交通大学、清华大学 统计学研究所 专题演讲 题　目：Security Enhancement on Linear Random Number Generators via Mutual Shuffling 主讲人：邓利源教授 (University of Memphis, USA) 时　间：99年12月23日(星期四)上午10:00-10:50 (上午10:50-11:10茶会於交大统计所429室举行) 地　点：交大综合一馆427室 Abstract Classical linear generators like Linear Congruential Generators (LCGs) and Multiple Recursive Generators (MRGs) have been very popular for simulation applications due to their simplicity, generating efficiency, and great empirical performance. However, linear generators, especially LCGs, are not suitable for security applications because it is fairly easy to predict (mainly because of its linearity structure) the whole sequence from just a small number of past values. To break the linearity and improve the empirical performance of LCGs, MacLaren and Marsaglia [1965] proposed using an auxiliary LCG to shuffle the output generated by a baseline LCG. In this paper, we propose a mutual-shuffling method to improve upon the original shuffle method---first construct two shuffle generators using two large-order efficient MRGs to serve as the baseline generators themselves and simultaneously as the auxiliary generator to each other; then couple the output variates of the two shuffle generators. We provide some justifications and discuss the security, efficiency, and distributional properties of the proposed generators. 敬 请 公 布 欢 迎 参 加 --

※ 发信站: 批踢踢实业坊(ptt.cc) ◆ From: 111.251.162.112