Résumé de séminaire

Résumé :

In a series of empirical tests we consider the quality of the pseudorandomness delivered by AES when using highly structured input patterns. Is AES able to destroy all these patterns? How does the statistical quality of the output compare to other methods used in the area of pseudorandom number generation? The talk is based on a common ACM TOMACS publication with Prof. Peter Hellekalek and will stress connections between the areas of cryptography and pseudorandom number generation.

Prerequisites: Interest and little understanding of cryptography, basic statistical testing theory (chi-square goodness-of-fit tests).

Références :

Home page : http://random.mat.sbg.ac.at/team/ste.html

Title: Tests for randomness: Return-time, Serial and Entropy-Tests

In ergodic systems, expected return times to initial states are thightly related to the relative frequency of visits in these states. This relation can be extended to statistics depending on return-times and relative frequencies and leads to two classes of tests for randomness. Both classes are finally linked by the concept of Entropy. We discuss mathematical foundations as well as practical applications within the field of pseudorandom number generation.

Prerequisites: Interest in theory and practice of statistical testing, a little knowledge in probability theory will help in the understanding of the beautiful mathematical concepts behind the construction of stringent tests for modern pseudorandom number generators - the workhorses of stochastic simulation.

This latter talk can also be given on a more basic level including the Monkey-typewriter stories if more students are interested in participation.