<s>
Neural	B-Algorithm
cryptography	I-Algorithm
is	O
a	O
branch	O
of	O
cryptography	O
dedicated	O
to	O
analyzing	O
the	O
application	O
of	O
stochastic	O
algorithms	O
,	O
especially	O
artificial	B-Architecture
neural	I-Architecture
network	I-Architecture
algorithms	O
,	O
for	O
use	O
in	O
encryption	O
and	O
cryptanalysis	O
.	O
</s>
<s>
Artificial	B-Architecture
neural	I-Architecture
networks	I-Architecture
are	O
well	O
known	O
for	O
their	O
ability	O
to	O
selectively	O
explore	O
the	O
solution	O
space	O
of	O
a	O
given	O
problem	O
.	O
</s>
<s>
At	O
the	O
same	O
time	O
,	O
neural	B-Architecture
networks	I-Architecture
offer	O
a	O
new	O
approach	O
to	O
attack	O
ciphering	O
algorithms	O
based	O
on	O
the	O
principle	O
that	O
any	O
function	O
could	O
be	O
reproduced	O
by	O
a	O
neural	B-Architecture
network	I-Architecture
,	O
which	O
is	O
a	O
powerful	O
proven	O
computational	O
tool	O
that	O
can	O
be	O
used	O
to	O
find	O
the	O
inverse-function	O
of	O
any	O
cryptographic	O
algorithm	O
.	O
</s>
<s>
The	O
ideas	O
of	O
mutual	O
learning	O
,	O
self	O
learning	O
,	O
and	O
stochastic	O
behavior	O
of	O
neural	B-Architecture
networks	I-Architecture
and	O
similar	O
algorithms	O
can	O
be	O
used	O
for	O
different	O
aspects	O
of	O
cryptography	O
,	O
like	O
public-key	B-Application
cryptography	I-Application
,	O
solving	O
the	O
key	O
distribution	O
problem	O
using	O
neural	B-Architecture
network	I-Architecture
mutual	O
synchronization	O
,	O
hashing	B-Algorithm
or	O
generation	O
of	O
pseudo-random	B-Algorithm
numbers	I-Algorithm
.	O
</s>
<s>
Another	O
idea	O
is	O
the	O
ability	O
of	O
a	O
neural	B-Architecture
network	I-Architecture
to	O
separate	O
space	O
in	O
non-linear	O
pieces	O
using	O
"	O
bias	O
"	O
.	O
</s>
<s>
It	O
gives	O
different	O
probabilities	O
of	O
activating	O
the	O
neural	B-Architecture
network	I-Architecture
or	O
not	O
.	O
</s>
<s>
Two	O
names	O
are	O
used	O
to	O
design	O
the	O
same	O
domain	O
of	O
research	O
:	O
Neuro-Cryptography	O
and	O
Neural	B-Algorithm
Cryptography	I-Algorithm
.	O
</s>
<s>
In	O
1995	O
,	O
Sebastien	O
Dourlens	O
applied	O
neural	B-Architecture
networks	I-Architecture
to	O
cryptanalyze	O
DES	B-Algorithm
by	O
allowing	O
the	O
networks	O
to	O
learn	O
how	O
to	O
invert	O
the	O
S-tables	O
of	O
the	O
DES	B-Algorithm
.	O
</s>
<s>
The	O
bias	O
in	O
DES	B-Algorithm
studied	O
through	O
Differential	O
Cryptanalysis	O
by	O
Adi	O
Shamir	O
is	O
highlighted	O
.	O
</s>
<s>
Hardware	O
application	O
with	O
multi	O
micro-controllers	O
have	O
been	O
proposed	O
due	O
to	O
the	O
easy	O
implementation	O
of	O
multilayer	O
neural	B-Architecture
networks	I-Architecture
in	O
hardware	O
.	O
</s>
<s>
One	O
example	O
of	O
a	O
public-key	B-Application
protocol	O
is	O
given	O
by	O
Khalil	O
Shihab	O
.	O
</s>
<s>
He	O
describes	O
the	O
decryption	O
scheme	O
and	O
the	O
public	B-Application
key	I-Application
creation	O
that	O
are	O
based	O
on	O
a	O
backpropagation	B-Algorithm
neural	B-Architecture
network	I-Architecture
.	O
</s>
<s>
The	O
encryption	O
scheme	O
and	O
the	O
private	B-Application
key	I-Application
creation	O
process	O
are	O
based	O
on	O
Boolean	O
algebra	O
.	O
</s>
<s>
A	O
disadvantage	O
is	O
the	O
property	O
of	O
backpropagation	B-Algorithm
algorithms	O
:	O
because	O
of	O
huge	O
training	O
sets	O
,	O
the	O
learning	O
phase	O
of	O
a	O
neural	B-Architecture
network	I-Architecture
is	O
very	O
long	O
.	O
</s>
<s>
The	O
most	O
used	O
protocol	O
for	O
key	O
exchange	O
between	O
two	O
parties	O
and	O
in	O
the	O
practice	O
is	O
Diffie	B-Protocol
–	I-Protocol
Hellman	I-Protocol
key	I-Protocol
exchange	I-Protocol
protocol	O
.	O
</s>
<s>
The	O
tree	O
parity	O
machine	O
is	O
a	O
special	O
type	O
of	O
multi-layer	O
feedforward	B-Algorithm
neural	I-Algorithm
network	I-Algorithm
.	O
</s>
<s>
The	O
output	O
of	O
neural	B-Architecture
network	I-Architecture
is	O
then	O
computed	O
as	O
the	O
multiplication	O
of	O
all	O
values	O
produced	O
by	O
hidden	O
elements	O
:	O
</s>
<s>
It	O
can	O
be	O
improved	O
by	O
increasing	O
of	O
the	O
synaptic	O
depth	O
L	O
of	O
the	O
neural	B-Architecture
network	I-Architecture
.	O
</s>
<s>
In	O
the	O
case	O
of	O
neural	B-Algorithm
cryptography	I-Algorithm
,	O
we	O
improve	O
it	O
by	O
increasing	O
of	O
the	O
synaptic	O
depth	O
of	O
the	O
neural	B-Architecture
networks	I-Architecture
.	O
</s>
<s>
The	O
output	O
of	O
neural	B-Architecture
network	I-Architecture
with	O
two	O
or	O
more	O
hidden	O
neurons	O
can	O
be	O
computed	O
as	O
the	O
exclusive	O
or	O
of	O
the	O
values	O
produced	O
by	O
hidden	O
elements	O
:	O
</s>
<s>
A	O
quantum	B-Architecture
computer	I-Architecture
is	O
a	O
device	O
that	O
uses	O
quantum	O
mechanisms	O
for	O
computation	O
.	O
</s>
<s>
That	O
gives	O
a	O
quantum	B-Architecture
computer	I-Architecture
in	O
comparison	O
with	O
a	O
conventional	O
computer	O
the	O
opportunity	O
to	O
solve	O
complicated	O
problems	O
in	O
a	O
short	O
time	O
,	O
e.g.	O
</s>
<s>
It	O
is	O
based	O
on	O
the	O
difference	O
between	O
unidirectional	O
and	O
bidirectional	O
synchronization	O
of	O
neural	B-Architecture
networks	I-Architecture
.	O
</s>
