There exists a mapping between the impact factors and the quantified value of each parameter for a given trust model. In our algorithm, the quantitative procedure of a parameter includes extracting the impact factors of each parameter, scoring the impact factors, and performing normalization and fuzzy integration of the impact factors to obtain the quantitative value of single parameter.
For a single parameter , C denotes the factor set: , and is the number of factors. The factor is extracted based on three considerations: the definition of parameter, evaluated points, range of parameter and some experience of experts.
For example, scalability is related to time complexity, space complexity, transmission, and efficient storage of data; the impact factors of sensitivity include the changing speed of trust value and handing speed of malicious attack and the speed of searching and timely reaction when network topology changes. For a single impact factor , , we evaluate it with specific measure, range. A simple method is using fuzzy theory to determine the range and level of the evaluated factor according to the experience of observer i.
Considering that there are some manufactured discrepancies for each factor, Delphi method can be introduced to collect and filter the divergent answers and obtain the quantified value. The Delphi method is an interactive forecasting method that relies on a group of experts. The experts answer questions in two or more rounds. Finally, the process is stopped after a predefined stop criterion e.
In this paper, several questions are defined firstly based on the multiple factors of a particular. Each question is followed by certain options that denote the level of possible answers i. These questionnaires are provided to several experts. After several rounds, the final correct feedback will be determined, and the final quantized value of a factor can be obtained. Repeat above quantization until all factors are quantized, denoted as. Notice that elements in may be measured in different unit; take the scalability as an example; the units of time complexity and space complexity are time ms and capacity kb.
Firstly, we normalize these different units. The popular method of normalization is max—min method: where and are the maximum quantized value and the minimum value determined by the range of j th factor. Repeat above disposal until all factors are normalized, denoted as. The final procedure is integrating the impact factors to obtain the overall quantized value of single parameter. As the impact factors are independent of each other, a simple integration is weighted sum of the quantized value of each factor.
The integral is defined as follows: where , , denote the weight of and satisfy. The weights are determined by the experience of the experts in consideration of importance degrees of evaluation criteria. The final integration is finished through 3 , and the quantized value of parameter is obtained, denoted as Q. Repeat all the above procedures until all parameters are quantized, denoted as where is a quantification function with all the above procedures and varied for different parameter.
A fusion algorithm based on fuzzy inference is proposed to combine the parameters in hierarchical structure in Figure 1. In Figure 1 , the middle criteria layer and lower parameters have certain relation. Several parameters are related to one or more factors in criteria layer. As the parameters are dependent on each other, fuzzy integral in 3 is inappropriate.
Without loss of generality, we suppose that one factor in criteria layer is related to all the parameters, and the goal layer is related to all the factors in criteria layer. Firstly, the weights of parameters to single factor in criteria layer and the weights of factors in criteria layer to the goal layer are calculated.
The entropy-weight coefficient method is a quantitative objective method and will be applied in our paper. Entropy-weight coefficient method is a quantitative risk evaluation method [ 22 ].
The relative importance of a risk factor to an evaluated system can be measured by its entropy, which is calculated by the fusion of probability values denoting the supporting degree of risk factors to indexes of evaluation set for the system.
In this paper, the parameters are considered as risk factors; one factor in the criteria layer is considered as evaluated object. Set several statuses for the evaluated object, give the probability of each parameter at each status, and apply entropy-weight coefficient method calculating the relative importance weight of each parameter to one upper factor. The statuses can be set based on certain evaluation set i. The detailed procedure of entropyweight coefficient method is referred to in related book i.
Repeat the above calculation until all the weights are obtained. The weight of parameters of k th , being the number of factors in criteria layer factor in middle criteria layer is denoted as , where is the number of parameters. For the sake of simplicity, discard the weights that equaled 0 e. Repeat the above filtering; the weights of parameters to one factor in criteria layer and the weights of factors in criteria layer to the goal layer are obtained.
The evaluated values of single factor in criteria layer and the evaluated value of goal layer will be fused by fuzzy inference in succession. In fuzzy set theory, a variable , denoting the value of object at the point k -level value according to the defined membership functions in a given discourse domain, and the problem is how to obtain under a given tree i.
We evaluate k th factor in the criteria layer followed by parameters. Set a discourse domain for the k th factor e. The quantitative values of parameters have been achieved in Section 4. The weighted vector is , and then the overall vector of the k th factor is denoted as. Define the evaluated value of the k th factor: , where is the maximum membership degree of. Repeat the above procedure until all the evaluated values of the factors in the criteria layer are obtained.
Based on the evaluated value of the factors in the criteria layer and the weights of factors in middle layer to the final goal layer, the fuzzy comprehensive judgment is performed with the same method as that used in calculating the evaluated value of the k th factor to obtain the comprehensive analytic value for a trust model.
Then, the observer can compare the eventual evaluated value with the threshold to judge whether the trust model is qualified. The threshold is set based on some factors, for example, accuracy and fee. We can evaluate a set of trust models, sort the evaluated values, and choose an optimal trust model usually the model with maximal evaluated value for implementation. The main steps of the proposed method are summarized: 1 Based on the structure in Figure 1 , apply entropy-weight coefficient method to calculate the weights of parameters to factors in criteria layer and the weights of factors in criteria layer to goal layer.
Meanwhile, determine the distributive parameters for k th factor in criteria layer. And judge whether the given model satisfies the request according to the defined threshold. In this section, some discussion, a concrete evaluation experiment, and the effectiveness of the proposed method are addressed in Sections 6. Consider the following: 1 Notice that the hierarchical model is an open structure that other parameters and decision factors can be integrated into this model, which reflect the flexibility of the proposed method.
Moreover, the hierarchical model is a reference model, and more than three layers might exist when subfactors are being linked to the parameters or the factor in criteria layer. And the quantized values of parameters varied from one trust model to another.
A concrete evaluation experiment is performed. Six traditional trust models analyzed in Section 3. Some conditions are set as follows: 1 A concrete scene: one user service requester performs the file download in P2P network. And the service requester pays more attention to the speed and quality of file download.
According to the procedures in Section 5. From Table 2 , we can see that multidimensional trust model reaches the highest score, as it has more parameters than others, and the robustness and scalability receive higher score in quantitation. PeerTrust compared to EigenTrust, although with better transitivity; worse scalability eventually leads to a smaller overall evaluated score, as the weight of scalability is larger than that of transitivity under the service requester policy.
We can see that none of the candidate trust models satisfies all the parameters. If the threshold is 0. Nevertheless, we can select the relatively optimal trust model i. The efficiency of the proposed method: for a given model, for parameters, suppose that there are factors mostly.
Seven experts carry out two rounds of consultation, each of which needs time , and the combination of factors costs. The overall time complexity is , being small around 3—5 for each parameter , so the time complexity is controlled.
For the weights of parameters, it is needed to calculate the process of the entropy-weight coefficient, time complexity being. Moreover, the weights can be reutilized for the same scene and task. We further validate the effectiveness of the proposed method by comparing it with previous methods [ 15 , 20 , 21 ]. Firstly, the proposed method adopts multiple parameters to evaluate trust model; it is more comprehensive than other works in characterizing the trust issues.
Wojcik introduced a series of factors classified into four aspects in establishing a trust model, but the parameter functions were not considered. In Schlosser, three parameters were used to reflect trust. These methods had failed to reflect the comprehensive characteristics of a trust model. Secondly, in terms of accuracy, Yang proposed a black box model and compared a set of trust history sequences in the input with the output and then determined the performance of the trust model with sensibility and foreseeability.
Its accuracy depends on the initialization of trust and behavioral characteristic. Wojcik displayed entire process of establishing trust comprehensively, but no specific assessment is performed. Schlosser presented a formal model for describing multiple reputation systems, but only reputation systems are taken into account.
In our proposal, objective disposal of parameters as well as fuzzy inference is used to quantify the evaluated value of a trust model, the results are more objective and with higher accuracy. Thirdly, in terms of efficiency, the overhead for our method is controllable and man-made evaluation in Delphi method and the calculation of weights and the fuzzy inference contribute to the calculation load. Schlosser simulated the reputation system in the performance of resisting attacks with the granularity of single node, and the consumption increases with the increase of nodes.
The analysis results are shown in Table 3. Table 3 explains the superiority of the proposed method. The accuracy and efficiency are compared among three methods. Accuracy simulation: reflect the change of deviation -axis of evaluated results with the increasing experiment time -axis. The conditions are the same as that set in Section 6. We perform the experiment 20 times. Efficiency simulation: reflect the relationship between resource consumption i.
The initial number of evaluated trust models is 1 and increases by 1 when the experiment time increases by 1. The simulation results are shown in Figure 2. Figure 2 a describes the accuracy of the three methods.
Therefore, the proposed method is more accurate. Figure 2 b describes the efficiency of the three methods; the calculation load increases with the increasing of evaluated models. The results are in accord with analysis in Table 3. A new method is proposed to compare and evaluate the trust models with quantitative parameters in P2P file downloading scene in this paper.
The evaluated parameters are extracted from the trust related concepts and modeled into a hierarchical structure. The Delphi method, entropy-weight coefficient method, and fuzzy inference are applied to obtain a comprehensive evaluated value of a trust model.
The optimal trust model is selected according to the sorted overall quantized values of candidate trust models. Analysis and simulation results show that the proposed evaluation algorithm is reasonable and effective. The proposed method resolves the individuality issues, assisting a decision maker in choosing an optimal trust model to implement in specific context.
Moreover, the method also can be used to guide the newly generated trust model in theory so that it has better performance in parameter function and adaptability.
This is an open access article distributed under the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Article of the Year Award: Outstanding research contributions of , as selected by our Chief Editors. Read the winning articles. Journal overview. Special Issues. Academic Editor: Arash Habibi Lashkari. Received 26 Dec Revised 29 Apr Accepted 09 Jun Published 13 Jul Abstract Varied P2P trust models have been proposed recently; it is necessary to develop an effective method to evaluate these trust models to resolve the commonalities guiding the newly generated trust models in theory and individuality assisting a decision maker in choosing an optimal trust model to implement in specific context issues.
It may not display this or other websites correctly. You should upgrade or use an alternative browser. P2P or Scene? Peer 2 Peer P2P Votes: 14 Scene Votes: 33 Total voters Status Not open for further replies. CyberAff Active Member. Holla guys, we people were talking about P2P and Scene Release what's better! Scene Releases are better or P2P Releases? The Scene: The Scene is a term used to refer to a collection of communities of pirate networks that obtain and copy new movies, music, and games, often before their public release, and distribute them throughout the Internet and previously through BBSes.
Each specific subsection within The Scene has its own community and rules governing releases, and are made up of many smaller groups.
Groups gather in private IRC channels where they can easily coordinate with other members to "pre" and distribute releases. This has been the case for ten years. Click to expand Scene Blogs. P2P Blogs. Frequenters of release blogs know a thing or two about finding the freshest releases. And without question; these sites beat out public torrent indexers pre-times by hours, and even hold their own with some of the best private trackers.
Below, we take a look at an assortment of release blogs, and delve into what makes them so popular. May 28, , AM. Re: p2p vs scene as i was the one suggesting you should start making a list, i didn't mean you should make a thread for it.
Canadian is right, judging anything is subjective to each individual. Originally posted by Canadian View Post. Making a list of the "good" and the "bad" groups is pretty subjective, just like which tracker is better IPT and TL for example. Last edited by kingpete ; May 28, , AM. May 28, , PM. Originally posted by securecrt View Post. I have my Samsung 40' connected via hdmi to my computer but don't see much difference either. Some of the HDBits. There are a whole load of others that are good, and if you check on these forums, and tracker forums, you will find plenty of discussions on the various merits of the different groups.
This will give you a place to start but the most important thing is try them for yourself, and see which ones you prefer.
0コメント