Title:

A  Family of Statistical Tests Based on Generalized Gambler's Ruin Problem

Lecturer:

Bartlomiej  Surma

Time:

15:40, May 4, 2016

Venue:

Lecturer  Profile

Bartlomiej  Surma is graduate student of Computer Security at Wroclaw University of  Technology in Poland. His current research include expanding NIST Statistical  Test Suite. He is satrting pursuing PhD degree at Max Planck Institute in  Saarbrücken, Germany in October 2016.

Lecture  Abstract

We present a new family  of statistical tests and present preliminary results of its application to  selected PRNGs. Our tests are based on recent results for generalized, multidimensional  Gambler's ruin problem:

1. The  probability of winning is not constant, but depends on current fortune.

2. 2.  The game can be multidimensional - gambler plays against many banks  independently.

We propose a method  which uses source bits (e.g. PRNGs) more efficiently.