Cryptography Reference
In-Depth Information
Claim 3.7.8.1: For every m
1, conditioned on
¬ ζ m , the R i 's are uniformly and
independently distributed over
n , and the L i 's are uniformly distributed over
the n -bit strings not assigned to previous L j 's. Namely, for every
{
0
,
1
}
α 1 ,...,α m
{
0
,
1
}
n ,
1
2 n m
i = 1 R i = α i ¬ ζ m =
m
Pr
(3.18)
whereas for every distinct β 1 ,...,β m ∈{
0
,
1
}
n ,
m
i = 1 L i = β i ¬ ζ m =
1
Pr
m
(3.19)
2 n
i
+
1
i = 1
Proof Idea: Eq. (3.18) follows from the observation that the R i 's are determined
by applying the random function H (3 n to different arguments (i.e., the R i 's),
where the distinctness of the R i 's is implied by
¬ ζ m . Similarly, the L i =
R i 's
are determined by applying the random function H (2)
n
to different arguments (i.e.,
the R i 's), and
¬ ζ m also conditions the results (i.e., the R i 's) to be different. Thus,
Eq. (3.19) also holds.
Claim 3.7.8.2: For every m 1,
2 m
2 n
Proof Idea: Fixing any i m , we consider the probability that R m + 1 = R i . There
are two cases:
Pr
[ ζ m + 1 ζ m ]
1. If R i =
R m + 1 , then certainly (since ( L i ,
R i )
( L m + 1 ,
R m + 1 )) we have
=
H (1 n R i =
H (1 n R m + 1 =
H (1 n R m + 1 =
R i =
L i
L i
L m + 1
R m + 1
2. On the other hand, if R i =
R m + 1 , then
Pr R i =
R m + 1 = Pr H (1 n R i
H (1 n R m + 1 =
L m + 1 =
L i
2 n
where the last equality holds because the random function H (1)
n
is applied to
different arguments (i.e., R i
and R m + 1 ).
[ R i = R m + 1 ] 2 n . In similarity to the foregoing Case 2,
conditioned on R i = R m + 1 ,wehave
Pr
Pr
Thus, in both cases,
R i = R m + 1 = Pr
H (2 n R i H (2 n R m + 1 = R i R m + 1 = 2 n
Thus, for every i
m ,
R m + 1 R i =
R i =
R m + 1 Pr
R i =
R m + 1 + Pr
R i =
R m + 1
R m + 1
R i =
Pr
2 · 2 n
and the claim follows.
Using
Pr
[
ζ m ]
Pr
[
ζ m 1 ]
+ Pr
[
ζ m ζ m 1 ] and Claim 3.7.8.2, it follows, by induc-
m 2
tion on m , that
Pr
[
ζ m ]
<
2 n . By Claim 3.7.8.1, conditioned on
¬ ζ m , the answers
Search WWH ::




Custom Search