[Paper Review] Very Low Cost Entropy Source Based on Chaotic Dynamics Retrofittable on Networked Devices to Prevent RNG Attacks

Mattia Fabbri, Sergio Callegari|arXiv (Cornell University)|Dec 17, 2014

Chaos-based Image/Signal Encryption15 references4 citations

TL;DR

This paper proposes a sub-$10, retrofittable true random number generator (TRNG) based on chaotic dynamics using standard microcontroller ADCs, enabling high-entropy output (over 32 kbit/s) via feedback loop manipulation. The method exploits inherent ADC noise and chaotic behavior in a simple circuit, achieving NIST-compliant randomness with minimal post-processing, making it ideal for securing legacy networked devices.

ABSTRACT

Good quality entropy sources are indispensable in most modern cryptographic protocols. Unfortunately, many currently deployed networked devices do not include them and may be vulnerable to Random Number Generator (RNG) attacks. Since most of these systems allow firmware upgrades and have serial communication facilities, the potential for retrofitting them with secure hardware-based entropy sources exists. To this aim, very low-cost, robust, easy to deploy solutions are required. Here, a retrofittable, sub 10$ entropy source based on chaotic dynamics is illustrated, capable of a 32 kbit/s rate or more and offering multiple serial communication options including USB, I2C, SPI or USART. Operation is based on a loop built around the Analog to Digital Converter (ADC) hosted on a standard microcontroller.

Motivation & Objective

• To address the critical lack of hardware-based entropy sources in low-cost, legacy networked devices that are vulnerable to RNG attacks.
• To develop a retrofit solution compatible with existing firmware-upgradable devices, especially those without true RNGs.
• To achieve high entropy rates at minimal cost using widely available microcontroller components.
• To enhance security against side-channel and tampering attacks by leveraging chaotic dynamics rather than direct electronic noise amplification.
• To enable field deployment and integration via standard serial interfaces (USB, I2C, SPI, USART).

Proposed method

• The system uses a feedback loop involving a microcontroller's ADC and DAC, where the ADC output is processed and fed back to generate chaotic behavior.
• A key innovation is the artificial degradation of ADC resolution by masking higher bits, which stabilizes the system and prevents noise-induced instability.
• The feedback voltage $V_B$ is set to a value that ensures the output voltage $V_{\text{out}}$ remains within the ADC input range and avoids operational amplifier saturation.
• The chaotic signal is sampled at each cycle, and the resulting $m$ values are used to generate random bits via bit manipulation (e.g., $m \bmod 4$).
• The system leverages the microcontroller’s internal ADC and DAC without hardware modifications, enabling compatibility with off-the-shelf components.
• Post-processing such as Von Neumann correction or XOR of consecutive bits ensures full compliance with NIST statistical tests.

Experimental results

Research questions

• RQ1 Can chaotic dynamics in a standard microcontroller’s ADC be harnessed to generate high-entropy random bits without custom hardware?
• RQ2 How can ADC noise and non-idealities be exploited rather than mitigated to generate true randomness?
• RQ3 What level of entropy and throughput can be achieved using only off-the-shelf components and minimal post-processing?
• RQ4 Can such a system be retrofitted into existing networked devices with minimal cost and no firmware redesign?
• RQ5 How robust is the entropy source against side-channel and tampering attacks compared to noise-based alternatives?

Key findings

• The prototype achieved an entropy rate exceeding 32 kbit/s using a PIC18F2550 microcontroller.
• The system produced output with over 0.99 bits of entropy per bit, as estimated via marginal entropy analysis.
• NIST SP 800-22 statistical tests were passed after minimal post-processing, such as XORing consecutive bits or applying a Von Neumann corrector.
• The Linux entropy distiller successfully passed all NIST tests when fed with raw output from the generator.
• The method enables detection of system anomalies via high-resolution observation of the ADC output ($\hat{m}$), supporting self-diagnostic and tamper-detection capabilities.
• The cost of a self-contained unit with USB interface is under $10, and under $5 when reusing an existing microcontroller.