Jul 11, 2012 - group formed among nodes having different goals. They share their resources to either provide or use network services in the network to ...
Evaluating network layer selfish behavior and a method to detect and mitigate its effect in MANETs M.A.Hussain, A.Nadeem, Obaid Khan, S.Iqbal, A.Salam Department of Computer Science Federal Urdu University of Arts, Science & Technology Abstract—A Mobile Ad Hoc Network (MANET) is an infrastructureless network of autonomous devices formed dynamically. At network layer routing protocols in MANETs expect every node to cooperate in routing and data forwarding tasks. However, this cooperation leads towards a state in which each node consumes it scarce resources for example battery power. Resource constraints of MANET nodes can derive dynamic selfish behavior in terms of not forwarding packets & not cooperating in routing operations on and off which can lead to degradation in network performance. In this paper, we first perform a simulation based study to evaluate the effect of selfish nodes on network performance. Then in contrast with most of the selfish node detection approaches in literature which are based on fully watchdog methods using promiscuous monitoring, we propose a proactive mechanism that first detects nodes that are exhibiting the selfish behavior based on the factors which derives selfish behavior and then mitigate its effect on network performance. Initial simulations results of a case study show that our methods detects & mitigate the effect of selfish behavior with an affordable overhead on the network. Keywords: constraints
I.
MANET, selfish behavior of nodes, battery
INTRODUCTION
A Mobile Ad Hoc Network (MANET) is an infrastructureless network of autonomous devices formed dynamically. This desirable feature of MANETs allows various applications of this networking technology. MANETs can be categorized into two types [1]: closed & open. Closed MANETs are goal specific like MANET deployed for surveillance and monitoring of specific area, carrying rescue operations after any disaster where fixed infrastructure is not available, military operations. In this type, each node is bound to take part in networking operations e.g. route discovery and forwarding packets for other nodes without considering their own resources spending upon performing such tasks. In contrast to closed group, open group formed among nodes having different goals. They share their resources to either provide or use network services in the network to ensure global connectivity. The cooperation among nodes in open type of MANET resembles the social groups in society where members of different social groups facilitate members of other groups in order to share different services. In social groups each member cooperates with others by believing that other members of the groups are also cooperating thus forming an stable society. However, there is a possibility that for some reasons, any member of the group does not cooperate but keep getting the benefit of services provided by the other groups. This action of a member not only degrades the performance but could also lead to unstableness of the society.
Due to their inherit characteristics such as no centralized control and limited energy resources, MANETs are vulnerable to various active & passive attacks. In addition, most of the routing protocols proposed for MANET operate on the assumption that all the nodes must cooperate in routing operations such as route discovery & route maintenance and packet forwarding process. This is to ensure desirable network performance in terms of throughput, PDR (Packet Delivery Ratio) and efficient resources consumption. In MANETs nodes are required to cooperate with other nodes in forwarding packets and sharing services in the network. In MANET, nodes are heterogeneous e.g. PDAs, laptops and cell phones and belongs to different individuals gathered in same geographical area for some reasons. These nodes communicate each other in such a way that an individual node altruistically spends its scarce resources for serving other nodes. However, this cooperation leads towards a state in which each node consumes its scarce resources for example battery power which is considered to be most important in a mobile environment. Limited energy resource could motivate nodes to avoid taking parts in network services for other nodes for example the node does not process or forward packets for other nodes in the network while still enjoying network service for himself. The nodes exhibiting such a behavior consider as selfish nodes [2]. Selfish nodes is different from malicious nodes as the main intention of malicious node is to collapse the entire network by applying different kind of attacks on the network such as black hole attack or gray hole attack [3][10] whereas selfish nodes misbehaving only to save it's resource. Selfish behavior further categorized in two types: a) Selfish by nature and b) Selfish due to resource constraints. A node which is selfish by nature always exhibits selfish behavior regardless of its residual energy or other resources thus save its energy. A selfish node behave selfishly in different manners e.g. falsely reporting about routes, change sequence ID in packet to avoid being part of the route, while still enjoying the network services. Performance of the network degrades in the presence of selfish nodes as we will show in this paper. Therefore, researcher has focused on dealing with selfish nodes MANETs. Many algorithms have been proposed by different researchers for detection of selfish nodes and provide solutions to mitigate their effects on overall network performance. Most of these algorithms rely upon watchdog mechanism where every node requires monitoring its neighbor’s behavior all the time for detection of nodes which behave selfishly. Once a selfish node is detected, it is required to respond. Based on the way the algorithm respond to selfish node the proposed mechanism can be classified into two main categories [4]. First,
detection & exclusion (i.e. detect selfish nodes and isolate them from network) and second motivational (i.e. detect selfish nodes and motivate them to cooperate in network functions). In this paper, we focus on selfish behavior in which a properly behaving node exhibit selfish behavior due to its low energy resources. We first analyze the impact of selfish behavior on MANETs performance through a simulation based study and then propose a selfish node detection protocol to mitigate the effect of selfish behavior of nodes in the network. Rest of the paper is arranged as follows: In section II, we review the related work from the literature. In section III, we present a selfish node detection protocol (SNDP). In section IV, first present the impact of selfish behavior on network performance through simulation based study and then presents the initial evaluation of selfish node detection protocol. Finally, in section V we summarize our results and future work.
II.
RELATED WORK
Researcher has suggested fully watchdog mechanisms in literature. For example, in [5], Wang et al. proposed a selfish node detection protocol based on watchdog mechanism in which a node only monitors its neighbor behavior when it forwards data to its neighbor node by working in promiscuous mode. All the other neighbor nodes are still in sleeping mode and do not monitor the behavior of such node. During the monitoring period, if a node detect a suspicious behavior of its neighbor, it request all other neighbors nodes to monitor this node a predetermine time period. Depending upon the statistics collected from the neighbors during this time period, a node is marked as selfish and as a punishment, they suggest isolating it from rest of the network for a predefined time. In addition, authors in [5] also proposed that when a node reaches low battery, it explicitly informs its neighbors and withdraw from the packet forwarding service. In contrast to this explicit reporting of low battery from nodes, in this paper we propose an algorithm that monitors the nodes with low battery for a possible selfish behavior transparent to the node. But it will keep using network services with some limitation. In [6], D. Djenouri et al. proposed an approach, which he named two-hop ACK. In this protocol, a central security mechanism is necessary which keeps record of node’s public key and private key. When a node forwards a packet to next node in the route, in order to confirm successful delivery of the packet and detection of selfish node, node generates a random number and encrypts it with public key of node which is two hops away in the route and encapsulates it in packet header along with own address. When the two hops away node received this packet, it get this number by decrypting the packet and send a special acknowledgement packet to its two hop back node. Acknowledgement is authenticated by node’s public key and some encryption method. Thus no node is required to continuously monitor their neighbors for selfish behavior detection hence this algorithm supports power aware routing protocol. If packet delivery is not acknowledged by two hops away node, its predecessor node is accused as selfish by packet forwarder node. Author supposed a temper proof hardware and acknowledgement is implicitly defined at MAC layer to prevent a cooperative node from false accusation.
In [7], a trust based watchdog mechanism named Packet Conservation Monitoring Algorithm (PCMA) is proposed in which information about a suspected node is collected from only those neighbor nodes which directly involve in sending or receiving packets from this node to check the its reliability. Each node maintains a table of directly send/received packets for each of its neighbor node and each node compare its statistics with its neighbor table. If difference found, the predecessor node marked its successor as selfish. A distributed reputation based technique is proposed in [8] in which each node maintains a reputation table for its neighbor nodes and reputation index is increased when a packet is correctly delivered whereas reputation decreased upon non-delivery of packet. Successful packet delivery is confirmed by using TCP from destination. However, due to packet collision and interference, a cooperative node may be falsely accused as selfish thus isolated from network which could results in a decrease in network performance as suggested in [9]. Considering this, authors in [9] proposed a distributed adaptive reputation based mechanism to detect selfish node in the network. Collaborative Watchdog algorithm is proposed in [11], a cooperative approach in which detection of selfish node relies on watchdog mechanism and each node maintain a reputation table about other nodes. Upon detection of selfish node, a collaborative node has a positive and this information disseminate to other nodes upon contact. A node with bad reputation is marked selfish cooperatively and isolated from the network. In [12], author proposed concept of virtual currency that a node can earn upon forwarding packets for others that it will pay to other nodes when it required network services from other nodes. In order to get network service, each node required this currency which motivates them to take part in providing services to other. Author present two models, the Packet Purse Model and Packet Trade Model. In Packet Purse Model, the price of forwarding service is attached to the packet which each intermediate node collected upon forwarding packet to next node. However, as indicated by the author, the node which initiate packet has a chance to loss his currency by underestimating or overestimating the actual currency required for transmission of packet. In Packet Trade Model, each node in the route buys currency from previous node which it sells to next node or the destination node. Hence nodes can earn currency for getting services from the network. The total cost incurred upon packet forwarding finally pays by the destination node. Author assumed that each node in the network have a temper-resistant hardware and a security model which is independent of nodes to prevent from forgery. Another motivational algorithm which does not required a temper resistant hardware is proposed in [13]. A central service is implemented in the network called Credit Clearance Service (CCS) which is responsible for keeping records of virtual money. When a node send it own packet, its credit is decrease while credit will increase when the node forward packet for others. However, intermediate node may claim credit falsely to get extra benefit so the intermediate nodes credit will increase only when destination node report CSS about successful receiving of packet.
To sum up, the mechanism review in this section deal with the selfish behavior using either watchdog techniques based on promiscuous monitoring such as [5][6][7][11] or motivational techniques [8][9][12][13]. As indicated in [14], in promiscuous mode a node’s energy is depleting by a higher rate as compare to when it is in idle mode or just discarding the traffic. However, in this paper we propose a proactive approach based on hierarchical monitoring & focusing on factors that derive selfish behavior, aiming to deal with the selfish behavior with low overhead on the network.
III.
the GR extract the residual battery status of the nodes (RB(Vi)) and using the last three parameters of GR, it calculates the overall network performance using the following impact matrix at each time frame.
SELFISH NODE DETECTION PROTOCOL A. ASSUMPTION
We consider selfish behavior of nodes at network layer, although nodes may exhibit selfish behavior at MAC layer, many researchers has proposed algorithms to deal with MAC layer selfish behavior [15][16][17]. We assume a clustered organization of MANET where the most capable node is selected as cluster head (CH) and other nodes become cluster nodes (CNs), we also assume secure communication between CH and CNs. We further assumed that there is no malicious node in the network as we only consider a selfish behavior from properly behaving nodes in this paper, which allow us to assume that there is no false reporting by CNs. B. CORE FUNCTIONALITIES OF SNDP We propose a selfish node detection protocol with two major considerations. First, we focus on the factors that motivate proper nodes to behave selfishly (i.e. to save their limited residual battery). Secondly we aim to propose a slightly light-weight mechanism in terms of low energy consumption. It has low overhead as most of detection work is performed at CH, however it still rely on neighbors of the node to investigate for its selfish behavior. CH keeps track of the overall network performance during all stages of the proposed protocol. We now explain the core functionality of our proposed protocol. SNDP comprises on three main modules as shown in its architecture in Fig. 1: a) monitoring and data collection, b) detection and c) response. a) Monitoring and Data Collection This module collects data using discrete time base operations, where the time is divided into smaller units (time intervals) such as T1, T2,….., Tn. At each time interval Ti each cluster node (Vi) sends a general report to the CH as depicted in Fig. 2. The general report (GR) contains the following parameter: GR={RB, Nbr_List, Num_Pkt_Gen, Num_Pkt_Rcvd, Num_Pkt_Fwd}
Where RB is the nodes residual battery, Nbr_List is the list of neighbours nodes, Num_Pkt_gen is the total number of packet generated by this node, Num_Pkt_Rcvd is the total number of packets which a node received as destination, Num_Pkt_Fwd is total number of packets which node forwards for other CNs.CH broadcast request for GR to all CNs, it uses a limited broadcast procedure to efficiently reach all CNs in the network. Each CNs then sends their GR to the CH.CH from
Fig. 1: Architecture of SNDP Protocol
= , }
Impact matrix consists of two parameters throughput and PDR. Throughput is an important parameter which measures the ability of successful transmission of traffic in the network. Packet delivery ratio is a ratio between the numbers of data packets generated by source nodes and the number of data packet successfully received at destination node. This matrix is used by the CH to estimate of the overall network performance and use it in detection and response modules. b) Detection This module is comprises of three phases: a) accusation, b) investigation and c) confirmation. In accusation phase, CH scrutinizes the general report send by CNs for the status of their residual battery for each time interval Ti as shown in Fig.2 (a). Upon receiving GRs from a CNs, CH finds the residual battery status of the node Vi which fits in the predefine low battery criteria. CH accuses this node Vi, and add it to the list of accused nodes based on the suspicion that it may exhibit a selfish behavior. CH will initiate the investigate module to inspect each node Vi in the list of accused node. In case of no node with a low residual battery status in a time interval Ti, CH will not take any further action in Ti. After this CH will start the investigation procedure (Fig.2(b)) for each of the accused node Vi. CH extracts the neighbors of the each accused node from the Nbr_List parameter of GR. Then, CH sends an accusation packet (ACC_Pkt) to each of the neighbor Vj of Vi . The format of the Acc-Pkt is as follows: Acc-Pkt = {AccNodeID, AccNodeNbrID, NumTestPkt }
Where AccNodeID represents the ID of the accused node and AccNodeNbrIDs stores the list of accused node neighbors. After receiving the Acc-Pkt, each neighbor Vj of accused node Vi will generates a test packets (Test-Pkt) to all other neighbors through the accused node Vi. The Test-Pkt contains the following fields: Test-Pkt = {SrcNodeID, DestNodeID, NextHopNodeID }
It is pertinent to mention that the nodes marked as selfish are not selfish by nature therefore they are still allowed to enjoy network services. After getting recharged its battery, accused node Vi sends general report at next timer interval indicating its battery status as recharged (normal) then CH broadcasts a selfish behavior alarm revoked packet to the CNs, and CNs will unmark the accused node Vi.
Where SrcNodeID is source node ID, DestNodeID is destination node ID (another neighbor node taking part in investigation), NextHopNodeID is accused node ID. Each neighbor Vj monitors Vi for a time interval in terms of number of TestPkt send to Vi for forwarding and number of Test-Pkt received from Vi. After the time interval each Vj send the specific report (SR) to the CH. SR= { , ! }, where Xij is the number of Test-Pkt received by Vj from Vi and Yij is the number of TestPkt send by node Vj from Vi . Upon receiving the SR the CH estimate the accused node behavior ANB(vi) using equation 1. n
A N B (vi ) =
∑
(
i=1, j=1
X ij ) …(1) Yij
CH investigate packet forwarding behavior of Vi taking advantage of the principle of flow conservation as proposed in [18]. According to this principle, each packet received by a node for which it is not a destination should exit that node. Equation 2 represents this principle where PRij (Ti) represents the number of packets received by node Vi for Vj and PTij (Ti) represents the number of packets transmitted by node Vi for Vj in Ti.
∑ PR (T ) = ∑ PT (T ).. ij
i
ij
i
for ∀ ij
,..(2)
According to equation 2 the ANB(vi) must be 100% in ideal conditions, however; by considering the possibility of collision, noisy radio channel and mobility, a threshold of ANB(vi) from 65% to 80% is generally acceptable. Any accused node with ANB(vi) value less than the threshold could indicate its selfishness. However, in order to reduce the probability of false conviction, CH checks the effect of Vi behavior on overall network performance using impact matrix during the tested time interval. If network performance matrix found degraded significantly, CH confirm the accused node Vi as selfish as shown in (Fig.2(c)). If the network performance matrix indicator does not show any considerable deviation then CH requests the neighbor nodes Vj to run of investigation phase for the next time interval. If the results are persistent then CH confirms Vi as selfish. c) Response Once node Vi is marked as selfish then in order to mitigate its effects on overall network performance, CH broadcasts a selfish behavior alarm packet to all CNs. All CNs, after receiving selfish behavior alarm packet, marked node Vi as selfish in their routing table. All existing routes containing Vi will be avoided, rather CNs will use second best path to forward the traffic. CNs will not include Vi in new paths by avoiding it in route discovery process.
Fig. 2: Pseudo code of SNDP Protocol
IV.
SIMULATION BASED CASE STUDY
A. Impact of selfish behavior in MANETs In this section we first performed a simulation based case study to analyze the impact of selfish behavior on overall network performance. We have driven a scalable network simulator GloMoSim. While using the GloMoSim, we have appended the extra parameter in terms of coding to fulfill our suggested model. Simulation study for evaluating the network performance is based on two scenarios: the first scenario we estimate the effect of different proportion of nodes involve in exhibiting selfish behavior in the network. In the second scenario we estimate the effect of selfish nodes mobility on
network performance. The parameters we have selected for our simulation environment is mentioned in Table. I. SIMULATION PARAMETERS
25
Terrain Dimension
900 * 900 meters
Simulation Traffic
CBR
Simulation Time
1000 sec.
Routing Protocol
AODV
MAC Protocol
802.11
Mobility Model Node Mean Speed
RWP varies from 0 -20 m/sec
We have simulated a 25 nodes network using Random Waypoint (RWP) as mobility model. To illustrate the concept we used AODV as the underlying routing protocol, however, the SNDP can be applied to any other routing protocol. IEEE 802.11 is used as a MAC layer protocol in our simulations. Scenario 1: In this scenario we have simulated with varying numbers of selfish node and observed their effect on overall network performance. To achieve this we explicitly maintain the (RB(Vi)) for all CNs. At the start of the simulation we set the (RB(Vi)) to maximum and when a nodes performs a task ( forward a packet, process route request etc.) we decrements the residual battery with the predefine unit with respect to that task. We perform 10 runs when no node behave selfishly in the network and estimate the network performance in terms of throughput and packet delivery ratio. Then analyze network performance with various proportion of the nodes in the network behaves selfishly. We perform 10 runs with each varying number of selfish nodes in the network. The graph in Fig.3 shows average throughput as a function of number of selfish nodes in the network. In general the graph shows that the throughput decreases gradually with the increase in the number of selfish nodes in the network. The graph in Fig. 3 illustrates the PDR as a function of number of selfish nodes in the network. We observe from the graph that the PDR drops significantly when a single node starts to behave selfishly (i.e. not forwarding & processing packets for other nodes) in the network. When further nodes starts to behave selfishly the graphs shows the shows comparatively low decrease in the PDR. In our opinion, this is because in a presence of a node is not forwarding packets, the other nodes in the network receive much less number of packets, so does not affect the PDR as compare to the first selfishly behaved node. To evaluate the impact of selfish nodes on the network performance we ensure that the matrix we used demonstrates changes and consequence that are cause by selfish nodes in MANET. We use two parameters of our impact matrix ( i.e. Throughput (Th), Packet delivery ratio (PDR)) because we scrutinize as per our experience and from the literature that in the evaluation of network performance throughput and PDR are more significant. Therefore, based on these factors we
Throughput vs Number of Selfish Nodes Throughput (bits/sec)
Number of Nodes
"#$ = % ∗ ∆(ℎ + % ∗ ∆$+, … (3) Value
30000 27000 24000 21000 18000 15000 12000 9000 6000 3000 0 0
2
4
6
8
10
Number of Selfish Nodes
Figure 3: Average throughput as a function of number of selfish nodes in the network
PDR vs Number of Selfish Nodes Packet Delivery Ratio (PDR)
Parameter
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0
2
4
6
8
10
Number of Selfish Nodes
Figure 4: Average Packet Delivery Ratio as a function of number of selfish nodes in the network.
Where wi represents the weights of the respective impact matrix parameters and ∑0 12 . = 1 . We assign the weights of throughput and PDR in Equation 3 respectively w1=0.6 and w2=0.4. In Equation 3, ∆ represents the percentage change, i.e. ∆Th calculates as a percentage change in throughput value with no selfish nodes and in the presence of various selfish nodes in the network simulation.
ENP vs Number of Selfish Nodes 100 90 80 70 60 50 40 30 20 10 0
Effect on Network Performance (%)
TABLE I.
derive equation 3 to compute the effect on network performance (ENP).
0
2
4
6
8
10
Number of Selfish Nodes
Figure 5: Effect of network performance as a function of number of selfish nodes in the network.
Scenario 2 In this scenario we analyze the performance of overall network on fixed number of selfish nodes but with different mean speed of the nodes in the network. We observe from the graph in Fig. 6 that when nodes including selfish nodes moves with higher means speeds, the throughput of the overall network drops. The behavior in this scenario is understandable because if the nodes are moving with high speed the chance of the disconnection and packet drops increases which effect the throughput of the overall network.
Throughput vs Node Mean Speed
ENP vs Node Mean Speed 100 90 80 70 60 50 40 30 20 10 0
No. Of Selfish Node = 2
0
4
8
12
16
20
Node Mean Speed (m/s)
Figure 8: ENP as a function of number of nodes mean speed.
B. Case Study based Evaluation of SNDP In this scenario we have evaluate our proposed SNDP model and simulate the simple case study environment in GloMoSim using the SNDP parameters mentioned in Table. II. TABLE II.
55000
Throughput (bits/sec)
that the performance of the overall network decreases with increasing mean speed of the node in the network.
Effect on Network Performance (%)
The Fig. 5 shows the effect of network performance in percentage as a function of number of selfish nodes in the network. In general the graph shows the rise in the effect of network performance with increase in the number of selfish nodes in the network. We can observe from the graph that the network performance degrade up to 50% when 10 nodes behave selfishly in the network.
SNDP PARAMETERS
No. Of Selfish Node = 2 50000
Parameter
Setting
45000
Number of Selfish Nodes
2
40000
No. of Time Interval
10
Time Interval
35000 30000 0
4
8
12
16
20
Node Mean Speed (m/s)
100 sec
Simulation Time
1000 sec.
Routing Protocol
AODV
Mobility Model
RWP
Node Mean Speed
2.5 m/sec
Figure 6: Average network throughput as a function of nodes means speed.
In general the fig. 7 shows that the PDR decreases when the nodes are moving with the high mean speed. In Fig. 7 at different mean speed this graph shows inconsistency due to the effect of moving selfish node. The inconsistence scenario in this figure is observed in between the mean value of 6m/s to 10m/s, the graph shows increase in the PDR which is in our opinion is due to rapid change in the position of both selfish nodes.
Packet Delivery Ratio (PDR)
PDR vs Node Mean Speed 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
We have simulated the scenario in GloMoSim which is not the GUI based simulator, therefore we illustrate the graphical representation of the different steps involved during the SNDP tasks in this case study. We have run the simulation using two random nodes which behave as a selfish but for investigating the impact of our proposed model on the network performance we have chosen to deal with one selfish node. The graphical representation of collecting GR from CNs after time Ti by CH is shown in Fig.9 after collecting the GR reports from CNs the CH during the simulation extracts (RB(Vi)) from the GR report and recognize that node V10 has low battery status and node V6, V12 are its neighbor.
No. Of Selfish Node = 2
0
4
8
12
16
20
Node Mean Speed (m/s)
Figure 7: Average PDR as a function of nodes mean speed.
Finally, we analyze the ENF against the nodes mean speed in the presence of two selfish nodes. The graph in Fig. 8 shows
Fig. 9: Snap shot of network at time Ti
c Ac
analyzing the effect of ENP on overall network at time Ti and Ti-1. If observation of ENP values shows significant degradation of network performance, CH confirms the accused node V10 as selfish. If the network performance matrix indicator does not show any considerable difference then CH requests the neighbor node V6 and V12 to repeat the investigation phase for node V10.
-Pk t
Fig. 10: CH sends Accused Packet (Acc-Pkt) to accused node’s neighbor.
In Fig. 10 shows when CH accused V10 as selfish node and sends Acc-Pkt to its neighbor V6 and V12 to start the investigation process after T1during the simulation. After receiving Acc-Pkt, node V6 generates Test-Pkt and sends to accused node V10 to be forwarded to node V12 as a destination. Node V12 performs same operation for node V6 as shown in Fig. 11.
At this stage, CH marked accused node V10 as selfish node and in order to mitigate its effects on overall network performance, CH broadcasts a selfish behavior alarm packet to all CNs as shown in Fig. 13. All CNs, after receiving selfish behavior alarm packet, marked node V10 as selfish in their routing table. All CNs avoid routes with Node V10, rather CNs will use second best path to forward traffic if available otherwise another route discovery will be proceed. After the simulation run on this scenario on fixed number of selfish node i.e. 2, we calculate the total number of control packets of SNDP that is GR request packet, accusation packets, test packets and selfish alarm behavior packets. Then we include the SNDP control packets with the routing protocol (AODV) control packets and the results indicate 1% increase in the total control packets in the network while using SNDP. However the size of SNDP control packet is different from AODV but it still reflects the estimated overhead of the purposed mechanism. V8 V17 V2
V9
Test
Selfish Behavior Alarm Packet
et Pack
V13
V1 Selfish Behavior Alarm Packet
After completion of investigation, neighbor of accused selfish node V10 i.e. (V6 and V12) send a specific report to CH on next time interval Ti+1as shown in Fig. 12.
Fig.12: Neighbor of accused node send Specific Report to CH.
CH analyzes the specific reports submitted by node V6 & V12 and calculates ANBV10. We fixed the node V10 as selfish in our simulation network environment, therefore ANBVi shows the packet forwarding ratio of V10 is less than predefined threshold (65%) in this case. The behavior of accused node V10 will be treated as suspicious. However; in order to avoid false conviction, CH also verify the behavior of node V10 by
io r av eh et h B Pack lfis Se larm A
V4
V14
Selfish Behavior Alarm Packet
V7
V15
V10
V5 CH
V11
Selfi sh Alarm Behavio r Pack et
V16 V12
Fig. 13: CH broadcasts Selfish behavior Alarm Packet to all cluster nodes
V.
R Sp ep e or c i f t ( ic V1 2)
V6
ior av eh et h B Pack lfis Se larm A
ior av e h et h B Pac k lfis Se larm A
Selfish Behavior Alarm Packet
Fig. 11: Neighbors of accused node send Test Packet (Test-Pkt) to accused node
Selfish Behavior Alarm Packet
V3
CONCLUSION AND FUTURE WORK
Limited resources of nodes in MANETs without any centralized administration motivate nodes to behave selfishly to preserve its resources. Therefore we consider selfish behavior of nodes in MANETs and proposed a proactive and comparatively lightweight mechanism SNDP which uses hierarchical monitoring and focuses on factors which derives selfish behavior from properly behaving nodes. We present a simulation-based study to analyze & estimate the effect of selfish behavior on network performance. We further presented a simulation based scenario to perform an initial evaluation of SNDP. In future we are aiming to comprehensively test SNDP with various scenarios. We will also generalize SNDP to accommodate all types of selfish behavior in MANET.
REFERENCES [1]
H. Miranda and L. Rodrigues, “Preventing selfishness in open mobile ad hoc networks,” In Proceedings of the 7th CaberNet Radicals Workshop, October 2002. [2] L. Buttyan and J-P. Hubaux, “Security and cooperation in wireless networks,” a book published by Cambridge University Press, Publication date: 11 July 2012. [3] Wang, Yongwei, "Enhancing Node Cooperation in Mobile Wireless Ad Hoc Networks with Selfish Nodes" (2008). Doctoral Dissertations. Paper 602. http://uknowledge.uky.edu/gradschool_diss/602 [4] F. Kargl, A. Klenk, S. Schlott, and M. Weber, “Advanced Detection of Selfish or Malicious Nodes in Ad hoc Networks”, In Proceedings of the 1st European on Security in Ad-Hoc and Sensor Networks (ESAS 2004) [5] Y. Wang, and M. Singhal, "A Light-Weight Solution for Selfish Nodes Problem Considering Battery Status in Wireless Ad-Hoc Networks", In Proceeding of IEEE WiMob'05, Vol. 3, pp. 299-306, Montreal, Canada, August 22-24 2005 [6] D. Djenouri, N. Badache, “New approach for selfish nodes detection in mobile ad hoc networks,” in the first IEEE/Creat-net Workshop on Integration of Security and Quality of Service (SecQoS’05), Athens, Greece (2005) [7] T. Fahad and Robert, “A Node Misbehaviour Detection Mechanism for Mobile Ad-hoc Networks”, in proceedings of the 7th Annual PostGraduate Symposium on The Convergence of Telecommunications, Networking and Broadcasting, June 2006. [8] M. T. Refaei, V. Srivastava, L. DaSilva, and M. Eltoweissy. “A reputation-based mechanism for isolating selfish nodes in ad hoc networks,” In Proceeding of IEEE Second Annual International Conference on Mobile and Ubiquitous Systems: Networking and Services (MobiQuitous ’05), pages 3–11, San Diego, CA, July 2005. [9] J. J. Jaramillo, and R. Srikant “DARWIN: Distributed and Adaptive Reputation mechanism for WIreless ad-hoc Networks”, In proceedings of the 13th annual ACM international conference on Mobile computing and networking MobiCom’07, Montréal, Québec, Canada. September 9–14, 2007
[10] A.Nadeem and M.Howarth, “A Generalized Intrusion Detection & Prevention Mechanism for Securing MANETs”, In a Proceeding of IEEE 5th International Conference on Ultra Modern Telecommunications & workshops (ICUMT 09), St Petersburg Russia, th 12-14 October 2009. [11] E. Hernandez-Orallo, M. D. Serrat, J. Cano, C. T. Calafate, and P. Manzoni, “Improving selfish node detection in MANETs using a collaborative watchdog”, IEEE Communications Letters, Accepted. May 2012. [12] L. Buttyan and Jean-Pierre Hubaux. “Enforcing Service Availability in Mobile Ad-Hoc WANs” In Proceedings of IEEE First Annual Workshop on Mobile and Ad Hoc Networking and Computing, 2000. MobiHOC. 2000 [13] S. Zhong, Y. R. Yang, and J. Chen. “Sprite: A simple, cheat-proof, credit-based system for mobile ad-hoc networks” In Proceedings of INFOCOM 2003. Twenty-Second Annual Joint Conference of the IEEE Computer and Communications. IEEE Societies, Mar 2003. [14] L. M. Feeney, “An Energy Consumption Model for Performance Analysis of Routing Protocols for Mobile Ad Hoc Networks”, In Journal Mobile Networks and Applications, Vol.6, No.3, June 2001, pp 239-249. [15] P. Kyasanur and N. Vaidya, “Detection and handling of mac layer misbehavior in wireless networks,” In Proceedings of International Conference on Dependable Systems and Networks (DSN). pp. 173 – 182. June 2003, [16] J. Choi, A. W. Min, and K. G. Shin, "A Lightweight Passive Online Detection Method for Pinpointing Misbehavior in WLANs", IEEE Transactions On Mobile Computing, VOL. 10, NO. 12, pp 1681-1693, December 2011. [17] P. Kyasanur and N. H. Vaidya, “Selfish MAC Layer Misbehavior in Wireless Networks,” IEEE Transaction on Mobile Computing, Volume , No. 5, pp. 502-516, September/October 2005. [18] O. F. Gonzalez, M. Howarth, G. Pavlou, “Detection of packet forwarding misbehavior in MANETs”, In Proceedings of International. Conference on Wired/Wireless Internet Communications (WWIC 2007), Coimbra, Portugal, 23-25 May 2007.