Trust - Peer-to-Peer Computing: Applications, Architecture, Protocols, and Challenges - page 140

for e-transactions should consider the unbalanced transactions among
users.

2. Second, to address the lesser impact from small users, a reputation sys-
tem should not apply the same evaluation cycle for all peers. The super
users should be updated more often than the small users.

3. Third, with skewed transaction amount, it makes sense to evaluate the
large transactions more often than the small ones.

Global Reputation
Aggregation Weight

Buyer's Local
Trust Score

Seller's Local
Trust Score

Fuzzy Inference

Fuzzy Inference Fuzzy Inference

Remote Peer's
Trust Score

Transaction
Amount

Transaction
Date

Payment
Method

Payment
Time

Goods
Quality

Delivery
Time

(a) local trust inference (b) global trust aggregation

FIGURE 6.7: Local and global trust management mechanisms in FuzzyTrust
[Song et al., 2005].

As shown in Figure 6.7, the FuzzyTrust system works by performing two
major inference steps: local-score calculation and global reputation aggrega-
tion.

Local-Score Calculation. In FuzzyTrust, peers perform fuzzy inference on
local parameters to generate the local scores. Figure 6.7(a) illustrates
the local-score calculation for eBay transactions. The fuzzy logic sidebar
shows a detailed fuzzy inference procedure. The fuzzy inference mecha-
nism can capture some uncertainties and is self-adjusting. It can adap-
tively track the variation of local parameters such as payment method
and time, goods quality, and delivery time, etc.

Global Reputation Aggregation. The FuzzyTrust system must then ag-
gregate local trust scores collected from all peers to produce a global
reputation for each peer. The system uses fuzzy inference to obtain the
global reputation aggregation weights, as illustrated in Figure 6.7(b).
The aggregation weights are determined by three variables: the peer's
reputation, transaction date, and transaction amount.
Listed below are several frequently used fuzzy inference rules in the pro-
totype FuzzyTrust system construction. In a full-scale P2P reputation
system, the number of fuzzy inference rules could be extended to several
hundreds.