hardware-trng progress report

Project Title: True Random Number Generator (TRNG)

Project Goal: To design and implement a hardware-based TRNG that makes use of environmental noise and ambient light to generate high-quality random numbers suitable for various Information Security applications.

Project Status: In progress

Current Progress

Hardware Layer

The hardware layer is based on the ESP32 microcontroller, which provides ample processing power and flexibility. We have successfully integrated a noise sensor (MAX9814) and a light sensor (TSL2561) to capture acoustic noise and ambient light variations, respectively. The sensors are calibrated to provide continuous entropy input to the system, providing a steady stream of raw data.

Backend Service

The backend service is being developed using Rust, providing a strong and scalable foundation. We have implemented Websockets for entropy collection and random number distribution, allowing easy integration with various client applications. The entropy pooling and processing system is there, which includes cryptographic mixing of entropy sources to enhance randomness. We are currently working on health monitoring and quality assessment features with custom-built FSBCI (Firmware Socket Broadcast Control Interface) to make sure that the system is reliable and produces high entropy.

Client Applications

A web interface is currently being developed using TypeScript/JavaScript that allows users to request random numbers and visualize the system's status. We are also thiking about the development of a CLI tool for system integration and API clients for various programming languages for different use cases.

Algorithms and Processing

Our system uses a combination of algorithms to extract entropy from the raw sensor data, including LSB extraction, spectral entropy calculation, and zero-crossing time interval analysis. We are also implementing a Von Neumann extractor for debiasing the entropy and a SHA-256 hash function for final cryptographic hashing of the entropy. The system is designed to continuously monitor the entropy quality and perform statistical randomness tests to ensure the generated numbers' unpredictability.

We sampled the MAX9814 at 44.1 kHz with a 12-bit depth and the TSL2561 at 1 sample every 100ms with a 16-bit depth, but these values can be adjusted based on specific testing scenario. While TSL2561’s slower sampling rate captures subtle variations while minimizing unnecessary data, MAX9814’s sampling rate is limited by the ESP32's ADC conversion speed.

To illustrate how much data we can collect, let's imagine we're taking sound and light readings for 5 seconds. If we sample MAX9814 at 44.1 kHz (i.e. how many sound samples we take per second) and each sample is 2 bytes, we'll collect 441,000 bytes of sound data. TSL2561 takes 1 sample every 200ms (i.e. 0.2 seconds) and each sample is 4 bytes, so we get 25 light samples. If we combine all the data, it's about 441,100 bytes. But after we use the SHA-256 hashing algorithm, we end up with just 256 bits or 32 bytes. This shows how much the data is reduced during the process.

Data from both sensors is combined into a single array for clear separation between sources for enhanced unpredictability. The Von Neumann debiasing algorithm removes predictable patterns, and the SHA-256 hash function further hashes and secures the raw entropy.

Challenges and Next Steps

Real-time Entropy Processing

Optimizing the real-time entropy processing and cryptographic mixing algorithms to ensure minimal latency and high throughput.

Statistical Testing

Implementing a comprehensive suite of statistical randomness tests to rigorously evaluate the quality of the generated random numbers.

Security Hardening

Conducting thorough security audits and implementing additional entropy sources to further enhance the system's resilience against potential attacks.

Client Application Development

Finalizing the development of the CLI tool and API clients for various programming languages to expand the system's accessibility and usability.

TL;DR

This project is progressing well, and we are on track to deliver a functional and reliable TRNG system. We have successfully integrated the hardware components, including the ESP32 microcontroller and the noise and light sensors. The backend service, based on Rust, provides a robust foundation for entropy collection and processing. We have also made progress in developing the client applications, including a web interface for user interaction. Our system utilizes a combination of algorithms to extract entropy from the raw sensor data, and we are implementing additional measures to ensure the quality and unpredictability of the generated random numbers.