🌊 Raspberry Pi True Random Number Generator using Light and Water

A True Random Number Generator (TRNG) built using a Raspberry Pi 5 and a BH1750 light intensity sensor, capturing the chaotic behavior of bubbling water. This project collects unpredictable, real-world data from the environment to generate high-entropy randomness, usable for simulations, cryptographic experiments, or chaotic modeling.

---

📸 Project Idea

Most random number generators (like NumPy's or Python's random) are pseudo-random — deterministic with a seed. This project instead uses natural, physical processes to collect real randomness:

🔦 A light sensor (BH1750) is placed above a vessel of bubbling water (e.g., boiling or air-pumped), and the light intensity is read multiple times over time. The randomness comes from the unpredictable movement and refraction caused by bubbles, splashes, and flickering.

---

🧪 Why Is This Truly Random?

This is a hardware entropy source:

• Bubbling water creates random, chaotic changes in how light reflects and refracts.
• The BH1750 captures rapid analog light fluctuations caused by these changes.
• The setup is sensitive to:
  • Bubble size and burst timing
  • Surface ripples
  • Ambient lighting shifts
  • Steam or vibration

Since all of these are governed by thermodynamic noise, they are unpredictable and non-deterministic, satisfying the criteria for true randomness.

---

🧠 Hardware & Software Stack

🔌 Hardware

• Raspberry Pi 5
• BH1750 Light Intensity Sensor (connected via I2C)
• Water container + bubbling source (e.g., pump or heat)
• Optional: paper shade or LED to enhance lighting contrast

🖥️ Software

• Python 3
• smbus2, socket, numpy, pickle, struct
• PC/Client to request and receive random values over LAN

---

🔄 How It Works

📍 Sensor Side (Raspberry Pi)

The Pi runs a Python server that:

1. Initializes the BH1750 sensor over I2C
2. Waits for incoming TCP requests from a client
3. Collects N light readings based on the request
4. Sends the readings (as a NumPy array) using pickle over the network

📱 Client Side (Laptop/PC)

The client connects via socket to the Pi and requests N values.

It:

1. Sends an integer (number of samples needed)
2. Receives and unpacks the light readings
3. Normalizes the readings to [0, 1] to serve as usable random floats

---

🧠 Algorithm (Client)

# 1. Normalize readings
normed = (readings - readings.min()) / (readings.max() - readings.min() + 1e-9)

# 2. Use directly for rand(), or convert using scipy for randn()
from scipy.stats import norm
rand = normed
randn = norm.ppf(np.clip(normed, 1e-6, 1 - 1e-6))

• rand() → Uniform random numbers in [0, 1)
• randn() → Gaussian (normal) random numbers using inverse CDF
• randint(a, b) → Integer randoms via scaling

---

🧪 Sample Output

Function	Example Output
rand(5)	[0.45, 0.78, 0.23, 0.65, 0.91]
randint(0, 10, size=4)	[2, 9, 3, 7]
randn(3)	[-0.34, 0.83, -1.75]

These values are entirely based on live ambient randomness, not any software PRNG.

---

🧪 Testing Randomness

We verified the randomness using:

• Histograms (to check uniform or Gaussian distribution)
• Autocorrelation (should be low for true random)
• Boxplots & Entropy (distribution spread and unpredictability)

Example plots:

will be shown in future after more experimentations

---

📦 LightRandom Class (Client-Side)

The following Python class fetches and returns true random numbers:

class LightRandom:
    def __init__(self, host, port):
        self.host = host
        self.port = port

    def _request_data(self, length):
        # Connect to Pi and fetch `length` readings
        # Normalize and return as array
        ...

    def rand(self, *shape):
        ...

    def randint(self, low, high=None, size=None):
        ...

    def randn(self, *shape):
        ...

    def shuffle(self, arr):
        ...

    def choice(self, a, size=None):
        ...

📌 Every function triggers a real-time data request to the Pi and returns fresh randomness.

---

📂 Project Structure

.
├── collected_data        # Generated True Random Numbers
├── first_steup           # Setting Up The Raspberry Pi 5
├── lib                   # Main Python Script For Requesting And Receiving Random Numbers With Post Processing
├── test_notebooks        # To Test And Visualise The Generated True Random Numbers With Post Analysis And Further Implementation
├── Pi_scripts            # The Functional Python Script On RAspberry Pi 5 (Server) To Collect Data Over Bubbling Water
└── README.md             # This file

---

🚀 Future Improvements

• Add whitening functions (logistic maps, XOR-fold, entropy amplification)
• Save readings to a database for analysis
• Package into a Python module (pip install lightrandom)
• Stream entropy to multiple clients

---

🙌 Acknowledgments

• Inspired by nature's chaos, light, and physics!
• Thanks to the creators of the BH1750 sensor and Raspberry Pi ecosystem.
• Special credit to all open-source projects that power scientific creativity 🌍

---

© Copyright

This work is the intellectual property of the Nirjhar Debnath and intended for research publication. Redistribution or reproduction in any form without permission is prohibited.

---

🤖 Made with Real Randomness

This README was written while coding under an LED light — which also powered some of the randomness you're reading from ✨