)

✅ What Is Processed Locally:

• Ambient BLE Advertisement Packet Frequency
• Received Signal Strength Indication (RSSI in dBm)
• Anonymous Manufacturer IDs (e.g., Apple, Google, Tile)
• Ephemeral timestamps for rolling scan windows

❌ What Is NEVER Accessed:

• Personal Names, Phone Numbers, or Emails
• Media, Messages, or Contact Lists
• Bluetooth Pairing, Bonding, or Data Streams
• Device Storage, Browsing History, or Location Logs

♻ Ephemeral Memory & MAC Address Rotation

Modern smartphones and smartwatches employ Resolvable Private Addresses (RPA) that rotate their MAC address every 15 minutes. Crowd Density Radar embraces this privacy feature:

• Observations are stored strictly in volatile device memory (RAM) during active tracking.
• Stopping the scan or clearing the history instantly purges all recorded observations.
• No persistent tracking profiles or historical user identifiers are ever generated or stored.

🧠 Embedded ONNX Neural Model

Crowd density is evaluated using an on-device ONNX machine learning model (CrowdDensityModel.onnx). It evaluates multi-variable radar features including total active transmitters, detection frequency per scan, and RSSI distribution to classify crowd levels from Very Low to Very High.

📏 Log-Distance Path Loss & Kalman Filter

Distance is estimated using the radio frequency path-loss formula calibrated to environmental RF dynamics:

Distance = 10 ^ ((A - RSSI) / (10 * n))

Where A is the 1-meter reference power (dBm) and n is the path-loss exponent (configurable via Environment Presets: Open Air, Indoor Office, Dense Urban, or Stadium).

A real-time 1D Kalman Filter filters out multipath RF fading, signal reflections, and human body absorption spikes for stable, smooth distance telemetry.

📍 Proximity Stratification Zones

Immediate Zone (< 3m)

High-density proximity cluster. Strongest signal (≥ -60 dBm).

Proximity Zone (3m – 8m)

Active room / medium distance cluster (-60 dBm to -80 dBm).

Distant Zone (8m – 15m)

Perimeter devices on the edge of detection (< -80 dBm).

Far Zone (15m+)

Ambient peripheral Bluetooth beacons and background trackers.

🔑 Operating System Permissions

To perform local passive Bluetooth packet inspection and render the proximity radar, the operating system requires the following runtime permissions:

BLUETOOTH_SCAN

Bluetooth Radio Scanning

Allows the app to passively detect surrounding BLE advertisement packets and read signal power (RSSI).

ACCESS_FINE_LOCATION

Location Services (Android OS Requirement)

Android requires location permission to perform BLE discovery (as beacons can theoretically determine physical proximity) and to position your central scanning station on the OpenStreetMap radar overlay.

HARDWARE_ACCELERATION

Graphics & Canvas Rendering

Used to render 60 FPS real-time radar sweeps and historical telemetry sparklines directly via the GPU.

⚙ User Control & Revocation

You may revoke any permission at any time via your device's Settings → Apps → Crowd Density → Permissions. Revoking Bluetooth will prevent live scanning, but historical metrics and preset calculators remain fully functional.