what-when-how
In Depth Tutorials and Information
his model in Reference 16 is simple and just considers a few factors that affect
trustworthiness.
A more sophisticated model which takes more factors into account in trust
computation is proposed in Reference 1.
1. Direct trust:
=
× +
×
×
×
dr
dr
t M i
( , )
j
L i
( , )
j
w L i
( ( , ))
j
F j
( )
(11.8)
ij
t
−
1
+
×
×
×
K j
( )
f x
( )
P x
( )
F y
( )
dr
t-
1
is the accumulative direct trust values of node
j
until
t
−
1 times.
t
= −
1 ( ( , )
L i
j M i
⋅
( , ))
j
(11.9)
he above is defined as the trust decay in terms of time.
M
(
i,j
) is the satisfac-
tion degree from node
i
to node
j
.
L
(
i,j
)is the shortest length between node
i
and node
j
.
w L i
( ( , ))
j
=
exp(
−
L i
( , ))
j
(11.10)
he above is the weight of the link.
=
+
−
P x
( )
1 1
(
exp(
n
))
(11.11)
he above is the recommendation punishment and
n
is the number of false
recommendations.
F
(
j
) is the risk that node
j
takes in the transaction.
F
(
y
) is
the transaction risk of node
y
which is recommended by node
j
. If node
j
is
the member of the clique,
K
(
j
) equals 1. Otherwise
K
(
j
) is 0. If the recom-
mendation is false,
f
(
x
) equals −0.5. Otherwise
f
(
x
) is 0.
2. Indirect trust:
⎧
N
⎫
1
⎪
⎪
⎡
⎢
Cxi
Cxi Rxi
⎤
⎥
⎪
⎪
∑
1
(11.12)
ir
=
ir
× +
t
×
L x j
( , )
×
w L x j
( ( , ))
M x j
×
( , )
xj
t
−
1
+
N
N
is the number of recommendation nodes.
Cxi
/(
Cxi+Rxi
) is the trust degree of
recommendation nodes.
he expression of the overall trust is the same as Equation 11.7. According
to simulation results of [1,16], these two models can recognize malicious behav-
iors such as boast and cheating in STNs. Based on the small world theory they
can efficiently handle relationships such as establishing new connections and
terminating old connections. he trust feedback mechanism, which consists of
