Sunteți pe pagina 1din 7

2014 International Conference on Computer Communication and Informatics (ICCCI -2014), Jan.

03 05, 2014, Coimbatore, INDIA

Reputation Management for Data Forwarding in


Opportunistic Networks
S Premalatha, V Mary Anita Rajam
Dept. of Computer Science and Engineering
CEG, Anna University

Abstract In opportunistic networks, contacts between nodes are Section IV analyzes the performance of the routing
viewed as an opportunity to move data closer to the destination. mechanisms and Section V concludes the paper.
Social routing is a method in which the messages are selectively
forwarded through the nodes with which encounters are more II. RELATED WORK
likely to occur. One such social routing technique is forwarding
the data through nodes that are known to be friends. But there is Iain et al. [1] proposed a privacy enhanced social routing in
a possibility that the nodes which were friends earlier might which two schemes namely Statisticulated social network
become selfish or malicious in future. Hence it is necessary to routing (SSNR) and Obfuscated social network routing
monitor the friend nodes and identify the possibly malicious or (OSNR) were used. In SSNR, the friends list is modified by
selfish friends. This paper proposes a social routing protocol for adding or removing nodes for each message transmission.
opportunistic networks, which uses the reputation of the nodes in Hence it is not easy for a node to identify the original friends
the network to make the decision of forwarding data. list of a sender by just interpreting a single message. In OSNR
the friends list of source node is embedded in a bloom filter.
Keywords-Opportunistic networks; friends; reputation; privacy Lilien et al. [9] discussed security and privacy challenges in
opportunistic networks such as helper privacy, opportunistic
I. INTRODUCTION (HEADING 1) network privacy, data privacy and intrusion detection. This
work discussed the use of trusted devices for more critical tasks
Opportunistic networks use a store, carry and forward and secured routing was enabled by selecting a route that
architecture where each node that generates or receives data, passes through only trusted devices.
carries it until it finds a suitable carrier and then forwards
it. Thus, contacts between nodes are viewed as an opportunity Na Li et al. designed a trust based framework for data
to move data closer to the destination. The routes between forwarding in opportunistic networks [2]. A watchdog
nodes are dynamically created, and any node can be component was included in the trust framework to monitor the
opportunistically used as a next hop, if they can bring the behaviour of the forwarding node. The PFM was generated by
message closer to the destination. One such routing is social the receiving node to the source to inform the behavior of the
routing in which the data is forwarded through nodes that are forwarding node. Based on the received PFM, trust to the
friends. forwarding behavior of a node was calculated. The forwarding
decision of a node was taken based on the trust and the
The main problem with social routing is that there is a forwarding ability of a node.
possibility that the nodes which had been friends earlier might
become selfish or malicious in future. The nodes which Yining et al. introduced a reputation computation model
become malicious will drop all the incoming messages to it. based on subjective logic [3] which identified and prevented
The selfish nodes may forward the incoming messages only for selfish behaviour. This model worked in two phases namely
a period of time and later start dropping the incoming reputation query and reputation computation. In the reputation
messages. Hence it is necessary to monitor the friends for query phase, a node that received a service request
identifying the misbehaving nodes. accumulated the recommended opinions on the requestor from
their common neighbours who had interacted with it. In the
In this paper we introduce a social routing protocol in reputation computation phase, the node evaluated the
which a reputation value is calculated for all nodes. The reputation with the opinions accumulated in the reputation
reputation value is updated using two mechanisms, namely the query phase.
acknowledgement-based reputation (AR) system and the
message delivery based reputation system (MDR). Based on Greg et al. proposed an incentive mechanism [8] in which
the reputation values, a mechanism for calculating a reputation Self-reported social networks (SRSNs) were used to collect
threshold is also introduced. The reputation threshold is then social network data. These SRSNs were used to provide
used to categorize misbehaving nodes. reputation for nodes. When the network starts up, nodes assign
This paper is organized as follows: Section II discusses higher trust values to nodes in their SRSN. Selfishness is
about the work done earlier in social routing. Section III detected by storing a history of encounter times and
discusses the AR and MDR systems for updating the reputation exchanging the histories during encounters. Once a node is
values and the method of calculating the reputation threshold.

978-1-4799-2352-6/14/$31.00 2014 IEEE


2014 International Conference on Computer Communication and Informatics (ICCCI -2014), Jan. 03 05, 2014, Coimbatore, INDIA

detected as selfish the detecting node decrements the value of In Figure 1, node 1 generates a message M1 and wants to
selfish node by a behavior constant. send M1 to node 4. M1 is transmitted to node 2. Node 2 then
forwards the message to node 3, node M and node S. Here
In a packet dropping detection and routing misbehaviour node M is a malicious node and hence drops the message M1
mitigation scheme [11] introduced by Qinghua et al., the and node S is a selfish node, which would keep the message
misbehaving node was required to generate a contact record for some time in its buffer and later drop the message without
during each contact and report its previous contact records to forwarding it to destination node. Hence node 3 alone helps in
the contacted node. Based on the reported contact records, the forwarding the message to the destination node. So, when the
contacted node detected if the misbehaving node had dropped ACK is got back in the source, the reputation value is
packets. The misbehaving node may misreport to hide its incremented only for nodes 2, 3 and the reputation value for
misbehaviour, but forged records cause inconsistencies which nodes M and S remain same. In Figure 1, a direct line from the
make misreporting detectable. To detect misreporting, the destination to the source is shown just to indicate that the ACK
contacted node also randomly selected a certain number of is sent back to the source.
witness nodes for the reported records and sent a summary of
each reported record to them when it contacted them. The
witness node that collected two inconsistent contact records
could detect the misreporting node.

III. INTEGRATED SOCIAL ROUTING PROTOCOL


Each mobile node maintains a reputation table with structure
as shown in Table 1. Initially the reputation value is zero for all
the nodes in the network. It is assumed that each node also
maintains a list of friend nodes. Any message sent by any
source node is first forwarded only through the friends of the
source node. These nodes which carry the messages can
directly handover the message to the destination or forward it Figure 1. AR system
through their friends in turn. This continues till the message
reaches the destination. The decision as to whether the message Table 2 shows the initial reputation values of all the nodes
can be forwarded through a node, though a friend is made in the network.
based on the reputation maintained for that node in the Table 2. Initial reputation table
reputation table.
Table 1. Structure of reputation table
Nodes in Oppnet Reputation Value
Nodes in Oppnet Reputation Value
1 0
2 0
M 0
3 0
In this paper, we have introduced two ways in which the S 0
reputation of nodes is updated in reputation tables, which are
explained in the following subsections. 4 0

A. Acknowledgement-based reputation system (AR)


In AR method, whenever a node forwards a message, id of
that node is also stored in the message. After each message is
received at the destination node, the destination node creates an
ACK. The destination node uses the path information stored in
the message received to find the nodes which help in
forwarding the message to it. The destination node now sends
the ACK message to the source node. Once the source node
receives the ACK from the destination node, it increments the
reputation of nodes which had contributed for the message
transfer as forwarding nodes. The intermediate nodes which
forward the ACK also increment the reputation value for the
forwarding nodes, thereby learning that these nodes are good
nodes. The reputation value for the other nodes remains the
same.

Figure 2. Updation of reputation table using AR system


2014 International Conference on Computer Communication and Informatics (ICCCI -2014), Jan. 03 05, 2014, Coimbatore, INDIA

Figure 2 shows the reputation table with values updated by all the nodes is compared with the calculated reputation
the source node after receiving the ACK message from the threshold and if the reputation value of a node is less than the
destination node. The reputation values of nodes 1, 2 and 3 are calculated threshold then that node is added to a suspicious list
incremented as they helped for forwarding the data whereas the of nodes. That is, if the reputation value of a node is smaller
reputation values of nodes M and S remain same. Thus, than the mean reputation by a value equal to the standard
whenever a node contributes in forwarding a message, the deviation, then the node is categorized as a suspicious node. If
reputation of that node gets incremented. the node remains in the suspicious list of nodes for a long time,
then that node is identified as misbehaving node and is
B. Reputation calculation using Message Delivery based removed from the original list of friends. Hence the original
reputation system friends list is updated periodically and after several updates it
will consist of only good reputed friend nodes.
In this system, similar to the AR system, the source node
sends the message through friend nodes and the message
reaches the destination carrying the information about the IV. SIMULATION RESULTS
forwarding nodes. The destination node now uses this path Simulations were done using the ONE (Opportunistic
information for finding the nodes that helped in the forwarding Network Environment) simulator. A transmission range of 50
and increments the reputation values of those nodes which meters was considered and the data rate used was 250kbps. We
helped in transmitting the messages to it. The intermediate have considered five groups of nodes and the simulations were
nodes that helped in forwarding the message also update the run for 10020 seconds.
reputation values in their respective reputation tables.
The performance of the routing protocol is analyzed by
Figure 3 shows that message M1 is transmitted from source varying the number of nodes, number of selfish nodes and
node 1. The source node then forwards the message to node 2 number of malicious nodes.
and node M. Node 2 alone forwards the message to node 3 and
node M drops the message M1. Node 3 after receiving message The performance is evaluated in terms of packet delivery
forwards M1 to node 4 and node S. Node S puts the message in probability, average delay for packet transfer and routing
its buffer and drops it after some period of time. Here, node 4 is overhead.
the destination node and it receives its message from node 3.
Then it finds the node which helped in forwarding the message
using the path information.

Figure 4. Comparison of Message delivery probability by


Figure 3. Reputation value updation using MDR system varying number of nodes
Figure 4 shows the message delivery probability of the
C. Threshold based updation of friends list MDR system and the AR system when the number of nodes is
We also introduce a mechanism to determine suspicious varied. It is seen that the delivery probability is high for the
nodes. The friends list of each node is updated at regular MDR system. This is because the ACK generated in the
intervals of time based on the reputation values in the destination node may not reach the source and hence the
reputation table maintained by the node using the mechanism updation about good nodes is not done in the source as well as
described below. Based on the AR and the MDR systems, the the intermediate nodes. Hence there is less knowledge about
reputation values of nodes are updated in the reputation table. good nodes in the network. When the number of nodes in
The mean () and standard deviation () of reputation values increased, there is a higher possibility of the ACK reaching the
source.
from the reputation table is calculated. Then the reputation
threshold (RT) -is calculated. The reputation value of
2014 International Conference on Computer Communication and Informatics (ICCCI -2014), Jan. 03 05, 2014, Coimbatore, INDIA

Figure 5 shows the average delay of the MDR system and high for MDR system when compared to that of the AR
the AR system when the number of nodes is increased. It is system.
seen that the AR system provides less delay than the MDR
system. It is also seen that the delay gets reduced when the Figure 8 shows the average delay of the MDR system and
number of nodes is increased for both the systems. When the the AR system when the percentage of selfish nodes is
number of nodes in increased, there are more nodes to forward increased. It is seen that the AR system provides less delay
the message and the message gets forwarded in less number of than the MDR system.
hops.

Figure 7. Comparison of Message delivery probabilities by


Figure 5. Comparison of Average Delay by varying number varying percentage of selfish nodes
of nodes

Figure 8. Comparison of Average Delay of by varying


Figure 6. Comparison of Routing Overhead by varying percentage of selfish nodes
number of nodes
Figure 9 shows the routing overhead of the MDR system
Figure 6 shows the routing overhead of MDR system and and the AR system when the percentage of selfish nodes in the
AR system when the number of nodes in the network is network is increased. It is seen that the overhead is very high in
increased. It is seen from the figure that the overhead is very the MDR system compared to that of the AR system.
high in MDR system compared to that of AR system.
The message delivery probability of the MDR system and
We have also compared the performance when the the AR system by varying the percentage of malicious nodes is
percentage of selfish nodes and malicious nodes in the network shown in Figure 10. It is seen that the delivery probability is
are varied. Figure 7 shows the message delivery probability for
the MDR system and the AR system when the percentage of high for the MDR system when compared to that of the AR
selfish nodes is varied. It is seen that the delivery probability is system while increasing percentage of malicious nodes.
2014 International Conference on Computer Communication and Informatics (ICCCI -2014), Jan. 03 05, 2014, Coimbatore, INDIA

Figure 9. Comparison of Routing Overhead by varying Figure 11. Comparison of Average Delay by varying
percentage of selfish nodes percentage of malicious nodes

Figure 12. Comparison of Routing Overhead by varying


Figure 10. Comparison of Message delivery probabilities percentage of malicious nodes
by varying percentage of malicious nodes
Figure 11 shows the average delay for the MDR system and
the AR system by varying the percentage of malicious nodes. It
also shows that AR system provides less delay than the MDR
system.
Figure 12 shows the routing overhead for the MDR system
and the AR system while varying the percentage of malicious
nodes in the network. It is seen that the overhead is very high
in the MDR system compared to that of the AR system.
Based on the reputation threshold that is found, we also
have analyzed the number of good nodes wrongly classified as
bad nodes and also the number of bad nodes wrongly classified
as good nodes. Figure 13 shows the variation of the number of
false positives under the MDR system and the AR system by
varying the percentage of malicious nodes. It is seen that the
number of false positives is high for MDR when compared to Figure 13. Comparison of false positives by varying
that of AR system. percentage of malicious nodes
2014 International Conference on Computer Communication and Informatics (ICCCI -2014), Jan. 03 05, 2014, Coimbatore, INDIA

positives is zero after increasing the percentage of selfish


nodes.

Figure 14. Comparison of false negatives by varying


percentage of malicious nodes
Figure 16. Comparison of false negatives by varying
Figure 14 shows the variation of number of false Number of nodes
negatives under MDR reputation system and AR system by
varying the percentage of malicious nodes. The figure shows
that the number of false negatives is zero for MDR reputation
system after increasing the percentage of malicious nodes
whereas the number of false negatives is high for the AR
system.

Figure 17. Comparison of false positives by varying


Percentage of selfish nodes

Figure 15. Comparison of false positives by varying


Number of nodes
Figure 15 shows that the false positives are very high for
the MDR system when the number of nodes is increased
whereas for the AR system the number of false positives is
very low while increasing the number of nodes.
Figure 16 shows that the false negatives are very high for
MDR reputation system when the number of nodes is increased
whereas for the AR system the number of false negatives is
very low while increasing the number of nodes.
Figure 17 shows that the false positives are very high for
MDR reputation system when the percentage of selfish nodes
is increased whereas for the AR system the number of false
Figure 18. Comparison of false negatives by varying
percentage of selfish nodes
2014 International Conference on Computer Communication and Informatics (ICCCI -2014), Jan. 03 05, 2014, Coimbatore, INDIA

REFERENCES [7] Guohong Cao, Qinghua Li, Sencun Zhu, Wei Gao, A routing protocol
for socially selfish delay tolerant networks, Ad hoc networks, Vol. 10,
No. 8, pp. 1619 1632, 2012.
[1] Iain Parris, Tristan Henderson, Privacy-enhanced social-network [8] G.Bigwood, T. Henderson, IRONMAN: Using social networks to add
routing, Computer Communications, Vol. 35, No.1, Elsevier, pp. 62-74, incentives and reputation to opportunistic networks, In Proceedings of
2012. the IEEE Third International Conference on Social Computing, pp. 65
[2] Na Li, Sajal K. Das, A trust-based framework for data forwarding in 72, 2011.
opportunistic networks, Ad Hoc Networks, Vol. 11, No. 4, Elsevier, pp. [9] Leszek Lilien, Zille Huma Kamal, Vijay Bhuse, Ajay Gupta, The
1497-1509, 2013. Concept of Opportunistic Networks and their Research Challenges in
[3] Yining Liu, Keqiu Li, Yingwei Jin, Dr. Yong Zhang and Wenyu Qu, A Privacy and Security, Mobile and Wireless Network Security and
novel reputation computation model based on subjective logic for Privacy, pp. 85-117, 2007.
mobile ad hoc networks, Journal of Future Generation Computer [10] Towards a reputation-based routing protocol to contrast blackholes in a
Systems, Vol. 27, No. 5, Elsevier, pp. 547-554, 2011.
delay tolerant network, Ad Hoc Networks, Vol. 10, No. 7, pp. 1167
[4] Mtibaa, M. May, C. Diot, M. Ammar, PeopleRank: social opportunistic 1178, 2012.
forwarding, In Proceedings of IEEE INFOCOM, pp. 1 - 5, 2010.
[11] Qinghua Li, Guohong Cao, Mitigating Routing Misbehavior in
[5] Ning Wang, E.Yoneki, Impact of Social Structure on Forwarding Disruption Tolerant Networks, Information forensics and security, Vol.
Algorithms in Opportunistic Networks, Mobile and Wireless 7, No. 2, pp. 664-675, 2012.
Networking, pp. 82 88, 2011.
[6] A.Passarella, C.Boldrini, M.Conti, Exploiting users social relations to
forward data in opportunistic networks: The HiBOp solution, Pervasive
and mobile computing, Vol. 4, No. 5, pp. 633 657, 2008.

S-ar putea să vă placă și