Information Technology Reference
In-Depth Information
The resulting sequence
B
1
B
2
...B
2
r
consists of data such that every pair
B
2
i−
1
B
2
i
,i
=1
, ..., r
, contains increasing addresses. Let us consider
B
1
B
2
...B
2
r
to encode 1 in the covert channel. The data in this sequence is split into pack-
ets. Packet sequences are transmitted to the corresponding addresses. These
sequences can contain additional packets, e.g. for establishing other connections,
so GNH(0) agent should not consider the additional packets, though some errors
are possible. The received sequence is 1 of the covert channel. 0 is encoded by
a sequence of decreasing address pairs of data. A sequence of unordered address
pairs encodes a delimiter
x
. We use an assumption that the PROXY server estab-
lishes connections with other PROXY servers according to the addresses in data
queue buffer, and packet block with the same destination address is transmitted
in a single connection.
There emerge the problems of estimation of the value of
r
for reliable ex-
traction of 1, 0 and
x
, and of investigation of transmitter fault tolerance. Due
to the fact that data permutations in a queue are stochastic, and there exists
a possibility of errors in address sequence in GNH(0), there exist the following
errors in covert channels:
- data loss (loosing an address s in address sequence restored in GNH(0));
- data insertion.
3M them lMod l
Let
s
=(
s
(1)
,s
(2)
, ..., s
(2
r
)) be the data address sequence determining one bit.
To restore this bit taking into consideration possible errors we count all increas-
ing and decreasing address pairs. The decision about the value of the bit is made
by the means of mathematical statistics. In this paper we consider the problem of
correct bit recognition by the sequence of data addresses. Bit recognition based
upon packet sequence is not considered.
Let input data addresses be random values
ξ
1
, ..., ξ
k
that are produced
independently with equal probabilities
P
(
ξ
j
=
s
i
)=
m
. Due to the fact that
packets with the same source address are transmitted in a single connection,
we delete all sequences with the same source address and replace them by a
single representative. After this transformation we get a sequence
η
1
, ..., η
2
r
.
This sequence is a simple Markov chain with the transition matrix
P
(
η
i
+1
=
s/η
i
=
s
)
,
where
1
P
(
η
i
+1
=
s/η
i
=
s
)is equalto
=
s
,
and 0
,
if
s
=
s
.
1
,
if
s
m
−
The initial distribution is uniform, and transition matrix is twice stochastic,
hence the Markov chain is stationary, with one acyclic ergodic class without
insignificant states.
