 Research
 Open Access
Reconfiguration time and complexity minimized trustbased clustering scheme for MANETs
 Sunho Seo^{1},
 JinWon Kim^{1, 2},
 JaeDong Kim^{1, 2} and
 JongMoon Chung^{1}Email authorView ORCID ID profile
https://doi.org/10.1186/s1363801709388
© The Author(s) 2017
 Received: 23 February 2017
 Accepted: 25 August 2017
 Published: 18 September 2017
Abstract
A trust management mechanism for mobile ad hoc networks (MANETs) is proposed to cope with security issues that MANETs face due to time constraints as well as resource constraints in bandwidth, computational power, battery life, and unique wireless characteristics. The trustbased reputation scheme GlobalTrust is a reliable trust management mechanism. In this paper, a clustering algorithm is applied to the GlobalTrust scheme (named Clusterbased GlobalTrust (CGTrust)) to find the optimal group size to minimize the configuration time, which consists of trust information computational time and complexity, while having to satisfy the trust reliability requirements. The optimal number of clusters is derived from the minimizing point of the computation complexity function. Simulation results show that the computational time and complexity of CGTrust are controllable and can be used effectively in time critical network operations that require trust analysis.
Keywords
 MANET
 Trust
 Cluster
 GlobalTrust
 CGTrust
1 Introduction

CGTrust evaluates trust at cluster heads (CHs) and at the TA in contrast to GlobalTrust that focuses on the TA. This approach has the benefit of requiring less time and uses less computing resources at the TA in the network.

CGTrust provides a computational complexity minimized mechanism to form MANET clustering, where the complexity considers both intra and inter cluster computations. This algorithm helps to drastically reduce the complexity associated with the trust evaluation and computation.

CGTrust provides a mechanism to evaluate the trust of both nonCH nodes and CHs to prevent false trustworthiness decisions of nonCH nodes by a nontrustworthy CH and minimize the setup and reconfiguration time.
2 CGTrust
where p _{ w,u } and n _{ w,u } are respectively the total number of positive events and the total number of negative events. If there are no events between node w and node u, the LTO is null. The sum of positive and negative events (p _{ w,u }+n _{ w,u }) is the sum of the number of events sent by node w to node u. Positive events represented by p _{ w,u } are events that node w is determined that a packet was transmitted well to node u. Negative events represented by n _{ w,u } are events that node w is determined that a packet was not delivered to node u for some reasons (noise, wrong decision, intentional packet drop, etc.). The transmission can be determined by an acknowledgement packet, etc. As can be seen from Eq. (1), the higher the positive events, the closer to 1 the LTO and the closer to 0 the negative events become. The higher the LTO, the trust opinion of node u approaches the reputation of an honest node.
An honest node correctly sends a packet to a predefined route when it receives a packet from another node, except for errors that occur during the transmission process. In addition, when the LTO is reported to the TA or CH, the reporting of the LTO is performed without distortion.
A malicious node may drop a packet when it is received from another node, or intentionally send a packet to a different route. In addition, information may be distorted during the transmission and reporting process of the LTO to the TA or CH.
where Sim(w,j) is the cosine similarity of the LTOs between node w and node j, S _{ u } is the set of nodes that have nonnull LTOs over node u (including w if w has one), and HR_{ j } is the hierarchical rank of node j. HR_{ j } can specify that administrators have different values, or they can all have the same value, depending on how trustworthy the node’s opinion is [2]. The TA should have a hierarchy rank higher than that of any CH or node, because the TA’s opinion is more strongly reflected than the CHs or nodes as it is where the reputation judgment is made. In this paper, the hierarchy rank is designated as TA = 3, CH = 2, and node = 1.
After calculating the SR, CH_{ i } makes a SR matrix. The SR tuple in node w’s view is denoted as a vector. CH_{ i } compares all SR tuples, and merges the two SR tuples with the least difference based on the agglomerative hierarchical clustering technique [2]. CHs use this technique to form its trusted quorum (D _{ CH }). The process of finding the trusted quorum ends when the set having the largest number of nodes has more than half of the total number of network nodes. When the total number of nodes is N, the number of nodes belonging to the trusted quorum may have a value between [N/2, N], depending on how it is calculated. Since all SR tuples are compared and a trusted quorum is found, the complexity does not change depending on the size of the trusted quorum.
where θ is a decision factor selected from the range in [0, 1]. The detection errors can be reduced by selecting the most appropriate θ value.
The TA evaluates each CH’s trustworthiness using direct trust computation, which can be conducted using the encrypted packet mode, which encrypts packets exchanged between nonCH nodes and the TA. In this mode, nonCH nodes send (encrypted) information packets to their CH and the CH forwards these packets to the TA without trust computation. Using the encrypted packet mode, the TA computes the CHs’ trustworthy level periodically considering β, which is the ratio of nodes that use encrypted packet mode in the cluster. In this computation process, β is a variable that represents the possibility that a CH is a malicious node. As the value of β increases, the number of nodes that the TA needs to directly compute increases, making it difficult to reduce the computational complexity. Considering the computational complexity of trust computation and the worst case where the majority of CHs are infected, the suitable value of β is [0, 0.5].
3 Clusterbased network analysis
3.1 Computational complexity analysis
The proposed scheme computes the GR of each node based on GlobalTrust that uses a trusted quorum D. GlobalTrust uses the agglomerative hierarchical clustering technique to find a minimum dominating cluster [2].
For an accurate complexity analysis, the method of [14] is applied to CGTrust, where the complexity of the pseudocode steps is computed. Each cluster has N/k nodes. Every node will collect the SRs of all other nodes in its cluster, which are ((N/k)1) SRs. Two nodes in a cluster will pair up and compare their collected SRs, but will exclude the SR of the paired node in this comparison process. Therefore, ((N/k)2) SRs will be compared by the node pair. In addition, since there are \(\binom {N/k}{2}\) combinations of possible node pairs in each cluster, the computational complexity of one cluster is O[((N/k) 2)(N/k)((N/k)1)/2] = O((N/k)^{3}) (step 3).
3.2 Trust information computation time
3.3 Minimization of computational complexity
4 Performance evaluation

Naive malicious attack (NMA): a malicious node provide improper services with probability α. However, it reports its LTOs honestly.

Collusive rumor attack (CRA): In addition to providing improper services with probability α, malicious nodes collude to report false LTOs. Malicious nodes report LTOs of 1 to malicious node and LTOs of 0 to honest nodes.

Noncollusive rumor attack (NRA): a malicious node can report a false LTO that is opposite to the observed evidence. For example, if an LTO is evaluated as p, the malicious node may report (1−p) as the LTO.

Malicious spy attack (MSA): some malicious nodes misbehave with probability α. Other malicious nodes behave honestly. These malicious nodes may collude and report LTOs of 1 to malicious node and LTOs of 0 to honest nodes to confuse the trust and reputation system.

Conflicting behavior attack (CBA): malicious nodes can collude to confuse the trust and reputation system such as CRA and MSA. However, they misbehave only to some of honest nodes, and report LTOs of 1 to malicious node and LTOs of 0 to honest nodes to confuse the trust and reputation system. This attack causes LTO disagreement among honest nodes, which makes it difficult to find malicious nodes.
CBA is considered the most demanding type of attack because it makes it difficult to distinguish malicious nodes by confusing LTO information of honest nodes with respect to other nodes. For the above reasons, CBA was selected and evaluated.
The simulation based performance analysis of CGTrust and GlobalTrust was conducted using Matlab with N nodes randomly distributed with a uniform density in a 2×2 km^{2} square area. Simulation parameters were set same to the experiments in [2], where the ratio of malicious nodes was set to 0.3 and every node randomly requests of its neighbor nodes to send a packet (which is multihop relayed) 100 times per minute, and β is in the range in [0, 0.5]. Honest nodes were made to drop packets based on a 0.05 packet error rate (PER) and the detection error probability of the monitoring system was set to 0.05. Each node was made to transmit trust data packets every 30 s and the TA computes the GRs based on the accumulated data of the past 30 min.
In addition, the probability that a malicious node drops a packet was set to 0.5, γ = 0.7, θ = 0.7, and the upper bound probability of FN and FP were set to 0.1 as used in [2].
In the simulation, the TA is not a target of a malicious node. If the TA is infected, trust decisions on network nodes will not be correct. It is assumed that CH and other nodes can be malicious nodes, based on the restriction that the malicious ratio is not more than 0.5. If the malicious node ratio is greater than 0.5, the malicious nodes can take control of all the opinions in the network and the trust decision cannot be determined correctly. Although it is assumed that the overall ratio of malicious nodes is less than 0.5, the proportion of malicious nodes in a cluster is not limited. Therefore, in some clusters, malicious nodes may not properly report to the TA because they have taken control of the cluster.
5 Conclusion
Mission supportive MANETs require fast updates on node conditions in order to properly support command and control operations. To support this objective, CGTrust was designed to minimize the time required to evaluate the trust profile of a MANET through optimal cluster size control applied to GlobalTrust. The simulation results show that for the number of nodes and malicious node ratios of practical interest (based on β = 0.1 and 0.5), CGTrust can be approximately 1000 and 10 times faster compared to GlobalTrust, respectively. In addition, the results also show that the FN probability is approximately 0.1 times lower when CGTrust is used instead of GlobalTrust for the malicious node ratio range of 0.05 to 0.5.
Declarations
Acknowledgements
This work was supported by the ICT R&D program of MSIT/IITP, Republic of Korea (B0101171276, Access Network Control Techniques for Various IoT Services).
Competing interests
The authors declare that they have no competing interests.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Authors’ Affiliations
References
 JH Cho, A Swami, IR Chen, A Survey on Trust Management for Mobile Ad Hoc Networks. IEEE Commun. Surv. Tutor.13(4), 562–583 (2011).View ArticleGoogle Scholar
 X Chen, J Cho, S Zhu, in Proceedings of the IEEE International Conference on Sensing, Communication and Networking (SECON) 2014. GlobalTrust: An AttackResilient Reputation System for Tactical Networks (Singapore, 2014), pp. 275–283.Google Scholar
 K Aberer, Z Despotovic, in Proceedings of the 2001 ACM International Conference on Information and Knowledge Management (CIKM). Managing trust in a peer2peer information system (Atlanta, GA, USA, 2001), pp. 310–317.Google Scholar
 SD Kamvar, MT Schlosser, H GarciaMolina, in Proceedings of the 2003 ACM International Conference on World Wide Web (WWW). The eigentrust algorithm for reputation management in p2p networks (Budapest, Hungary, 2003), pp. 640–651.Google Scholar
 L Xiong, L Liu, Peertrust: Supporting reputationbased trust for peertopeer electronic communities. IEEE Trans. Knowl. Data Eng. 16(7), 843–857 (2004).View ArticleGoogle Scholar
 S Buchegger, JY Le Boudec, A Robust Reputation System for Mobile ad hoc Networks. Technical Report IC/2003/50, EPFLDIICA, (Lausanne, 2003).Google Scholar
 Q He, D Wu, P Khosla, in Proceedings of the IEEE Wireless Communications and Networking Conference (WCNC) 2004. Sori: a secure and objective reputation based incentive scheme for adhoc networks (Atlanta, GA, USA, 2004), pp. 825–830.Google Scholar
 WL Teacy, J Patel, NR Jennings, M Luck, Travos: Trust and reputation in the context of inaccurate information sources. Auton. Agent Multi Agent Syst. 12(2), 183–198 (2006).View ArticleGoogle Scholar
 A Jsang, R Ismail, in Proceedings of the Electronic Commerce Conference 2002. The beta reputation system (Bled, Slovenia, 2002), pp. 2502–2511.Google Scholar
 H Chan, VD Gligor, A Perrig, G Muralidharan, On the distribution and revocation of cryptographic keys in sensor networks. IEEE Trans. Dependable Secure Comput. 2(3), 233–247 (2005).View ArticleGoogle Scholar
 M Raya, MH Manshaei, M Felegyhazi, JP Hubaux, in Proceedings of ACM Conference on Computer and Communications Security (CCS) 2008. Revocation games in ephemeral networks (Alexandria, VA, USA, 2008), pp. 199–210.Google Scholar
 S Reidt, M Srivatsa, S Balfe, in Proceedings of the ACM Conference on Computer and Communications Security (CCS) 2009. The fable of the bees: incentivizing robust revocation decision making in ad hoc networks (Chicago, IL, USA, 2009), pp. 291–302.Google Scholar
 X Chen, H Patankar, S Zhu, M Srivatsa, J Opper, in Proceedings of the IEEE International Conference on Sensing, Communication and Networking (SECON) 2013. Zigzag: Partial mutual revocation based trust management in tactical ad hoc networks (New Orleans, LA, USA, 2013), pp. 131–139.Google Scholar
 S Kim, JM Chung, Message Complexity Analysis of Mobile Ad Hoc Network Address Autoconfiguration Protocols. IEEE Trans. Mobile Comput. 7(3), 358–371 (2008).View ArticleGoogle Scholar