what-when-how
In Depth Tutorials and Information
x
=
⇒
=
Trust
( ,
x y
)
f Trust
(
)
i j
,
i j
,
+
x
y
N
is the set of direct mutual neighbors of nodes
i
and
j.
e
indicates the number of
unsuccessful interaction.
=
DirectTrustValue
f Trust
(
)
i j
,
i j
,
he computation of the
indirecttrust
is the same as the
directtrust
value and just
needs to replace
N
by
M
, which is denoted as the set of all the neighbors of
i
which
are connected to node
j
directly.
From the simulation and scenario analysis in Reference 31 we can see that this
new model revokes malicious nodes more quickly and performs with better tolerance
for the node's occasional failure than the other two schemes we mentioned above.
11.4.6 Gravity-Based Trust Model
In Reference 18 there is an interesting trust model called the gravity-based trust
model according to the mechanics theory in physics. It considers the trust attenu-
ation with time and trust context. Its goal is to use the local values to calculate the
global trust. his model includes two steps: computing trusted social neighbor-
hoods and computing trust lows.
Step1.Computingatrustedsocialneighborhood
According to the small world theory, a node usually interacts with only a small frac-
tion of nodes in an STN. hus, the trust distance, which is deined as the distance
in the trust spaces can be calculated only based on partial and local information
in whole STNs. he idea of computing the trust distance comes from distributed
virtual coordinate systems from the network area [20,21].
According to the theory of the virtual coordinates, the nodes in STNs are
mapped onto a Cartesian space based on the delay measurements in the network.
In Reference 18 a virtual coordinate
y
i
is associated with the trust space of
the node
i
. he initial value of the virtual coordinate is assigned randomly.
d
ij
is
denoted as the trustworthiness between node
i
and node
j
. Reference 18 defines the
total error as a sum of a squared-error function as below:
∑
∑
E
=
(
d
−
y
−
y
)
2
ij
i
j
(11.13)
i
j
Reference 20 demonstrates that simulating a network of mechanical springs pro-
duces coordinates that lead to the minimum error function shown previously in
