Embedded HW/SW Platform for On-the-Fly Testing of True Random Number Generators

Abstract

Abstract-We present a HW/SW platform for on-the-fly detection of failures and weaknesses in entropy sources. By splitting the operations between hardware and software, we achieve sufficient flexibility to control the level of significance of the tests. This approach also enables sharing resources between different tests thereby reducing the area and power. Statistical tests were selected from the NIST test suite. We propose several versions of hardware co-processors for monitoring random bit sequences, ranging from 52 slices (5 tests) to 552 slices (9 tests) on Spartan-6 FPGA. We are the first to provide implementations of the Serial test and the Approximate entropy test for on-the-fly monitoring.

TOTAL: TRNG On-the-fly Testing for Attack Detection using Lightweight Hardware

“…The features were selected based on the previous implementations of on-the-fly tests. We looked into the implementations presented in [2], [3], [4], [5], [6], [7], [8], [9] and selected the features that can be implemented in a simple and compact manner. In addition, we recommended using the auto-correlation coefficients.…”

“…Related work can be divided into two groups. The first group consists of implementations of standard statistical tests [2], [3], [4], [5], [6], [7], [8], [9]. These tests are originally designed for prototype evaluation, so some of them are too computationally intensive to be practical for on-the-fly testing.…”

Exploring active manipulation attacks on the TERO random number generator

“…Several successful attacks [1], [2], [3] that manipulate the TRNG output have been presented in the literature in recent years. Robustness against such attacks can be achieved by implementing online tests [4], [5], [6] that monitor the behavior of the entropy source and trigger an alarm signal in case a malfunction is detected.…”