An Alternative

Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Introduction
An Alternative Approach to Maurers Maurers Universal

Universal Statistical Test Statistical Test

An Example
Definition
Parameters

An Alternative

Ali DOGANAKSOY, Cihangir TEZCAN Approach

Comparison

Future Work

December 19, 2008

 An Alternative
Table of contents Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
Introduction
Introduction

Maurers Universal Statistical Test Maurers Universal

Statistical Test
An Example An Example
Definition
Definition Parameters

An Alternative
Parameters Approach

Comparison

An Alternative Approach Future Work

Comparison

Future Work
 An Alternative
Randomness in Cryptography Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Some uses of random numbers in Cryptography: Introduction

Maurers Universal
I Keystreams of one-time pads Statistical Test
An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Randomness in Cryptography Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Some uses of random numbers in Cryptography: Introduction

Maurers Universal
I Keystreams of one-time pads Statistical Test
An Example
I Secret keys of a symmetric cipher systems Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Randomness in Cryptography Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Some uses of random numbers in Cryptography: Introduction

Maurers Universal
I Keystreams of one-time pads Statistical Test
An Example
I Secret keys of a symmetric cipher systems Definition
Parameters
I Public key parameters An Alternative
Approach

Comparison

Future Work
 An Alternative
Randomness in Cryptography Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Some uses of random numbers in Cryptography: Introduction

Maurers Universal
I Keystreams of one-time pads Statistical Test
An Example
I Secret keys of a symmetric cipher systems Definition
Parameters
I Public key parameters An Alternative
Approach
I Session keys Comparison

Future Work
 An Alternative
Randomness in Cryptography Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Some uses of random numbers in Cryptography: Introduction

Maurers Universal
I Keystreams of one-time pads Statistical Test
An Example
I Secret keys of a symmetric cipher systems Definition
Parameters
I Public key parameters An Alternative
Approach
I Session keys Comparison
I Nonces Future Work
 An Alternative
Randomness in Cryptography Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Some uses of random numbers in Cryptography: Introduction

Maurers Universal
I Keystreams of one-time pads Statistical Test
An Example
I Secret keys of a symmetric cipher systems Definition
Parameters
I Public key parameters An Alternative
Approach
I Session keys Comparison
I Nonces Future Work

I Initialization vectors
 An Alternative
Randomness in Cryptography Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Some uses of random numbers in Cryptography: Introduction

Maurers Universal
I Keystreams of one-time pads Statistical Test
An Example
I Secret keys of a symmetric cipher systems Definition
Parameters
I Public key parameters An Alternative
Approach
I Session keys Comparison
I Nonces Future Work

I Initialization vectors
I Salts
 An Alternative
Frequency Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I Calculates weight of the binary sequence.
Introduction

Maurers Universal
Statistical Test
An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Frequency Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I Calculates weight of the binary sequence.
Introduction
I Probability of weight being w is: Maurers Universal
Statistical Test
  An Example
1 n Definition
P(w ) = n (1) Parameters
2 w An Alternative
Approach

Comparison

Future Work
 An Alternative
Frequency Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I Calculates weight of the binary sequence.
Introduction
I Probability of weight being w is: Maurers Universal
Statistical Test
  An Example
1 n Definition
P(w ) = n (1) Parameters
2 w An Alternative
Approach
w   Comparison
1 X n
p value = (2) Future Work
2n1 w
i=1
 An Alternative
Frequency Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I Calculates weight of the binary sequence.
Introduction
I Probability of weight being w is: Maurers Universal
Statistical Test
  An Example
1 n Definition
P(w ) = n (1) Parameters
2 w An Alternative
Approach
w   Comparison
1 X n
p value = (2) Future Work
2n1 w
i=1

I Sequence:
1111111111111111111100000000000000000000
 An Alternative
Test Suites Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Introduction

Maurers Universal
Statistical Test
I NIST An Example
Definition
I Knuth Parameters

An Alternative
I DIEHARD Approach

I TestU01 Comparison

Future Work
I Crypt-X
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Introduction

Maurers Universal
Statistical Test
I Introduced by Ueli M. Maurer in 1992. An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Introduction

Maurers Universal
Statistical Test
I Introduced by Ueli M. Maurer in 1992. An Example
Definition
Parameters
I Closely related to the per-bit entropy of a stream. An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Introduction

Maurers Universal
Statistical Test
I Introduced by Ueli M. Maurer in 1992. An Example
Definition
Parameters
I Closely related to the per-bit entropy of a stream. An Alternative
Approach
I Designed to detect statistical defects that can be
Comparison
modeled by an ergodic design with finite memory.
Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
Sequence: 01011010011101010111
Introduction

Maurers Universal
Statistical Test
An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
Sequence: 01011010011101010111
Introduction
L: 2
Maurers Universal
Statistical Test
An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
Sequence: 01011010011101010111
Introduction
L: 2
Maurers Universal
Sequence: 01 01 10 10 01 11 01 01 01 11 Statistical Test
An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
Sequence: 01011010011101010111
Introduction
L: 2
Maurers Universal
Sequence: 01 01 10 10 01 11 01 01 01 11 Statistical Test
An Example
New Sequence: 1 1 2 2 1 3 1 1 1 3 Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
Sequence: 01011010011101010111
Introduction
L: 2
Maurers Universal
Sequence: 01 01 10 10 01 11 01 01 01 11 Statistical Test
An Example
New Sequence: 1 1 2 2 1 3 1 1 1 3 Definition
Parameters
Distances: 3 6 2 1 1 4
An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
Sequence: 01011010011101010111
Introduction
L: 2
Maurers Universal
Sequence: 01 01 10 10 01 11 01 01 01 11 Statistical Test
An Example
New Sequence: 1 1 2 2 1 3 1 1 1 3 Definition
Parameters
Distances: 3 6 2 1 1 4
An Alternative
Approach
K
1 X Comparison
fn = log2 ci = 1.1949875 (3) Future Work
K
1
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
Sequence: 01011010011101010111
Introduction
L: 2
Maurers Universal
Sequence: 01 01 10 10 01 11 01 01 01 11 Statistical Test
An Example
New Sequence: 1 1 2 2 1 3 1 1 1 3 Definition
Parameters
Distances: 3 6 2 1 1 4
An Alternative
Approach
K
1 X Comparison
fn = log2 ci = 1.1949875 (3) Future Work
K
1

p value = 0.767189 (4)

 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

I Let {an } = a1 , a2 , . . . , aN be a binary sequence of Introduction

Maurers Universal
length N. Statistical Test
An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

I Let {an } = a1 , a2 , . . . , aN be a binary sequence of Introduction

Maurers Universal
length N. Statistical Test
An Example
I Partition this sequence into adjacent non-overlapping Definition
Parameters
blocks of length L. An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

I Let {an } = a1 , a2 , . . . , aN be a binary sequence of Introduction

Maurers Universal
length N. Statistical Test
An Example
I Partition this sequence into adjacent non-overlapping Definition
Parameters
blocks of length L. An Alternative
Approach
I Compute the integer values of these blocks and obtain a
Comparison
new sequence {tn } = t1 , t2 , . . . , tk where k = b NL c and
Future Work
ti {0, 1, 2, . . . , 2L 1}.
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

I Let {an } = a1 , a2 , . . . , aN be a binary sequence of Introduction

Maurers Universal
length N. Statistical Test
An Example
I Partition this sequence into adjacent non-overlapping Definition
Parameters
blocks of length L. An Alternative
Approach
I Compute the integer values of these blocks and obtain a
Comparison
new sequence {tn } = t1 , t2 , . . . , tk where k = b NL c and
Future Work
ti {0, 1, 2, . . . , 2L 1}.
I The remaining bits at the end of the sequence are
discarded.
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

I The first Q blocks of {tn } is called the initialization Introduction

part (Q = 10 2L ). Maurers Universal
Statistical Test
An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

I The first Q blocks of {tn } is called the initialization Introduction

part (Q = 10 2L ). Maurers Universal
Statistical Test
I The remaining K blocks are called the test part where An Example
Definition
K + Q = bn/Lc (K 1000 2L ). Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

I The first Q blocks of {tn } is called the initialization Introduction

part (Q = 10 2L ). Maurers Universal
Statistical Test
I The remaining K blocks are called the test part where An Example
Definition
K + Q = bn/Lc (K 1000 2L ). Parameters

An Alternative
I For every block in the test part, we calculate the Approach
distance of that block to its previous occurrence. Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

I The first Q blocks of {tn } is called the initialization Introduction

part (Q = 10 2L ). Maurers Universal
Statistical Test
I The remaining K blocks are called the test part where An Example
Definition
K + Q = bn/Lc (K 1000 2L ). Parameters

An Alternative
I For every block in the test part, we calculate the Approach
distance of that block to its previous occurrence. Comparison

I If we denote these distances by ci , the test statistic fn is Future Work

 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

I The first Q blocks of {tn } is called the initialization Introduction

part (Q = 10 2L ). Maurers Universal
Statistical Test
I The remaining K blocks are called the test part where An Example
Definition
K + Q = bn/Lc (K 1000 2L ). Parameters

An Alternative
I For every block in the test part, we calculate the Approach
distance of that block to its previous occurrence. Comparison

I If we denote these distances by ci , the test statistic fn is Future Work

K
1 X
fn = log2 ci . (5)
K
1
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali
The reference distribution for the test statistic is the
DOGANAKSOY,
Cihangir TEZCAN
half-normal distribution. The p-value is obtained as follows:
Introduction

Maurers Universal
Statistical Test
An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali
The reference distribution for the test statistic is the
DOGANAKSOY,
Cihangir TEZCAN
half-normal distribution. The p-value is obtained as follows:
Introduction
32 K 3/L
 
0.8 Maurers Universal
c = 0.7 + 4+ (6) Statistical Test
L L 15 An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali
The reference distribution for the test statistic is the
DOGANAKSOY,
Cihangir TEZCAN
half-normal distribution. The p-value is obtained as follows:
Introduction
32 K 3/L
 
0.8 Maurers Universal
c = 0.7 + 4+ (6) Statistical Test
L L 15 An Example
Definition
Parameters
r
variance(L) An Alternative
=c (7) Approach
K Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali
The reference distribution for the test statistic is the
DOGANAKSOY,
Cihangir TEZCAN
half-normal distribution. The p-value is obtained as follows:
Introduction
32 K 3/L
 
0.8 Maurers Universal
c = 0.7 + 4+ (6) Statistical Test
L L 15 An Example
Definition
Parameters
r
variance(L) An Alternative
=c (7) Approach
K Comparison
  Future Work
 fn expectedvalue(L) 
p value = erfc 
  (8)
2 
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali
The reference distribution for the test statistic is the
DOGANAKSOY,
Cihangir TEZCAN
half-normal distribution. The p-value is obtained as follows:
Introduction
32 K 3/L
 
0.8 Maurers Universal
c = 0.7 + 4+ (6) Statistical Test
L L 15 An Example
Definition
Parameters
r
variance(L) An Alternative
=c (7) Approach
K Comparison
  Future Work
 fn expectedvalue(L) 
p value = erfc 
  (8)
2 
If the obtained p-value is less than the probability of type I
error, which is a small value between 0.01 and 0.001, we
assume that the sequence is obtained from a non-random
resource.
 An Alternative
Parameters Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
Table: Parameters in NIST Statistical Test Suite
n L Q = 10 2L Introduction

Maurers Universal
387, 840 6 640 Statistical Test

904, 960 7 1, 280 An Example

Definition
Parameters
2, 068, 480 8 2, 560
An Alternative
4, 654, 080 9 5, 120 Approach

10, 342, 400 10 10, 240 Comparison

22, 753, 280 11 20, 480 Future Work

49, 643, 520 12 40, 960

107, 560, 960 13 81, 920
231, 669, 760 14 163, 840
496, 435, 200 15 327, 680
1, 059, 061, 760 16 655, 360
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Introduction

Maurers Universal
Statistical Test
Sequence: 01011010011101010111 An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Introduction

Maurers Universal
Statistical Test
Sequence: 01011010011101010111 An Example
Definition
L: 2 Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Introduction

Maurers Universal
Statistical Test
Sequence: 01011010011101010111 An Example
Definition
L: 2 Parameters

Sequence: 01 01 10 10 01 11 01 01 01 11 An Alternative
Approach

Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Introduction

Maurers Universal
Statistical Test
Sequence: 01011010011101010111 An Example
Definition
L: 2 Parameters

Sequence: 01 01 10 10 01 11 01 01 01 11 An Alternative
Approach
New Sequence: 1 1 2 2 1 3 1 1 1 3 Comparison

Future Work
 An Alternative
Maurers Universal Statistical Test Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Introduction

Maurers Universal
Statistical Test
Sequence: 01011010011101010111 An Example
Definition
L: 2 Parameters

Sequence: 01 01 10 10 01 11 01 01 01 11 An Alternative
Approach
New Sequence: 1 1 2 2 1 3 1 1 1 3 Comparison
Distances: 0 0 0 1 3 6 2 1 1 4 Future Work
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I Let {an } = a1 , a2 , . . . , aN be a binary sequence of
length N. Introduction

Maurers Universal
Statistical Test
An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I Let {an } = a1 , a2 , . . . , aN be a binary sequence of
length N. Introduction

Maurers Universal
I Partition {an } into adjacent non-overlapping blocks of Statistical Test
An Example
length L. Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I Let {an } = a1 , a2 , . . . , aN be a binary sequence of
length N. Introduction

Maurers Universal
I Partition {an } into adjacent non-overlapping blocks of Statistical Test
An Example
length L. Definition
Parameters
I Compute the integer values of these blocks. An Alternative
Approach

Comparison

Future Work
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I Let {an } = a1 , a2 , . . . , aN be a binary sequence of
length N. Introduction

Maurers Universal
I Partition {an } into adjacent non-overlapping blocks of Statistical Test
An Example
length L. Definition
Parameters
I Compute the integer values of these blocks. An Alternative
Approach
I We obtain a new sequence {tn } = t1 , t2 , . . . , tk where Comparison
k = b NL c and ti {0, 1, 2, . . . , 2L 1}. Future Work
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I Let {an } = a1 , a2 , . . . , aN be a binary sequence of
length N. Introduction

Maurers Universal
I Partition {an } into adjacent non-overlapping blocks of Statistical Test
An Example
length L. Definition
Parameters
I Compute the integer values of these blocks. An Alternative
Approach
I We obtain a new sequence {tn } = t1 , t2 , . . . , tk where Comparison
k = b NL c and ti {0, 1, 2, . . . , 2L 1}. Future Work
I Obtain another sequence {cn } where ci is the distance
between the integer ti and its previous occurrence.
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I Let {an } = a1 , a2 , . . . , aN be a binary sequence of
length N. Introduction

Maurers Universal
I Partition {an } into adjacent non-overlapping blocks of Statistical Test
An Example
length L. Definition
Parameters
I Compute the integer values of these blocks. An Alternative
Approach
I We obtain a new sequence {tn } = t1 , t2 , . . . , tk where Comparison
k = b NL c and ti {0, 1, 2, . . . , 2L 1}. Future Work
I Obtain another sequence {cn } where ci is the distance
between the integer ti and its previous occurrence.
I If the integer ti is its first occurrence in the sequence
{tn }, we assign the value 0 to ci .
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test
If ci is r for some i, this means that ti and tir are the same Ali

DOGANAKSOY,
integers and the integers between them are different than ti . Cihangir TEZCAN

Introduction

Maurers Universal
Statistical Test
An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test
If ci is r for some i, this means that ti and tir are the same Ali

DOGANAKSOY,
integers and the integers between them are different than ti . Cihangir TEZCAN

The probability of such a situation is

Introduction

(2L 1)r 1 Maurers Universal

Statistical Test
prob(ci = r ) = (9)
(2L )r An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test
If ci is r for some i, this means that ti and tir are the same Ali

DOGANAKSOY,
integers and the integers between them are different than ti . Cihangir TEZCAN

The probability of such a situation is

Introduction

(2L 1)r 1 Maurers Universal

Statistical Test
prob(ci = r ) = (9)
(2L )r An Example
Definition
Parameters

An Alternative
Approach
Note that since t1 is the first element of the sequence {tn },
Comparison
c1 must be 0. Similarly c2 is either 0 or 1. Thus a distance r
Future Work
in the sequence ci can be observed only in k r different
places.
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test
If ci is r for some i, this means that ti and tir are the same Ali

DOGANAKSOY,
integers and the integers between them are different than ti . Cihangir TEZCAN

The probability of such a situation is

Introduction

(2L 1)r 1 Maurers Universal

Statistical Test
prob(ci = r ) = (9)
(2L )r An Example
Definition
Parameters

An Alternative
Approach
Note that since t1 is the first element of the sequence {tn },
Comparison
c1 must be 0. Similarly c2 is either 0 or 1. Thus a distance r
Future Work
in the sequence ci can be observed only in k r different
places. Hence the expected number of appearance of the
value r in the sequence {cn } is

(2L 1)r 1
E (r ) = (k r ) . (10)
(2L )r
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I The number of appearance of 0 in the {cn } sequence is
equivalent to the number of distinct integers in the {tn } Introduction

sequence. Maurers Universal

Statistical Test
An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I The number of appearance of 0 in the {cn } sequence is
equivalent to the number of distinct integers in the {tn } Introduction

sequence. Maurers Universal

Statistical Test
An Example
I We will consider the cases when the probability of every Definition

possible integer values not appearing in {tn } is less than Parameters

An Alternative
104 . Approach

Comparison

Future Work
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I The number of appearance of 0 in the {cn } sequence is
equivalent to the number of distinct integers in the {tn } Introduction

sequence. Maurers Universal

Statistical Test
An Example
I We will consider the cases when the probability of every Definition

possible integer values not appearing in {tn } is less than Parameters

An Alternative
104 . Approach

Comparison
I Hence to perform the test, the largest block size L that
Future Work
can be chosen for a {tn } sequence with length k is the
largest L value satisfying the following equation:
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN
I The number of appearance of 0 in the {cn } sequence is
equivalent to the number of distinct integers in the {tn } Introduction

sequence. Maurers Universal

Statistical Test
An Example
I We will consider the cases when the probability of every Definition

possible integer values not appearing in {tn } is less than Parameters

An Alternative
104 . Approach

Comparison
I Hence to perform the test, the largest block size L that
Future Work
can be chosen for a {tn } sequence with length k is the
largest L value satisfying the following equation:
 L k  L k  k
2 1 2 2 1
1 L
L
. . . L > 0.9999 (11)
2 2 2
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Let di denote the number of appearance of the value i in the Cihangir TEZCAN

sequence {cn }. We apply the test by calculating di s and Introduction

performing 2 of goodness of fit test to di and E (i) values. Maurers Universal

The degree of freedom d is k 1 and 2 value is Statistical Test

An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Let di denote the number of appearance of the value i in the Cihangir TEZCAN

sequence {cn }. We apply the test by calculating di s and Introduction

performing 2 of goodness of fit test to di and E (i) values. Maurers Universal

The degree of freedom d is k 1 and 2 value is Statistical Test

An Example
Definition
Parameters
k
X (di E (i))2 An Alternative
2 = (12) Approach
E (i)
i=1 Comparison

Future Work
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Let di denote the number of appearance of the value i in the Cihangir TEZCAN

sequence {cn }. We apply the test by calculating di s and Introduction

performing 2 of goodness of fit test to di and E (i) values. Maurers Universal

The degree of freedom d is k 1 and 2 value is Statistical Test

An Example
Definition
Parameters
k
X (di E (i))2 An Alternative
2 = (12) Approach
E (i)
i=1 Comparison

Future Work
The p-value is:

d 2
 
p value = gammq , (13)
2 2
 An Alternative
An Alternative Approach Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Table: Largest Possible Block Sizes Cihangir TEZCAN

Sequence Length Block Size Introduction

66 n 206 2
207 n 571 3 Maurers Universal
572 n 1, 454 4 Statistical Test
1, 455 n 3, 509 5 An Example
3, 510 n 8, 224 6 Definition
8, 225 n 18, 839 7 Parameters
18, 832 n 42, 407 8 An Alternative
42, 408 n 94, 269 9 Approach
94, 270 n 207, 448 10
207, 449 n 452, 663 11 Comparison
452, 664 n 980, 823 12
980, 824 n 2, 112, 599 13 Future Work
2, 112, 600 n 4, 257, 059 14
4, 257, 060 n 9, 657, 775 15
9, 657, 776 n 20, 522, 858 16
20, 522, 859 n 43, 460, 243 17
43, 460, 244 n 91, 749, 460 18
91, 749, 461 n 193, 156, 859 19
193, 156, 860 n 405, 629, 468 20
405, 629, 469 n 849, 890, 425 21
849, 890, 426 n 1, 777, 043, 709 22
 An Alternative
Comparison Approach to
Maurers Universal
Statistical Test
1. There is no initialization part in our method which Ali
allows us to test the whole sequence without wasting
DOGANAKSOY,
Cihangir TEZCAN
any parts of the sequence.
Introduction

Maurers Universal
Statistical Test
An Example
Definition
Parameters

An Alternative
Approach

Comparison

Future Work
 An Alternative
Comparison Approach to
Maurers Universal
Statistical Test
1. There is no initialization part in our method which Ali
allows us to test the whole sequence without wasting
DOGANAKSOY,
Cihangir TEZCAN
any parts of the sequence.
Introduction
2. In Maurers Universal Test, sequences which are shorter
Maurers Universal
than 387, 840 bits cannot be tested. However, our Statistical Test
An Example
approach can be applied to test sequences as short as Definition
Parameters
66 bits.
An Alternative
Approach

Comparison

Future Work
 An Alternative
Comparison Approach to
Maurers Universal
Statistical Test
1. There is no initialization part in our method which Ali
allows us to test the whole sequence without wasting
DOGANAKSOY,
Cihangir TEZCAN
any parts of the sequence.
Introduction
2. In Maurers Universal Test, sequences which are shorter
Maurers Universal
than 387, 840 bits cannot be tested. However, our Statistical Test
An Example
approach can be applied to test sequences as short as Definition
Parameters
66 bits.
An Alternative
3. Maurers test is not suitable for block size length larger Approach

Comparison
than 16. This number is increased to 22 in our method.
Future Work
 An Alternative
Comparison Approach to
Maurers Universal
Statistical Test
1. There is no initialization part in our method which Ali
allows us to test the whole sequence without wasting
DOGANAKSOY,
Cihangir TEZCAN
any parts of the sequence.
Introduction
2. In Maurers Universal Test, sequences which are shorter
Maurers Universal
than 387, 840 bits cannot be tested. However, our Statistical Test
An Example
approach can be applied to test sequences as short as Definition
Parameters
66 bits.
An Alternative
3. Maurers test is not suitable for block size length larger Approach

Comparison
than 16. This number is increased to 22 in our method.
Future Work
4. Speed of the both algorithms:
 An Alternative
Comparison Approach to
Maurers Universal
Statistical Test
1. There is no initialization part in our method which Ali
allows us to test the whole sequence without wasting
DOGANAKSOY,
Cihangir TEZCAN
any parts of the sequence.
Introduction
2. In Maurers Universal Test, sequences which are shorter
Maurers Universal
than 387, 840 bits cannot be tested. However, our Statistical Test
An Example
approach can be applied to test sequences as short as Definition
Parameters
66 bits.
An Alternative
3. Maurers test is not suitable for block size length larger Approach

Comparison
than 16. This number is increased to 22 in our method.
Future Work
4. Speed of the both algorithms:

Table: The time comparison of the two tests based on 1,000

random sequences of length 800,000 bits
Maurers Universal Test 102 seconds
New method 87 seconds
 An Alternative
Future Work Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Introduction

Maurers Universal
Statistical Test
An Example

This new approach allows us to test short sequences and we Definition

Parameters

are using it to test candidate Hash functions of NISTs Hash An Alternative

Approach
Function Competition.
Comparison

Future Work
 An Alternative
Thank You Approach to
Maurers Universal
Statistical Test

Ali

DOGANAKSOY,
Cihangir TEZCAN

Introduction

Maurers Universal
Statistical Test
An Example
Definition

THANK YOU FOR YOUR Parameters

An Alternative
Approach

ATTENTION Comparison

Future Work