Trust - Peer-to-Peer Computing: Applications, Architecture, Protocols, and Challenges - page 135

transaction. The second part scales the first part by an increase or decrease
of the trust value based on community-specific characteristics and situations.
For the first part, there are two variations in defining the credibility mea-
sure. The first one is to use a function of the trust value of a peer as its credibil-
ity factor. Thus, feedback from trustworthy peers is considered more credible
and, consequently, weighted more than that from untrustworthy peers. This
definition of credibility measure is based on two assumptions: (1) untrustwor-
thy peers are more likely to submit false or misleading feedback in order to
cover up their own malicious behavior; (2) trustworthy peers are more likely
to be honest on the feedback they provide. Accordingly, considering only the
first component, the trust metric is now given by:

I(u)

T (p(u, i))

T T V M (u) = S(u, i) (6.7)

I(u)
j=1

T (p(u, j))

i=1

The second possible credibility measure is for a peer w to use a personal-
ized similarity measure to rate the credibility of another peer v through w's
prior interactions experience. Specifically, peer w uses a personalized similar-
ity between itself and another peer v to weight the feedback by v on any other
peers. Let IS(v) denote the set of peers that have interacted with peer v.
Thus, the common set of peers that have interacted with both peer v and w,
denoted by IJ S(v, w), is given by IS(v)∩IS(w).
To measure the feedback credibility of peer v, peer w computes the feed-
back similarity between w and v over the common set IJ S(v, w) of peers that
they have interacted with in the past. Here, the feedback by v and the feedback
by w over IJ S(v, w) are modeled as two vectors. As a result, the credibility
can be defined as the similarity between the two feedback vectors. The root-
mean-square or standard deviation (dissimilarity) of the two feedback vectors
can then be used to compute the feedback similarity. Accordingly, the trust
metric (considering only the first component) is given by:

I(u)

Sim(p(u, i), w)

T P SM (u, w) = S(u, i)

(6.8)

I(u)
j=1

Sim(p(u, j), w)

i=1

where:

2

x∈IJ S(v,w)

P I(x,v)
i=1 S(x,i)

P I(x,w)
i=1 S(x,i)
I(x,w) I(x,v) −

Sim(v, w) = 1− (6.9)

|IJ S(v, w)|

Figure 6.4(a) gives a sketch of the system architecture of PeerTrust. First
of all, we can see that there is no central database, implying that trust data
that are needed to compute the trust measure for peers are stored across
the network in a fully distributed manner. The trust manager performs two
main functions. Firstly, it submits feedback to the network through the data