Machine Learning Predictors for Min-Entropy Estimation

TL;DR

This paper explores machine learning methods, including deep learning models, for estimating min-entropy in RNGs, showing they can outperform traditional predictors and emphasizing the importance of target bit count in entropy assessment.

Contribution

It introduces machine learning predictors for min-entropy estimation, compares their performance with traditional methods, and highlights the impact of target bit count on entropy evaluation.

Findings

ML predictors outperform traditional methods in certain scenarios

Deep learning models effectively estimate average min-entropy

Target bit count significantly influences entropy assessment

Abstract

This study investigates the application of machine learning predictors for min-entropy estimation in Random Number Generators (RNGs), a key component in cryptographic applications where accurate entropy assessment is essential for cybersecurity. Our research indicates that these predictors, and indeed any predictor that leverages sequence correlations, primarily estimate average min-entropy, a metric not extensively studied in this context. We explore the relationship between average min-entropy and the traditional min-entropy, focusing on their dependence on the number of target bits being predicted. Utilizing data from Generalized Binary Autoregressive Models, a subset of Markov processes, we demonstrate that machine learning models (including a hybrid of convolutional and recurrent Long Short-Term Memory layers and the transformer-based GPT-2 model) outperform traditional NIST SP 800-90B predictors in certain scenarios. Our findings underscore the importance of considering the number of target bits in min-entropy assessment for RNGs and highlight the potential of machine learning approaches in enhancing entropy estimation techniques for improved cryptographic security.