1 Introduction

Internet users have increased rapidly by fast development spread of wired and wireless networks and the utilization range has also diversified while many services using the internet have become actively. In particular, the one of the services that are the most commonly used is the work to search information and get resources using network. Processing speed can be slow and system can stop because the throughput of the server is increased as the number of clients is increased in the client-server model so far. Emerging P2P(Peer-to-Peer) networks in order to complement these disadvantages is used widely for load balancing, abundant resources, and high transmission efficiency through a direct connection among nodes without the central server of high performance [1–3]. The application of P2P technique is widely used also in MANET(Mobile Ad Hoc Network) consisting of only mobile nodes due to these characteristics. It is required the technique that can manage the behavior of nodes using P2P and perform detection and removal of nodes doing malicious behavior because there is no central management authority [4--6]. For this, it is necessary to participate in data transmission of only trusted nodes through an authentication process of nodes constituting a MANET.

In this paper, we propose reputation-based trust management technique that can be done fairly TZ(Trust Zone) formation of the entire network, certification, and file transmission in order to provide P2P services having high reliability. The whole network are divided into non-overlapping TZ considering the mobility of the nodes and TZM(Trust Zone Manager) is used for the trust evaluation and certification of the member nodes in the TZ. The ratio which transmits packet participating in network is calculate and the node having the highest value is elected to TZM in each TZ. The node elected as TZM manages TZMT(Trust Zone Member Table) to save the certificate issued information and trust information issued to the member nodes. The member nodes in TZ can transmit data only when issued certificate from TZM have and improve the reliability of the network by excluding that node which does not issue a certificate or has low confidence participates completely. And the reputation-based trust evaluation is performed in order to block malicious behavior of nodes issued the certificate and transmit data fairly. That is, the reputation is calculated by using evaluation value received from each nodes and the trust value for each node by using this value is calculated. The selfish behavior of nodes by using the trust value are able to block. The performance of authentication technique and the excellent trust evaluation proposed in this paper is confirmed through experiments.

This paper is organized as follows. The characteristics of P2P network is searched in the chapter 2 and the trust evaluation technique in P2P environments studied before is described in the chapter 3. The hybrid trust evaluation technique proposed in this paper is described in the chapter 4. The performance of the proposed technique is evaluated through experiments in the chapter 5. Finally, chapter 6 concludes.

2 P2P network

2.1 P2P service method

P2P network is classified depending on the presence of a server managing centrally. That is, there are pure P2P method that all users are directly connected and transmits information of each other by relay method, and hybrid P2P method that links users while the central server maintains P2P list like file list and IP address of client [7].

Pure P2P method is entirely defendant on users and works without relying on any central server. The user looks for other users on the network and communicates directly. This model has the advantage of having a high scalability in terms that can use directly by connecting to a network [8]. However, it has the disadvantage that the bandwidth can increases and the efficiency of the search can be lowered because the work sending a query among peers is repeatedly generated for needed information.

Hybrid P2P has a characteristic that facilitates the search among peers by managing the information such as the identification information of peers which is connected to the current network or a list of shared resources [9]. The central server basically manages only list to other server and contents access of server on the list. The contents may be distributed in several places and the communication are made directly to each other (Fig. 1).

2.2 P2P service type

P2P technique has been applied to a variety of electric transaction which may occur in cyber space like internet as well as a file sharing.

Distributed file sharing service using P2P is a service that allows to share file between the internet visitors using the same P2P program. P2P collaborative service model starts initially to connect instant messaging protocol that is not compatible with each other on the basis of the XML [10, 11]. This is a collaborative model for communication between people and application and between applications as well as dialogue between peoples. P2P e-commerce is that transaction costs, marketing costs, and system implementation costs can be reduced and a highly efficient e-commerce service is available because each transaction actors exchange product information freely and transact through P2P network [12, 13]. The auction or reverse auction field by the existing B2B e-commerce can be replaced because real-time exchange of information and confidentiality about information exchange of buyers and sellers is available than the existing Web if P2P method is used.

3 Trust evaluation technique

3.1 Trust attack

The one of the most dangerous attacks among the trust attacks is on-off attack. On-off attack is that malicious user acts alternately normal or malicious behavior in order to participate continuously in the network and it is similar to sybil attack [14–16]. Sybil attack is a method which uses obtained the number of legitimate node ID to attack. That is, the malicious user participates in the network by creating a new node ID using identity generator. And it attempts to normal or malicious behavior. If malicious behavior is detected by a trust technique, It creates other node and participates again in network. Bad mouthing attack is an attack that a malicious node does a false evaluation of normal node and makes a lot of nodes to recognize normal node to the malicious node. There is a collusive cheating that forms a group and tries after the attackers participating in the network that trust technique is applied collude with each other. This attack is affected to the entire network by doing good evaluation each other in the group which malicious user forms. Also, bad mouthing attack doing bad evaluation to user of other group is accompanied [17].

3.2 The previous trust evaluation technique

The trust evaluation technique is to find nodes performing malicious behavior discussed earlier and exclude the network participation. This is divided into centralized method, global trust mechanism, and regional trust mechanism according to the calculating method. The centralized method is a method that a particular node manages the entire network and calculates the reliability. This technique is very effective, but the confusion in entire network is caused when the node having responsibility is out of order. This may be appropriate to small network but this is not suitable for large scale system such as distributed P2P.

Global Trust Mechanism is that all nodes participating in the network have a each unique reliability [18]. The evaluation of the neighboring nodes is reflected through an iterative calculation. A node finally evaluates the all nodes in the entire network. The typical technique is EigenTrust technique and PowerTrust technique using this technique.

EigenTrust [19] is a algorithm for determining the overall reliability of the other nodes in the distributed environment. Each node obtains the reliability of nodes in the entire network using reliability of the neighboring node. This is a technique for reducing the malicious peer sharing untrust resources and constrains the participation of malicious peer by reducing the reliability of peer when choosing a transaction subject.

Abrams et al. [20] proposes the technique that can evaluate entire reliability based on EigenTrust. This prevents malicious user operating reliability by providing a specific benefits to node sharing file. The comprehensive evaluation algorithm for the user is proposed using the download speed, the accuracy of data, and the size of file when each node evaluates reliability of neighboring node in the FSEM-EigenTrust that Shen Rao et al. proposes.

PowerTrust [21] selects the most trust node to a super node unlike EigenTrust, changes adaptively in the environment, and obtains reliability. The accuracy of reliability is increased through LRW(Look-ahead Random Walk) technique and super node based on DHT(Distributed Hash Table).

The regional trust mechanism is that the node itself calculates the reliability directly using evaluation received from a limited number of other node unlike Global Trust evaluation. The subjective evaluation of each node can be reflected. There is PeerTrust and SFTrust in regional trust mechanism and is mainly used in a distributed computing environment.

PeerTrust [11] as the reputation-based framework calculates the confidence by utilizing many parameters. The user comments with a higher reliability utilizing the weight are much reflected than otherwise user when calculating the reliability.

SFTrust [22] is proposed framework storing reliability, calculating and updating in the distributed P2P environment. The reliability of service and feedback is calculated by separating.

PET [23] is a model that Liang et al. [18] proposes for the purpose of file sharing in P2P network and this evaluates reliability and risk of suppliers as resource shares. The risk is used for supplement the reliability and it can mean the reliability of user for a short time if the reliability expresses reliability for a long time. The risk of user providing reliability is decided and evaluated according to indirect reliability providing other user than the risk of supplier in R2Trust [24].

FileTrust [25] technique manages the reliability of sharing resource as well as user by separating sharing resource and reliability of individual user. It is blocked traffic is concentrated to user having a particular resources by selecting the resources having the highest reliability and the node having lower contribution among users providing resource when a transaction subjection is selected.

3.3 Trust management technique

Trust management in P2P service is directly related to the quality of service(QoS) provided for users. Figure 2 shows the classification of the trust management system.

Fig. 1
figure 1

Operation of the hybrid P2P

Full size image
Fig. 2
figure 2

The classification of trust management system

Full size image

The policy-based trust management system uses a certificate validation to establish a trust relationship. The main goal of this system is to enable the access control. The concept of the trust management restricts to limit access of resources by the defined policy according to the certificate validation and application.

The reputation-based trust management system is a mechanism that allows to establish a trust between peers. The reputation is the value obtained by the direct or indirect experience in the initial interaction with user. The peer requesting the resources in reputation-based system can be evaluated based on the reliability of the peer providing resources and resources. The following information should be considered for the reputation-based trust management.

  1. A.

    The save of the reputation information

The reputation information can be used easily at any time only when this for participants is distributed to multiple users and stored because the participants using P2P service do not maintain always online state. In addition, this should be searched effectively because this is opened information to use freely anybody.

  1. B.

    Integrity of reputation information

The reputation information can be easily processed using the access control mechanism because this is managed by server when the central server exists. However, the other method such as encryption should be used because a separate trust organ does not exist in a distributed environment.

  1. C.

    The reputation calculation

The calculation method using a sum or average of evaluation received from users is most used when the reputation of user is calculated because the performance of the overall system can reduce by complex calculation.

  1. D.

    The change of ID

It is easy to change ID due to the distributed feature of P2P and there is no the additional cost. The reputation should not give the any benefit to new subscriber and even though ID change of peer having a low reputation because this is a value maintained continuously by the various transaction.

The social network-based trust management system uses a social trust relation between peers in calculating a confidence and reputation value. In particular, this system constitutes a confidence regarding the peers through a social network analysis representing a relationship in the community.

4 The proposed hybrid trust evaluation model

In this chapter, we propose a hybrid trust evaluation mechanism for reliable P2P support. First, trust zone formation and authentication is made to check the confidence for all nodes participating in the P2P network. Second, the reputation-based trust management technique is proposed in order to perform file transmission safety and fairly.

4.1 Trust zone

MANET can share a variety of information through a direct connection between nodes without the central server because it is consisted of only mobile nodes. Thus, all nodes operate with the same authority and responsibility and can use effectively rapid exchange of information, cost reduction, and communication bandwidth. Many problems in reliability can cause because there is no central management authority that can manage the behavior of nodes participating in the P2P and control malicious nodes. In particular, sharing of invalid data by malicious nodes or add of malicious code causes to lower the reliability of P2P and decrease the QoS of the sharing data. However, the trust check between the nodes cannot be assisted efficiently if the entire network is in a plane structure. The accurate reliability evaluation of node cannot be made by only collection of limited information of mobile nodes. Therefore, network type of hierarchical structure is needed to collect properly an accurate confidence information of nodes in the entire network and provide efficient authentication. This can ensure somewhat the system performance such as throughput and delay even if the number of nodes is large or frequent movement. In this paper, MANET is divided into non-overlapping Trust Zone(TZ) of constant size. The nodes in TZ divided to part network like this is classified largely to Trust Zone Manager(TZM) and member node. Each TZM receives a unique ID of each zone and manages the trust information of member nodes in the TZ. Information exchange between TZM nodes is performed to exchange the trust information of the node if the member node is moved to the inside TZ from the outside. The member nodes can participate in the network through the trust evaluation and authentication of member nodes on the basis of this information. Figure 3 shows the structure of TZ formed for a trust evaluation of nodes in this paper.

Fig. 3
figure 3

The structure of trust zone

Full size image

4.2 Trust zone manager

The election of TZM that evaluates the trust of member nodes in the TZ and manages is very important because TZM is possible to exclude the network participation of malicious nodes according to the confidence for member nodes and the certificate issue. All nodes internal TZ transmit a trust value of neighboring nodes with HELLO message for election of TZM in this paper. The trust value is the ratio participated in packet transmission and has between 0 and 1. The highest node among the broadcast trust information values is elected to TZM. The member nodes of the TZ is able to participate after a certificate is issued by elected TZM. So, TZM has TZMT(Trust Zone Member Table) for trust evaluation of the member nodes internal TZ and the trust information of member nodes will be updated. The information is collected from the closest neighboring TZM. The structure of TZMT is shown in Fig. 4.

Fig. 4
figure 4

The structure of TZMT

Full size image

The zone history field in Fig. 3 represents TZ that the node is joined and this information can be utilized usefully to figure out the location of malicious nodes. The node k request a certificate issue to TZM of TZ 1 if a node k moves to the TZ 1. TZM of TZ 1 broadcasts the trust information of node k to neighboring TZMs to process this request. The certificate is issued if the trust value of node k is over the reference value, otherwise the certificate cannot be issued. The member nodes that do not issued the certificate cannot request a data and only increase the trust value by participating in the data transmission. The trusted reference value of the certificate issue reference is calculated by the following equation.

$$ T(v)=avg\Big({\displaystyle \sum_{i=1}^n\left(T{Z}_i\right)\times t} $$
(1)

Here, t is newly set each time a node is selected and set to 30 s. If the trust information of node k does not receive from neighboring TZM, it is determined that the node participates first in network and the value of node k is initialized and stored in TZMT. The certificate of the node k is not issued and the certificate issue is determined by measuring the reliability again after the time participating in network is provided for a certain time. Figure 5 shows the process that ZH node issues the certificate of member nodes described above.

Fig. 5
figure 5

Certificate request and issuance process

Full size image

4.3 The node authentication and security routing

TZM performs as a certificate authority that issues the certificate through a trust check of member nodes. The process that TZM issues the certificate to member node k is as follows. First, TZM has a master key pair and member node k request the certificate issue (k, P k ) to TMZ after pairing-based encryption key pair ((s k , P k ), P k ) is created. TZM issues the certificate (D k , Q k ) to member node k after this checks the trust value of the node. The member node k performs data transmission using IBS(ID based Signature) electronic signature. That is, member node k creates a random number and transmits a signature made using authentication (D k , Q k ) to the destination node. Figure 6 shows the progress of issuing certificate to member node in the ZH node.

Fig. 6
figure 6

Detailed procedures for the certificate

Full size image

The malicious nodes can be excluded and the secure P2P service can be provided by transmitting data after only member nodes issued certificate participate in network.

4.4 The reputation-based trust management

The member nodes issued certificate from TZM will be able to share files with the neighboring nodes. There is no unconditionally trusted despite the node issued certificate from TZM because all nodes participating in the service should cooperate with each other in order that file sharing between nodes is made smoothly. It is an important factor that the performance of the system is degraded if nodes perform a selfish behavior like that they do not share their file or transmit the query of other nodes. Therefore, the reputation is calculated using evaluation value received from each node after a transaction of nodes participating in P2P is performed and the trust value of each node is calculated using this value. The reputation value of each node manages in the TZM and all nodes transmit the evaluation results to TZM after they complete the transaction. The nodes refer the trust value to selection of transaction target and the service use is differentiated according to the trust value by accepting transmission request or refusing through comparison of the trust value. In this way, the fairness is provided because the node can use resource that other nodes provides and service as much as the node contributes depending on the quality of resource that the node provides and behavior. Figure 7 shows the proposed reputation-based trust management process.

Fig. 7
figure 7

Detailed procedures for the certificate

Full size image

The node broadcasts to neighboring nodes to download the required file. The node received a response from a neighboring nodes requests the trust check of the neighboring nodes to own TZH. The file download is rejected if the trust value of neighboring node is lower than itself and the file is downloaded by selecting the neighboring node that has the higher trust value than itself. Thus, service use will be limited if the trust value is lowered due to malicious behavior of nodes. The evaluation value of node that is 1 if it is same to the request file, and is −1 otherwise is transmitted to TZH after the file service is completed. However, −1 is assigned to the node and 0 is assigned to a node providing a file if the service is terminated without evaluation. The selfish behavior of nodes are blocked and the fair P2P service is able to perform by managing reputation of nodes in this way.

5 Performance analysis

5.1 Experimental environment

In this chapter, we comparative experimented with PeerTrust technique to evaluate the performance of hybrid trust evaluation technique proposed in this paper. The size of network used in the experiment was 1500 × 1500, the transmission range was 200 m, and experiment time was 300 s. The mobile node model used in the experiment was random-way point model and moved at a speed between 0 ~ 20 m/s. The battery consumption of the mobile nodes was not considered. The experiment was based on the file sharing and it was evaluated using on-off, sybil, bad mouthing attack scenario according to the number of malicious node. Table 1 shows the value of an environment variable used in the experiment.

Table 1 Simulation parameters
Full size table

5.2 Experiment result

The excellent performance of the trust evaluation of nodes is measured through comparative experiments with PeerTrust technique in this paper. The criteria of performance evaluation is measured data transmission rate by the number of malicious nodes and attack types. In other words, we measured exactly how the data is transmitted without being affected by a malicious node.

Figure 8 is the result measured success rate of data transmission for the on-off attack. The PeerTrust technique as shown in Figure is that success rate is significantly less as the number of malicious node. This technique gives a higher weight to user who has a high reliability and calculates when the reliability is calculated. The accurate trust evaluation of malicious node can not perform because a higher value is given accordingly when malicious nodes do normal behavior. This technique is not respond actively to on-off attack that gives the impact on the network a long time. However, The trust evaluation is performed when the node is evaluated for a certificate issue of nodes participating in the network in the proposed technique. The evaluation performance of the malicious nodes showed excellent results because the evaluation information of nodes participating in data transmission is managed consistently by TZM.

Fig. 8
figure 8

Success rate of data transmission for on-off attack

Full size image

The result of data transmission success rate by bad mouthing attack scenario is shown in Fig. 9. It is confirmed that the trust evaluation of malicious nodes is made exactly because this attack scenario is relatively simple attack. The PeerTrust technique performed accurate evaluation for bad mouthing of malicious nodes because it use many parameters when it evaluates the trust of nodes. The success rate was decreased because the number of node which has high reliability is decreased when the number of malicious node increases. On the other hand, the performance of the proposed technique was not affected significantly even if the number of malicious nodes increases because it performs the trust evaluation of the nodes in two stages.

Fig. 9
figure 9

Success rate of data transmission for bad mouthing attack

Full size image

Figure 10 shows the results for the performance of node trust evaluation in the sybil attack. It is not easy that Sybil attack does the trust evaluation of malicious nodes because this attack performs normal and abnormal behavior of malicious nodes and attacks again using the legitimate ID. As shown in the Figure, the performance was significantly less as the number of malicious nodes became many in the PeerTrust technique. However, it is not easy to attack by malicious node because the certificate issue is limited when the node participates first in the network in the proposed technique. The malicious node must receive a certificate to participate in the network and perform malicious behavior when it participates again in the network using a legitimate ID after the malicious node is evaluated. The proposed technique showed excellent result because this behavior is easily detected by the trust evaluation.

Fig. 10
figure 10

Success rate of data transmission for sybil attack

Full size image

6 Performance analysis

The computing environment has been changed largely through a wired or wireless network development. In particular, the distributed processing model that solves the problem using processing power of multiple computers is getting popular and P2P service among them has attracted. P2P service is a technology that it connects user searching information and user having information without going through a central server and shares resources with each other. It is a set of service made by appling the technology. The P2P service in MANET is being actively researched because this structural feature is very similar to the MANET. The trust evaluation and maintenance between each nodes is very important because P2P in MANET is configured only mobile nodes. The evaluation method for nodes is very important as provision by the bad data and the insertion of malicious code can be decreased the reliability of the network. Therefore, the TZ structure not planar structure is used to provide accurate evaluation of nodes in this paper. TZM is elected for the trust check and authentication in member nodes of each TZ. TZM elects the node that has the high participation rate, ie, the high reliability. TZM issues the certificate if the trust information is above the reference value after it collects the trust information of member nodes requesting the certificate from neighboring TZM. TZMT is managed to store the trust information and certificate issue information of member nodes. Only member nodes issued certificate like this can transmit data. The nodes that do not issued certificate cannot participate in data transmission and they can transmit data after their trust value are over than a reference value by participating only data transmission. The nodes that have a lower value than there are rejected for data request by requesting the trust information of the node to TMZ when nodes transmit data. Each nodes implement the reputation-based trust evaluation performing evaluation of the node the TZM when data transmission is complete. In this way, the selfish behavior of nodes can be blocked. The proposed technique in this paper was comparative experimented with PeerTrust and the excellent performance was confirmed by experiment result.