Hostname: page-component-7b7567c8c-2n9c9 Total loading time: 0 Render date: 2026-11-18T04:59:41.816Z Has data issue: false hasContentIssue false

Taxonomy of reputation assessment in peer-to-peer systems and analysis of their data retrieval

Published online by Cambridge University Press:  03 October 2014

Farag Azzedin*
Affiliation:
Information & Computer Science Department, King Fahd University of Petroleum and Minerals, Saudi Arabia e-mail: fazzedin@kfupm.edu.sa
Rights & Permissions [Opens in a new window]

Abstract

The need for reputation assessment is particularly strong in peer-to-peer (P2P) systems because the peers’ personal site autonomy is amplified by the inherent technological decentralization of the environment. However, the decentralization notion makes the problem of designing a P2P-based reputation assessment substantially harder in P2P networks than in centralized settings. Existing reputation systems tackle the reputation assessment process in an ad hoc manner. There is no systematic and coherent way to derive measures and analyze the current reputation systems. In this paper, we propose a reputation assessment process and use it to classify the existing reputation systems. Simulation experiments are conducted and focused on the different methods in selecting the recommendation sources and collecting the recommendations. These two phases can contribute significantly to the overall performance owing to precision, recall, and communication cost.

Figure 0

Figure 1Classification of reputation systems

Figure 1

Figure 2Reputation assessment process

Figure 2

Figure 3Recommendation retrieval

Figure 3

Figure 4Recommendation evaluation

Figure 4

Table 1Classification of existing reputation systems

Figure 5

Figure 5Flow of recommendation retrieval simulation.

Figure 6

Table 2Exogenous and design parameters used in the simulation

Figure 7

Figure 6Algorithm I for unstructured overlay network construction

Figure 8

Figure 7Unstructured overlay network using Algorithm I (largest island)

Figure 9

Figure 8Algorithm II for unstructured network construction

Figure 10

Figure 9Unstructured overlay network using Algorithm II

Figure 11

Figure 10Average hop count in P-Grid

Figure 12

Figure 11Recall values (1024 peers)

Figure 13

Figure 12Node reachability (FL1)

Figure 14

Figure 13Node reachability (FL2)

Figure 15

Figure 14Precision values (1024 peers)

Figure 16

Figure 15F-measure values (1024 peers).

Figure 17

Figure 16Number of sent messages for 1024 peers (in thousands)

Figure 18

Figure 17Percentage of duplicate sent messages (1024 peers)

Figure 19

Table 3Number of reply messages (1024 peers)

Figure 20

Figure 18(a) Recommendation tree with depth-first search (DFS), (b) recommendation tree with breadth-first search (BFS)