A futuristic trust coefficientbased semiMarkov prediction model for mitigating selfish nodes in MANETs
 Janakiraman Sengathir^{1}Email author and
 Rajendiran Manoharan^{1}
https://doi.org/10.1186/s1363801503844
© Sengathir and Manoharan. 2015
Received: 2 March 2015
Accepted: 13 May 2015
Published: 7 June 2015
Abstract
In mobile ad hoc networks (MANETs), network survivability is considered as a potential factor required for maintaining maximum degree of connectivity among the mobile nodes even during failures and attacks. But, the selfish mobile nodes pose devastating influence towards network survivability. Hence, a prediction model that assesses network survivability through stochastic properties derived from nodes’ behaviour becomes essential. This paper proposes a futuristic trust coefficientbased semiMarkov prediction model (FTCSPM) that investigates and quantifies the impact of selfish behaviour towards the survivability of the network. This FTCSPM approach incorporates a non birthdeath process for manipulating futuristic trust coefficient since it does not consider the transition of a mobile node from the failed state to a selfish state into account. This semiMarkov prediction model also aids in framing a lower and upper bound for network survivability. Extensive simulations were carried out through ns2 and the results indicates that FTCSPM show better performance than the existing benchmark mitigation mechanisms like correlated node behaviour model (CNBM), probabilistic behavior model (PBM) and epidemic correlated node behavioural model (ECNBM) proposed for selfish nodes. Further, FTCSPM isolates the selfish nodes rapidly at the rate of 33 % than the considered benchmark systems. Furthermore, the validation of this prediction model performed through Weibull distribution has a high degree of correlation with the simulation results and thus assures the reliability and correctness of the proposed approach. In addition, this approach computes the mean transition time incurred by a mobile node to transit from cooperative to selfish mode as 6.49 s and also identifies the minimum and maximum selfish behaviour detection time as 140 and 180 s, respectively.
Keywords
1 Introduction
In mobile ad hoc networks (MANETs), network survivability is considered as an important entity for reliable data communication. But the dynamic change in mobile nodes’ behaviour pose a great challenge towards the survivability of the network since they do not possess a centralised infrastructure for communication [1]. Moreover, the transition of mobile nodes’ behavioural state from cooperation to selfish drastically affects the connectivity of the network [2]. Further, mobile ad hoc networks are highly prone to random failures and attacks due to their unique characteristics like limited energy availability, dynamic network infrastructure and errorprone communication link [3]. Thus, in practical, a mobile ad hoc network is said to be highly survivable when it establishes and maintains a collaborative environment among the mobile nodes [4].
From the recent past, researchers have analysed the survivability of the network in terms of network connectivity [5–7]. This network connectivity highly depends on the degree of cooperation attributed by the mobile nodes present in the routing path established between the source and the destination [8]. This degree of cooperation is drastically affected by means of selfish behaviour since selfish nodes deny forwarding either data or control packets of their neighbouring nodes for conserving its energy. In addition, the selfish behaviour of a mobile node prevents an individual node from participating in routing activity and affects the cooperation level rendered by the mobile nodes present in the routing path.
In the literature, various mitigation mechanisms proposed for selfish nodes either concentrates on the past history of mobile nodes or estimates conditional probabilistic factor based on both the past and present behaviour of the mobile node into account [9, 10]. However, these approaches lack to detect selfish nodes by forecasting the probability of transition between the behavioural states of the mobile nodes based on its present characteristics.
In this paper, we contribute a futuristic trust coefficientbased semiMarkov prediction model to forecast the likelihood ratio of a mobile node to turn into selfish based on stochastic properties derived from its present behaviour. This semiMarkov prediction model incorporates a non birthdeath Markov process, which is a special kind of Markov chain that does not emphasize on the restriction of nearest neighbour only transitions. Further, we assume that a failed mobile node cannot be converted into a selfish node. Hence, in this prediction model, a Markov chain is utilized for modeling the node behaviour transitions. Furthermore, this model considers three possible types of node behaviours viz., cooperative, selfish and failed. In addition, ad hoc ondemand distance vector (AODV) protocol has been modified with an additional metric called as expected level cooperation coefficient pertaining to the next hop neighbour in order to establish reliable route for data dissemination.
 1)
This paper investigates on semiMarkov prediction model to forecast the mobile nodes’ behaviour manipulated through the present state characteristics represented by means of futuristic trust coefficient.
 2)
The survivability of the mobile ad hoc network is estimated probabilistically, and the upper and lower thresholds for network connectivity are derived in this paper.
 3)
This paper provides a semiMarkov model that is validated using Weibull distribution since it is the predominant distribution used for evaluating lifetime in reliabilitybased engineering.
The remaining part of the paper is organized as follows. Section 2 presents indepth analysis about the existing selfish behaviour prediction models in the literature. Section 3 depicts the futuristic trust coefficientbased semiMarkov prediction model that aids in forecasting the change in mobile nodes’ behaviour towards selfishness. Section 4 briefly describes the influence of selfish nodes towards network survivability and model validation using Weibull distribution. Section 5 presents the details of the simulation environment and the comparative analysis performed with the benchmark systems and the proposed FTCSPM. Section 6 concludes the paper with future plan of work.
2 Related work
In the literature, considerable number of models which extensively analyses the survivability of mobile ad hoc networks in the presence of various types of malicious nodes is discussed. Xing and Wang [11] presented an analytical approach to evaluate network survivability in the presence of misbehaving nodes and failure nodes. In their work, authors investigated a novel mechanism based on semiMarkov process that categorizes mobile nodes’ behaviour into four types viz., i) cooperative node, ii) failure node, iii) selfish node and iv) malicious node. Authors also incorporated a special type of counter known as the Nuglet counter to estimate the probability of node misbehaviour at any instant of time. This approach also aids in deriving the closed form approximation of network survivability based on network characteristic functionalities and node behaviour distribution.
Further, Cardenas et al. [12] recommended a sequential probability ratio test (SPRT) detection scheme which incorporates a quantitative approach called DOMINO for mitigating selfish behaviour in an ad hoc network. This work adopts a dual step process for detecting selfish behaviour of a mobile node. In the first step, the transition probability of mobile nodes which are currently under direct communication was estimated. In the second step, Markov chain was derived for each mobile node with transition probability p and 1p. Finally, the identification of misbehaving mobile nodes was carried out by monitoring the state changes through Markov chain. Vallam et al. [13] incorporated a discrete time Markov chain (DTMC) with steadystate probabilities for analysing non saturated node. This mechanism addresses the problem of backoff manipulation that arises due the presence of misbehaving nodes. In this work, authors also incorporated a sequential probability ratio test for estimating the mean sample size of attack.
Hernandezorallo et al. [14] proposed a collaborative watchdog mechanism for detecting selfish nodes in an ad hoc scenario. In this paper, authors incorporated Poisson distribution for analysing cooperativity among the mobile nodes which defines two states for collaborative nodes viz., NOINFO and POSITIVE states. A NOINFO state of collaborative node is the state in which the particular node does not realize its neighbours’ selfish behaviour, while in case of POSITIVE state, an individual mobile node acknowledges the neighbouring selfish nodes. Authors have incorporated continuous time Markov chain (CTMC) for modeling the network, and the network is evaluated collaboratively using watchdog. This model also aids in determining the time and cost incurred in the detection of selfish nodes with the aid of watchdog mechanism.
Xing in [15] further contributed a novel approach based on generic semiMarkov model for characterizing nodes’ behaviours and their transitions. This approach also adapts stochastic properties for framing transient and limited probabilities for mobile nodes which aids in estimating the negative impact of nodes’ misbehaviour and the node failures. Lei Guang et al. [16] contributed a novel approach known as probabilistic random back off which is capable of mitigating selfish nodes. This mechanism mitigates the selfish nodes which specifically drops others nodes’ packets without forwarding it to the next hop neighbour in the routing path. This model incorporates 3D Markov chain for analysing the characterization behaviour of mobile nodes towards network survivability.
Furthermore, the benchmark systems used for comparing the proposed FTCSPM approach are discussed below.
Komathy and Narayanasamy [17] developed a probabilistic behaviour model (PBM) based on nodes’ residual energy for synchronizing collaborative mobile nodes. This approach measures the tradeoff that exists between the energy parameters (residual energy and energy utilization) and network parameters (packet delivery ratio and average endend delay). This probabilistic model also predicts the expected node behaviour which highly depends upon the packet delivery and drop ratio. For this prediction, they incorporated a dynamic memory table called neighbour table which regularly estimates the forwarding rates of neighbour nodes. Finally, the network reliability is manipulated based on the equilibrium value of a mobile node that quantifies the influence of other neighbouring mobile nodes towards them by means of packet delivery rate and endend delay.
Azni et al. [18] contributed a correlated node behavior model (CNBM) based on continuous time semiMarkov process for clustering the mobile nodes. This model characterizes the node behaviour based on probability of selfishness, probability of forwarding packets, probability of dropped packets and probability of recovery. Further, authors contributed [19] an epidemic correlated node behavioral model (ECNBM) to characterize the behaviour of the mobile nodes participating in the routing path. The ECNBM approach isolates the malicious nodes from the ad hoc environment in two phases, the first phase characterizes the behaviour of the mobile node and its dynamic transition in behaviour states based on semiMarkov process. The second stage derives the predicted correlation degree of a mobile node from its current state behavioural characteristics.
Further, PBM, CNBM and ECNBM are mainly considered for comparison since they are proven as the significant and primary semiMarkovbased correlated node behaviour models for efficient and effective isolation of selfish nodes. In addition, PBM, CNBM and ECNBM detect selfish nodes rapidly and improve the survivability of the network by 33 % than other existing works available in the literature.
2.1 Extract of the literature
 a)
A semiMarkovbased selfish behaviour forecasting model that utilizes non birthdeath process has not been explored to the best of our knowledge.
 b)
A forecasting mechanism that incorporates the futuristic trust coefficient for mitigating selfishness in order to enable high survivability in an ad hoc network has not been investigated.
Hence, these limitations motivated us to devise a futuristic trust coefficientbased semiMarkov prediction model for isolating selfish behaviour of nodes.
3 Futuristic trust coefficientbased semiMarkov prediction model
Thus, “Given a wireless ad hoc network ‘N’ with possible definitions of mobile node behaviours ‘B’, the problem can be formulated as a network survivability model ‘M (N,B)’ that estimates and isolates the selfish nodes from the routing path through futuristic trust coefficient which quantifies the likelihood probability incurred by a mobile node to get transited into the noncooperative state”.
 a)
Estimation of stochastic properties from mobile nodes.
 b)
Estimation of model parameters
 c)
Node behaviour modelingbased transition probability matrix
 d)
Manipulation of futuristic trust coefficientbased on non birthdeath semiMarkov process.
 e)
Isolation of selfish nodes based on futuristic trust coefficient.
3.1 Estimation of stochastic properties from mobile nodes
3.2 Estimation of stochastic probabilities

λ _{cs}—stochastic probability for a cooperative mobile node to turn into selfish node

μ _{sc}—stochastic probability for a selfish node to rehabilitate into cooperative mobile node

μ _{sf}—stochastic probability for a cooperative mobile node or a selfish node to turn into failure node.

μ _{fc}—stochastic probability for a failure node to rehabilitate into cooperative mobile node
where E _{ R } is the residual energy possessed by a mobile node and E _{ N } is the energy required for transmitting maximum of ‘N’ number of packets.
where τ is the selfish threshold parameter and L _{ T } is the mean lifetime of the mobile node.
Hence, a mobile node is considered to be highly cooperative when the value ‘λ _{cs}’ is very less, and it also infers that the particular node has very high mean life time.
Further, this stochastic probability is computed from the direct trust value identified between the neighbouring nodes of a routing path based on mutual packet forwarding process.
This direct thrust value identified gives a clear picture on the level of cooperation extended by a mobile μsc node to their neighbour. Thus, a mobile node in selfish mode gets into cooperative when the value of μsc reaches above a threshold as defined in [21].
where ‘d’ and ‘k’ denote the average distance of communication between mobile nodes and mobility proportion variable, respectively.
Therefore, a mobile node can reconfigure and enter into the network when the mean reconfiguration time of the network is minimum.
3.3 Node behaviour modeling through transition probability matrix
In the above representation, it is assumed that the stochastic probability of a node to convert from selfish to failure is equivalent to the stochastic probability of a cooperative node for turning into failure as defined in [23].
This transition matrix provides a snapshot about the possible behavioural transitions of a mobile node during the routing activity. The entry ‘0’ in the transition matrix emphasizes that this model does not consider the transition of a mobile node from the failed state to a selfish state into account.
3.4 Manipulation of futuristic trust coefficientbased on non birthdeath semiMarkov process
In FTCSPM, the stochastic probabilities derived are designated into failure time distributions (λ _{cs,} λ _{sf}) and repair time distributions (μ _{sc,} μ _{fc}). Further, the failure time distributions are assumed to be exponential but the repair time distributions can be divided into two phases viz., i) rehabilitating selfish nodes to cooperative nodes and ii) rehabilitating failure nodes to cooperative nodes.
These two repair phases are exponentially distributed with the expectation coefficient of \( \raisebox{1ex}{$1$}\!\left/ \!\raisebox{1ex}{${\mu}_{sc}$}\right. \) and \( \raisebox{1ex}{$1$}\!\left/ \!\raisebox{1ex}{${\mu}_{fc}$}\right. \), respectively. Furthermore, in this non birthdeath process, the cumulative sojourn period during failure is two staged and hypoexponential distribution, the FTCSPM approach is a semiMarkov prediction process rather than a homogeneous continuous time Markov chain. In this model, the steady state probability vectors for detection are considered as (π _{ c,} π _{ S,} π _{ F }).
In this vector, π _{ c } indicates the steady state probability of a mobile node during cooperation, while π _{ S } indicates the steady state probability of a mobile node during selfishness (futuristic trust coefficient) and π _{ F } the steady state probability of a mobile node under failure.
From the derived vector, the Futuristic trust coefficient π _{ S } aids in quantifying the degree of selfish behaviour attributed by a mobile node.
3.5 Isolation of selfish nodes based on futuristic trust coefficient
In this model, the mobile nodes are identified as selfish when the value of π _{ S } reaches below the futuristic trust threshold of 0.6 as defined in [24]. Then, the FTCSPM approach isolates the selfish nodes based on the value of futuristic trust coefficient π _{ S } possessed by each of the mobile node for enabling reliable routing path.
4 Validation and network survivability analysis of FTCSPM Model
Moreover, the mean transition time incurred in the conversion of a cooperative mobile node into a selfish node is identified as 6.49 s. Hence, we choose a = 2 and β ^{ S } = 6 from simulation.
In this approach, survivability of networks in the presence of selfish mobile nodes is quantified through selfish survivability factor which is defined as the ratio of observed number of mobile nodes infected through selfishness to the expected number of mobile nodes actually infected. Further, the selfish survivability factor infers that FTCSPM in an average withstands the degree of selfishness up to a maximum of about 39 % when compared to CNBM, PBM and ECNBM.
5 Simulation results and analysis
Simulation configuration
Parameter  Value 

NS version  2.32 
Number of nodes  100 
Protocol used  AODV 
Mac layer  802.11 
Terrain area  1000 × 1000 m^{2} 
Mobility model  Random way point 
Simulation time  250 s 
Traffic model  CBR (40 packets/s) 
Packet size  512 bytes 
Type of antenna  Antenna/omni antenna 
Type of propagation  Twoway ground 
Channel capacity  2 Mpbs 
Further, the main objective of these simulation experiments is to evaluate the correctness of the proposed FTCSPM approach. Hence, the efficiency of the proposed FTCSPM is proved not only through comparative analysis with the existing algorithms viz., CNBM, PBM and ECNBM, but also by incorporating a similar network configuration for implementing FTCSPM and the three benchmark mitigation mechanisms.
Furthermore, in simulation experiments, the change in behaviours of a mobile node according to the strategies defined in Section 3 is analysed. The routing protocol is chosen here is AODV, which is implemented to route the data packets using cooperative nodes of the network environment, while for routing data packets to the destination in the presence of selfish nodes is carried out through modified version of AODV so that the selfish mobile nodes are isolated without affecting the standard network activities.
In addition, the connectivity of an ad hoc network purely depends on the cooperation established by the intermediate nodes towards the establishment of a reliable routing path [25–29]. However, the connectivity is drastically influenced by the presence of selfish nodes, since these node decreases the packet delivery ratio and throughput while increasing the energy consumption rate, packet drop rate and average endtoend delay [30, 31]. Hence, the performance of the proposed FTCSPM is evaluated based on performance evaluation parameters viz., packet delivery ratio, energy consumption, average endtoend delay, packet drop rate and throughput.
Packet delivery ratio (PDR):
It is defined as the ratio between the number of packets received by the destination node and the number of packets sent to the destination node.
Energy consumption rate:
It is defined as the total energy consumed by a mobile node during the state of transmission, reception, idle and sleep.
Average endtoend delay:
It is defined as the average time taken by the data packets to reach its destination including the time taken for connection establishment and queuing delay.
Packet drop rate:
It is defined as the total number of data packets lost during the data transmission from source to the destination.
Throughput:
It is defined as the total number of packets successfully delivered to the destination in a specified period.
5.1 Performance evaluation of FTCSPM by varying the number of mobile nodes (experiment 1)
In general, the packet delivery ratio of a network decreases when there is an increase in the number of transmitting mobile nodes. This decrease in PDR value is mainly due to inadequate availability of bandwidth caused by enormous data generation. The plots depicted in Fig. 3a shows the packet delivery ratio exhibited by FTCSPM and the three benchmark mitigation mechanisms. However, the proposed mechanism FTCSPM shows considerable improvement in packet delivery rate when compared to the other three benchmark mechanisms. FTCSPM also exhibits an increase in PDR from 8 to 14 % than ECNBM, from 16 to 22 % than CNBM and from 19 to 25 % than PBM. On the whole, the proposed FTCSPM shows a significant improvement in packet delivery ratio by 14.2 %.
Similarly, the plots depicted in Fig. 3b shows the energy consumption rate of FTCSPM with the three benchmark systems. The energy consumption rate considerably increases when the number of mobile nodes participating in the transmission increases due to enormous amount of data flow. But the proposed FTCSPM shows significant improvement in performance by consuming the minimum energy even when enormous amount of data is transmitted. This is due to the effectiveness of the FTCSPM approach in predicting the change in behaviour of the mobile nodes during the period of transmission. FTCSPM also shows a decrease of 0.7 to 1.1 J of energy consumption from ECNBM, from 0.9 to 1.2 J from CNBM and from 1.8 to 2.2 J from PBM. In an average, FTCSPM shows considerable decrease in energy consumption rate of 2.34 J when compared to the other three selfish mitigation mechanisms.
Further, Fig. 3c depicts the plots of average endtoend delay for the given number of mobile nodes of the four mitigation mechanisms viz., FTCSPM, CNBM, PBM and ECNBM. The proposed method FTCSPM shows a significant decrease in average endtoend delay, since FTCSPM incorporates an effective and efficient selfishness isolation approach for enabling reliable routing path for data transmission. FTCSPM shows a decrease of 9 to 12 % in average endtoend delay from ECNBM, from 14 to 19 % from CNBM and from 16 to 22 % from PBM. FTCSPM also decreases the average endtoend delay factor by 37 % (mean value), when compared to the other three benchmark mechanisms.
Furthermore, Fig. 3d depicts the plots of packet drop rate derived for a given number of mobile nodes from all the four selfish mitigation mechanisms. The packet drop rate increases when the number of mobile nodes increases since large number of data packets has been transmitted. However, the proposed FTCSPM approach considerably decreases the packet drop rate due its inbuilt trustbased prediction mechanism for analysing change in behaviour of mobile nodes. Further, FTCSPM shows a decrease of 10 to 15 % in packet drop rate than ECNBM, from 15 to 22 % than CNBM and from 21 to 29 % than PBM. FTCSPM in an average decreases the packet drop rate by 28 % than the other three mitigation mechanisms.
Finally, Fig. 3e depicts the plots of throughput derived for a given number of mobile nodes from all four selfish mitigation mechanisms. The throughput decreases when the number of selfish nodes increases since increase in packet drop rate reduces the throughput. However, the proposed FTCSPM approach considerably increases the throughput of the network when compared to ECNBM, CNBM and PBM. Further, FTCSPM shows increase of 9 to 14 % in throughput from ECNBM, from 12 to 16 % than CNBM and from 14 to 19 % than PBM. Furthermore, FTCSPM increases the throughput in an average value by 23 % than the other three selfish mitigation mechanisms used for benchmarking.
5.2 Performance evaluation of FTCSPM by varying the number of selfish nodes (experiment 2)
The plots depicted in the Fig. 4a shows that there is decrease in packet delivery ratio exhibited by all of the above said four mitigation mechanisms when the number of selfish nodes increases in the ad hoc environment. However, the proposed mechanism FTCSPM shows a considerable improvement in packet delivery rate when compared to the other three benchmark mechanisms. FTCSPM also shows an increase in PDR from 9 to 14 % than ECNBM, from 12 to 16 % than CNBM and from 17 to 21 % than PBM. On the whole, FTCSPM shows a significant improvement in packet delivery ratio by 24.4 %.
Similarly, the plots depicted in Fig. 4b shows energy consumption rate for the given number of mobile nodes in order to deliver the data packets reliably. The energy conservation factor considerably increases when the number of selfish mobile nodes increases in the ad hoc environment. But the proposed FTCSPM shows significant improvement in performance by conserving minimum energy. Further, FTCSPM shows decrease of 1.4 to 1.9 J of energy consumption from ECNBM, from 1.9 to 2.7 J from CNBM and from 2.4 to 3.8 J from PBM. In an average, FTCSPM shows considerable decrease in energy consumption rate of 5.67 J when compared to the other three selfish mitigation mechanisms.
Figure 4c depicts the plots of average endtoend delay for the given number of selfish nodes from the four mitigation mechanisms viz., FTCSPM, CNBM, PBM and ECNBM. The proposed FTCSPM shows the decrease of 8 to 15 % in average endtoend delay from ECNBM, from 18 to 22 % from CNBM and from 23 to 28 % from PBM. Furthermore, FTCSPM decreases the average endtoend delay factor by 25.54 % (mean value), when compared to all the other three mitigation mechanisms.
Figure 4d depicts the plots of packets drop rate considered for comparative analysis for the four selfish mitigation mechanisms. The packet drop rate increases when the number of selfish nodes increases since it does not forward or cooperate with the other mobile nodes present in the routing path. However, the proposed FTCSPM approach considerably decreases the packet drop rate from 10–17 % than ECNBM, from 15–25 % than CNBM and from 21–27 % than PBM. FTCSPM also decreases the packet drop rate in an average by 24.45 % than the other three mitigation mechanisms.
In addition, Fig. 4e depicts the plots for thought derived from all the four selfish mitigation mechanisms. The throughput decreases when the number of selfish nodes increases. However, the proposed FTCSPM approach considerably increases the throughput of the network when compared to ECNBM, CNBM and PBM. Further, FTCSPM shows increase of 11–14 % in packet drop rate from ECNBM, from 14–19 % than CNBM and from 16 to 20 % than PBM. FTCSPM in an average increases the throughput by 19.18 % than the other three selfish mitigation mechanisms.
5.3 Performance evaluation of FTCSPM by varying the CBR traffic flows (experiment 3)
Figure 5a shows that there is a decrease in packet delivery ratio when the CBR traffic flow is varied from low level to high level since heavy packet loss occurs due to network congestion. However, the proposed mechanism FTCSPM shows a considerable improvement in packet delivery ratio than the other three benchmark mechanisms. FTCSPM also shows an improvement in PDR from 16 to 22 % than ECNBM, from 21 to 26 % than CNBM and from 24 to 31 % than PBM. On the whole, FTCSPM shows a considerable improvement in packet delivery ratio by 21.3 %.
Similarly, Fig. 5b shows energy consumption rate for the varying traffic flows. The energy consumption rate considerably increases when the CBR traffic flow rate linearly increases. But the proposed FTCSPM shows significant improvement in performance by consuming minimum energy even during the event of high traffic flows. FTCSPM also shows a decrease of 2.1 to 2.5 J of energy consumption than ECNBM, from 2.3 to 2.8 J than CNBM and from 2.9 to 3.3 J than PBM. In an average, FTCSPM considerably decreases the energy consumption rate by 3.45 J when compared to the other three selfish mitigation mechanisms.
Figure 5c depicts the plots of average endtoend delay for FTCSPM and the considered benchmark systems obtained by varying the CBR traffic flows. FTCSPM decreases the average endtoend delay of 11 to 14 % than ECNBM, from 15 to 21 % than CNBM and from 17 to 25 % than PBM. Furthermore, FTCSPM decreases the average endtoend delay factor by 21.54 % (mean value), when compared to all the other three mitigation mechanisms.
Figure 5d depicts the plots of packet drop rate obtained for comparative analysis of the four selfish mitigation mechanisms derived by varying the number of CBR traffic flows. The packet drop rate increases when the CBR traffic flow increases from lower intensity to higher intensity. However, the proposed FTCSPM approach considerably decreases the packet drop rate from 14 to 19 % than ECNBM, from 18 to 22 % than CNBM and from 23 to 29 % than PBM. FTCSPM also decreases the packet drop rate in an average by 20.15 % than the other three benchmark mitigation mechanisms.
In addition, Fig. 5e depicts the comparative analysis chart for throughput derived from all the four selfish mitigation mechanisms. The throughput decreases when the number of selfish nodes increases even when the CBR traffic flow increases. However, FTCSPM approach considerably increases the throughput of the network when compared to ECNBM, CNBM and PBM. Further, FTCSPM shows an improvement of 14 to 17 % in throughput than ECNBM, from 18 to 23 % than CNBM and from 23 to 28 % than PBM. FTCSPM in an average increases the throughput by 22.81 % than the other three selfish mitigation mechanisms.
Finally, FTCSPM is also valuated by varying the number of mobile nodes, number of selfish nodes and traffic flow rate based on control overhead and total overhead. FTCSPM in an average decreases the control overhead by 23, 29 and 32 % than the compared ECNBM, CNBM and PBM approaches, respectively. Similarly, FTCSPM in an average decreases the total overhead by 19, 25 and 29 % than ECNBM, CNBM and PBM approaches, respectively.
6 Conclusions
This paper has presented a futuristic trust coefficientbased semiMarkov prediction model formulated through non birthdeath process for mitigating selfish mobile nodes in an ad hoc network. The performance of FTCSPM is analysed based on packet delivery ratio, average endtoend delay, energy consumption rate, packet drop rate and throughput. The experimental results illustrated in this paper show the proximity performance of the FTCSPM approach with increase in packet delivery ratio and throughput. Further, our ns2based simulation study makes it evident that FTCSPM in an average improves the packet delivery ratio and throughput by 14.2 and 23 %, respectively, and at the same time, FTCSPM reduces the energy consumption rate, packet drop rate and average endtoend delay by 18.4, 37 and 28 %, when compared to the benchmark systems like CNBM, PBM and ECNBM. Further, the influence of selfish nodes towards the survivability of a network is also derived. Furthermore, the proposed model is also validated by means of Weibull distribution, and this validation clearly portrays that Weibull distribution has a high degree of correlation with the simulation results. In addition, FTCSPM detects and isolates the selfish nodes rapidly at the rate of 33 % than the considered benchmark systems.
Finally, as a part of our future work, we have been planning to devise a semiMarkov prediction model based on pure birthdeath process that relies only on neighbouronly transitions and a limited availability factorbased semiMarkov prediction model that quantises the degree of participation rendered by the mobile nodes in the routing process.
Declarations
Authors’ Affiliations
References
 F Xing, W Wang, On the survivability of wireless ad hoc networks with node misbehaviors and failures. IEEE Transactions On Dependable And Secure Computing 7(3), 284–299 (2010)View ArticleGoogle Scholar
 A Md, AK Akhtar, G Sahoo, Mathematical model for the detection of selfish nodes in MANETs. International Journal of Computer science and Informatics 1(3), 25–28 (2008)Google Scholar
 S Buchegger, JY Boudec, Nodes bearing Grudges: Towards routing security, Fairness and Robustness in Mobile AdHoc Network (presented at tenth Eurominicro workshop on Parallel, Distributed and Network based Processing (Canary Islands, Spain, 2002)Google Scholar
 S Marti, TJ Giuli, K Lai, M Baker, Mitigating routing misbehavior in mobile ad hoc networks. Mobile Computing and Networking 1(1), 255–265 (2000)Google Scholar
 SK Hwang, DS Kim et al., Markov model of link connectivity in mobile ad hoc networks. Telecommunication Systems 34(1–2), 51–58 (2006)Google Scholar
 L Guang, M Chadi, A Benslimane, Enhancing IEEE 802.11 random backoff in selfish environments. IEEE Transactions On Vehicular Technology 57(3), 125–132 (2008)Google Scholar
 G Corradi, J Janssen, R Manca, Numerical treatment of homogenous semiMarkov processes in transient case—a straightforward approach. Methodology and Computing in Applied Probability 6(1), 233–246 (2004)MATHMathSciNetView ArticleGoogle Scholar
 T Sundarajan, A Shanmugam, Modeling the behavior of selfish forwarding nodes to simulate cooperation in MANET. International Journal of Network Security and Its Application 2(2), 147–160 (2010)View ArticleGoogle Scholar
 J Sengathir, R Manoharan, A reliability factor based mathematical model for isolating selfishness in MANETs’. International Journal of Information and Communication Technology 6(3/4), 403–421 (2014)View ArticleGoogle Scholar
 J Sengathir, R Manoharan, Selfish conscious mathematical model based on reliable conditional survivability coefficient in MANET routing, in 3rd Third International Conference on Advances in Information Technology and Mobile Communication (AIM 2013). Bangalore, India, April, Proceeding published by Elsevier 1(1), 31–40 (2013)Google Scholar
 F Xing, W Wang, Modeling and analysis of connectivity in mobile ad hoc networks with misbehaving nodes (Paper presented at the IEEE international conference on communications, IEEE, Istanbul, June, 2006)Google Scholar
 A Alvaro, Cardenas, R Svetlana, S John Baras, Evaluation of detection algorithms for MAC layer misbehavior. Theory and Experiments. IEEE Transactions on Networking 17(2), 605–617 (2009)Google Scholar
 Rohith Dwarakanath Vallam, Antony Franklin A, Siva RamMurthy C, Modelling cooperative MAC layer misbehaviour in IEEE 802.11 ad hoc networks with heterogeneous loads. Paper presented at the 6th international symposium on modeling and optimization in mobile, ad hoc, and wireless networks and workshops, IEEE, WIOPT Berlin, Germany, 1–3 April 2008.Google Scholar
 E HernandezOrallo, MD Serraty, JC Cano, T Calafate, P Manzoni, Improving selfish node detection in MANETs using a collaborative watchdog. IEEE Letters 16(5), 642–645 (2012)View ArticleGoogle Scholar
 X Fie, Modeling, Design, and Analysis on the Resilience of Largescale Wireless MultiHop Networks (North Carolina State University, Raleigh (North Carolina, USA, Department Of Engineering, 2009)Google Scholar
 L Guang, M Chadi, A Benslimane, Enhancing IEEE 802.11 random backoff in selfish environments. IEEE Transactions on Vehicular Technology 57(3), 1806–1822 (2008)View ArticleGoogle Scholar
 K Komathy, P Narayanasamy, A probabilistic behavioral model for selfish neighbors in a wireless ad hoc network. International Journal of Computer Science and Network Security 7(7), 77–82 (2007)Google Scholar
 AH Azni, A Rabiah, N Zul, M Azri, B Abd SamadHasan, H Burairah, Correlated node behavior model based on semi Markov process for MANETS. International Journal of Computer Science 9(1), 25–32 (2013)Google Scholar
 AH Azni, A Rabiah, N Zul, M Azri, B Abd SamadHasan, H Burairah, Epidermic modelling for correlated node behavior in ad hoc networks. International Journal of Chaotic Computing 1(1), 22–30 (2013)Google Scholar
 AP Patil, KDK Rajani, S Bathey, MP Dinesh Kumar, J Malavika, Design of energy efficient routing protocol for MANETs based on AODV. IJCSI 8(1), 215–220 (2013)Google Scholar
 Srinivasan V, Nuggehali, Chiasserini, C.F., Rao, R.R., (2003). Stimulating Cooperation in Wireless ad hoc networks. Presented at 22nd Annual Joint Conference of the IEEE Computer and Communication Societies (INFOCOM’03), 2003Google Scholar
 N Sadagopan, F Bai, B Krihnamachari, A Helmy, PATHS: analysis of PATH duration statistics and their impact on reactive MANET routing protocols. In prod. ACM MobiHoc 1(1), 299–307 (2003)Google Scholar
 K Paul, RR Choudhuri, Bandyopadhyay, Survivability analysis of ad hoc wireless network architecture. In prod IFIPTC6/ European Commission, International Workshop Mobile and Wireless Communication Networks 1(1), 31–46 (2000)Google Scholar
 P Manohar, M Vereshechaka, D Manjanth, Survivability, analysis under nonuniform stochastically dependent node damages. National Conference on Communications 1(1), 1–5 (2010)Google Scholar
 Y Xu, W Wang, Characterizing the spread of correlated failures in larger wireless networks. In proc. IEEE INFOCOM 56(11), 1–9 (2010)Google Scholar
 Z Zhang et al., ECGcryptography and authentication in body area networks. IEEE Transactions on Information Technology in Biomedicine 16(6), 1070–1078 (2012)View ArticleGoogle Scholar
 Z Shen et al., Peertopeer media streaming: insights and new developments. Proceedings of the IEEE 99(12), 2089–2109 (2011)View ArticleGoogle Scholar
 A Attar et al., A survey of security challenges in cognitive radio networks: solutions and future research directions. Proceedings of the IEEE 100(12), 3172–3186 (2012)View ArticleGoogle Scholar
 Z Wan et al., Adaptive unequal protection for wireless video transmission over IEEE 802.11e networks. Multimedia Tools and Applications 1(1), 345–352 (2013)Google Scholar
 D He et al., Retrust: attackresistant and lightweight trust management for medical sensor networks. IEEE Transactions on Information Technology in Biomedicine 16(4), 623–632 (2012)View ArticleGoogle Scholar
 J Zhou et al., Securing mhealthcare social networks: challenges, countermeasures and future directions. IEEE Wireless Communications 20(4), 123–134 (2013)Google Scholar
Copyright
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.