what-when-how
In Depth Tutorials and Information
express truthfully, so (R) should be also taken into account. he factor (N) extends the
factor (O) to a distribution level that makes factor (O) more reliable. Age of the opin-
ion (T) indicates the freshness of the opinion. he inluence of the opinion on reputa-
tion will become weakened when the opinion exits for a long time in the sociotechnical
network. Reputation attaches importance to community context factor (W).
he node in an environment where most of the nodes communicate so honestly
(such that the community has a high overall reputation) is prone to be reliable.
Besides, another issue should be considered in reputation modeling; for example,
node i trusts node j for his advising about cars but does not trust j with his rec-
ommendation about computers. hus, reputation should be expressed by a trust
space like Trust 1 , Trust 2 , Trust 3 , .... Trust n . Every Trust represents one aspect of the
reputation in a particular context, so we just discuss one aspect of reputation. Other
aspects can be calculated in the same way.
11.4.1 Generic Reputation Function [19]
he Generic Reputation Function described in Reference 19 is shown below:
N
λ
α
T T
(
Tj
)
X
R
α
R
X
R
O
α
O O
X
e
N
j
j
j
j
=
1
=
+
Φ
λ
Trust t
( )
W
e
/
T
(11.1)
n
k
m
N
N
+
W W
N
α
X
W
α
X
=
k
1
j
=
1
In this function there are two types of impact parameters: the impact variable (X)
and the impact weight ( α ). hey are used to control the direction of inluence and
the amount of influence factors (O, R, T, N, W) on the reputation of the node [19].
Each factor has its own impact factors. he freshness of the opinion should decay
gradually not suddenly, so an exponential function is involved. he variable λ is the
opinion decaying rate. he variable ϕ is used to assign the initial reputation value,
and it serves the dual purpose of stabilization.
his generic reputation function is the foundation of many other reputation
computation models. hese models may change somewhere in the format but they
all consider a part of the factors mentioned in the generic reputation function.
According to the requirements of the environment, the factors in the function can
be changed to emphasize the influence of several factors. In the experimental evalu-
ation of the function in Reference 19, during the period when the attackers express
their dishonest opinions to the target, if attacker frequency is higher than the evalu-
ator frequency, the reputation of the target will be affected seriously, especially in
the case where attacker frequency overwhelms evaluator frequency totally. On the
other hand, if the attacker frequency is lower, the reputation of the target will be
hardly inluenced. hus, the sudden change in reputation slope should be paid
attention to as a possible attack. Generally, the reputation function cannot detect
Search WWH ::




Custom Search