Docstoc

An Efficient Trust Establishment Framework for MANETs

Document Sample
An Efficient Trust Establishment Framework for MANETs Powered By Docstoc
					                                                         (IJCSIS) International Journal of Computer Science and Information Security,
                                                         Vol. 8, No. 4, July 2010




      An Efficient Trust Establishment Framework for
                         MANETs

                                           Mohammad Karami, Mohammad Fathian
                                                Department of Industrial Engineering
                                             Iran University of Science and Technology
                                                            Tehran, Iran


Abstract— In this paper, we present a general trust establishment         and communication protocols for MANETs have been
framework comprising three components. The first part is the              developed optimistically, where the benign and cooperative
trust computation model that evaluates the trust level of each            behavior of all the participating nodes is presumed. However, it
participating node through monitoring and quantification of               may not be always the case and in the absence of a fixed trust
some relevant behavioral indicative metrics. The second part is           or security infrastructure; some nodes may decide to exhibit a
the trust evidence distribution scheme that distributes the trust         non-cooperative or malicious behavior for a variety of
evidences obtained by the first component. And finally the third          incentives including better service, selfishness, monetary
part is the reputation computation model that combines the                benefits or malicious intents.
collected trust evidences from other nodes to form an overall
reputation score and a judgment basis regarding the                              Due to the unique characteristics of MANETs such as
trustworthiness level of each node.                                       shared wireless medium, the lack of any fixed infrastructure,
                                                                          mobility and consequently dynamic topology changes, and
   The trust computation model is based on first-hand evidences           resource-constrained nodes in terms of battery and computation
obtained via direct observations at the MAC layer. The proposed           capability, these networks are seriously susceptible to a large
trust evidence distribution scheme is an efficient, scalable and          number of security attacks [2]. The aforementioned
completely distributed scheme based on ant colony optimization            characteristics also prevent traditional cryptographic-based
algorithm. For combination of collected evidences in the                  security methods to be directly applicable to MANETs.
reputation computation model, Dempster’s rule for combination
is applied. Dempster’s rule for combination gives a numerical                    As a result, in recent years researchers have taken a
procedure for fusing together multiple pieces of evidence from            trust-based approach which promotes modeling and computing
unreliable observers.                                                     trust by defining and monitoring some behavioral indicative
                                                                          metrics and coming up with some sort of belief in
   The paper, illustrates the applicability of the proposed               trustworthiness level of other nodes. This computed degree of
framework on data packet delivery functionality with Dynamic              trustworthiness may then be used in situations where a node
Source Routing (DSR) as the underlying routing protocol. We               has to rely on previously unknown and therefore unreliable
present simulation results which demonstrate the effectiveness
and efficiency of the proposed framework.
                                                                          nodes for accomplishment of a cooperative service. In a
                                                                          MANET context, trust is defined as a belief level that one node
                                                                          can put on another node for a specific action according to
                                                                          previous direct or indirect information from observation of
   Keywords- Trust establishment framework; mobile ad hoc                 behaviors. The belief level is the extent to which one node
network (MANAT); evidence distribution; ant colony optimization;          believes that another node is willing and able to obey the
Dempster-Shafer theory                                                    protocol and act normally [3].
                                                                                 In this paper, we present a trust establishment
                        I.   INTRODUCTION                                 framework that is based on first-hand evidences obtained via
    Mobile ad hoc networks (MANETs) are multihop wireless                 direct observations at the MAC layer as well as second-hand
networks spontaneously constructed by mobile nodes without                evidences that are obtained via an ant-based trust evidence
relying on any pre-established infrastructure [1]. In MANETs,             distribution scheme from other nodes. A common difficulty in
nodes can directly communicate with other nodes within their              trust-based schemes that incorporate various trust evidence
wireless transmission range that are often referred to as                 exchange mechanisms to reinforce their accuracy pertains to
neighbors. However, to communicate with non-neighbor nodes,               the combination of observational data from nodes that can vary
they have to follow a multi-hop scenario where the source                 in their reliability or trustworthiness. In this paper, we have
nodes rely on their neighbors and several other intermediate              employed the Dempster-Shafer evidence theory, which is well
nodes to relay their messages and deliver them to the                     suited to an ad-hoc network where doubt and uncertainty is
destination. Therefore, the cooperation of participating nodes            inherent.
plays a vital role for successful communications. Early routing




                                                                    252                              http://sites.google.com/site/ijcsis/
                                                                                                     ISSN 1947-5500
                                                          (IJCSIS) International Journal of Computer Science and Information Security,
                                                          Vol. 8, No. 4, July 2010




    The remainder of the paper is organized as follows. Section            method for combining observational data from nodes that can
II briefly reviews related work on trust establishment in                  vary in their reliability or trustworthiness. Previous approaches
MANETs. Section III is dedicated to the details of our                     have used simplistic combination techniques such as averaging
proposed trust establishment framework. Section IV presents                or majority voting [18,19]. Here we apply Dempster-Shafer
results from simulation experiments that demonstrate the                   mathematical theory of evidence to combine independent
effectiveness of the proposed scheme. The final section of the             pieces of evidence collected from other nodes in order to form
paper discusses concluding remarks.                                        an overall reputation score regarding the trustworthiness degree
                                                                           of a given node.
                      II. RELATED WORK
    In recent years, security establishment in MANETs by the                               III. THE PROPOSED FRAMEWORK
means of trust modeling and management has been a                              As in real life, in MANETs context, trust levels are
considerable topic of interest. The proposed trust management              determined for particular actions. Obviously, trust computation
frameworks in literature fall into two major categories,                   for any action of interest requires clear definition, monitoring
reputation-based [4,5] and trust establishment [6-9]. In the               and quantification of some relevant behavioral indicative
former category, trust in other nodes is evaluated by direct               metrics. We believe that our proposed framework is a general
observation and second-hand information distributed among a                framework and once corresponding metrics for a given action
network. In this category most of the proposed methods use a               of interest are properly defined, monitored and quantified, it
Bayesian approach based on Beta distribution [3, 5, 10, 11]. In            may be adapted for various scenarios. However, to give a
this approach, a random variable that follows the beta                     practical illustration, for the rest of the paper, we will be
distribution is associated with the trust value of a node. Also,           particularly considering the incorporation of the proposed
the posterior distribution that represents a notion of trust is            framework into data packet delivery functionality with
derived from a prior distribution. In the later category [6-9],            Dynamic Source Routing (DSR) as the underlying routing
trust in neighbors is evaluated by direct observation, and trust           protocol [20]. In the resulted trust-aware DSR protocol, the
relations between two nodes without previous direct interaction            trustworthiness degree of intermediate nodes is taken into
are established through a combination of opinions from                     account, so that, non-cooperative nodes could be avoided in
intermediate nodes.                                                        route selection decisions. The details of the proposed trust
                                                                           establishment framework are discussed in subsequent
        L. Eschenauer et al. [12] present a high-level framework           subsections.
for generation, revocation and distribution of trust evidence and
demonstrate the significance of estimation metrics in trust
establishment. A.A. Pirzada et al. [13] present a trust model              A. Trust Computation Model
that allows the evaluation of the reliability of the routes, using             The trust computation model is executed by each individual
only first-hand information. The notion of confidence as it                node. Each node operates independently and maintains its
relates to trust management was explored by G.                             individual perspective of the trust hierarchy. Each node uses a
Theodorakopoulos et al. [14]. L. Buttyan et al. [15] propose a             direct observation mechanism for monitoring data packet
framework for stimulating cooperation in MANETs. The                       forwarding behavior of its neighbor nodes and accordingly
approach is based on a credit system for packet forwarding                 quantifies trust level of each neighbor node.
while trusted hardware is assumed.                                                In the proposed scheme, each node buffers all the
        The majority of research works presented in the                    packets it has sent, puts itself in promiscuous mode, initiates a
literature have mainly concentrated on trust modeling and                  timer and then overhears its neighbor’s forwarding behavior. If
quantification, while little attention has been paid to efficient          a packet is properly forwarded within the expected timeout,
distribution of trust information. In most of the proposed trust           then a successful forwarding event is recorded, otherwise an
establishment schemes participating nodes are required to                  unsuccessful forwarding event is recorded. The trust level is
periodically disseminate their trust information acquired                  simply computed by dividing the number of successful
through direct observations.                                               forwarding observations for a particular node by the total
                                                                           number of packets sent to that node to be forwarded. In
   These trust information are received by other nodes and                 particular, the trust value, t, assigned to node j by node i is
combined to form an overall reputation score for each node.                defined as follows:
This proactive approach suffers scalability, efficiency and
robustness problems in resource-constrained environments                                  Ns
[16]. Tiang and Baras [17] propose an efficient ant-based                      tij =                                                     (1)
approach for the distribution of trust certificates in MANETs.                         N s + Nu
However, their proposed scheme does not involve any trust or
reputation computation model. In this paper we use an efficient
on-demand trust evidence discovery protocol based on ant
colony optimization algorithm for the distribution of trust                    Where 0 ≤ t ≤ 1 and Ns and Nu respectively represent
evidences.                                                                 the cumulative number of successful and unsuccessful
                                                                           forwarding events of node j recorded by node i. A trust value of
       Yet another challenge in reputation-based schemes is                0 for a given node represents complete distrust and a value of 1
related to employing an accurate, robust and straightforward




                                                                     253                              http://sites.google.com/site/ijcsis/
                                                                                                      ISSN 1947-5500
                                                          (IJCSIS) International Journal of Computer Science and Information Security,
                                                          Vol. 8, No. 4, July 2010




implies absolute trust in packet forwarding functionality of that              The idea of the proposed ant-based scheme is inspired by
node.                                                                      the process used by real ant colony. The ant can seek path
                                                                           between the nest (source node) and multiple food sources
    The trust value computed for each neighbor node is signed              (nodes hosting relevant trust evidences). They accomplish the
by observer’s private key and therefore can’t be modified by               mission with great efficiency. As the environment changes,
intermediate nodes. We assume that the public key of the                   ants can also quickly discover new routes. Since trust evidence
signer is well known and authenticated, and the corresponding              discovery is a process to find relevant evidences with the best
private key cannot be compromised. Trust evidence is a                     efficiency, utilizing the ant colony optimization proves to be
foursome tuple denoted as TE=<provider, target, TV, time>.                 helpful.
Provider is the observer node which has computed the trust
value, target represents the node for which this trust evidence                To obtain desired trust evidences hosted by other nodes, a
is produced, TV is the trust value of target node computed by              node generates several artificial ants. The probabilistic
the provider and finally time is the last update time of the trust         movement of the ant allows it to explore new paths and find the
evidence. Trust evidences are locally stored by observer nodes.            proper trust evidence provider. During the trust evidence
                                                                           discovery period, Forward ants (Fa) and backward ants (Ba)
    In the proposed framework as it applies to the data packet             are used. Fa is generated by trust evidence requester to explore
delivery functionality of DSR protocol, whenever a node needs              a path to a proper trust evidence provider. Ba which contains a
to choose among available paths to communicate with a given                relevant piece of trust evidence is generated from the trust
destination, it first evaluates the reliability of each available          evidence provider and routes back to the requester.
path and consequently chooses the most reliable one. Path
reliability is computed as the probability that a packet won’t be             The formats of Fa and Ba packets are shown in Fig. 1. The
dropped by the nodes along the route and will be safely                    Fa packet contains RID – requester’s ID, TID – target’s ID (the
delivered to its destination.                                              node for which we are interested to obtain trust evidences),
                                                                           SeqN – the unique sequence number, TTL – the maximum
       To compute reputation scores, a node first employs the              number of intermediate nodes allowed to forward the Fa
trust evidence discovery protocol to collect relevant trust                packet and pass list– the dynamically increasing list which
evidences and then applies the reputation computation model to             consists of the passed nodes’ IDs. In the Ba packet PID is the
combine multiple pieces of independent trust evidences                     ID of trust provider node which creates the backward ant and
collected from other nodes. The details of these two steps are             TimeStamp is the creation time of the Ba packet.
discussed in following subsections.

B. Trust Evidence Discovery Protocol                                               RID     TID         SeqN          TTL           Pass List …
    Although there exist some literature on trust evidence
discovery in P2P networks [21,22], very little attention has                                                   (a)
been paid to exclusive study of trust evidence                                    RID      PID         TimeStamp               Pass List …
discovery/distribution problem in MANETs. Typical
approaches for trust evidence discovery in P2P networks rely                                                   (b)
on either flooding or centralized storage. The flooding
                                                                                           Figure 1. (a) FA packet         (b) BA packet
approach imposes efficiency and scalability problems and the
centralized storage approach is against the decentralized and                 Along the path of delivering requested trust evidences,
infrastructure-less nature of MANETs and also imposes                      backward ants modify the information stored in the trust
robustness risks.                                                          evidence table (TET) of each node. The structure of trust
    Almost all of the trust establishment schemes that utilize             evidence table (TET) is shown in Fig. 2.
trust information sharing mechanisms take a proactive
approach, where nodes periodically broadcast their first-hand
trust information to their neighbors. This approach also suffers                                   N1            N2            …          Nm
scalability, efficiency and uneven distribution of trust                                 TE1     P11           P12         …            P1m
evidences across the network.                                                            TE2     P21           P22         …            P2m

    Here we introduce an efficient on-demand ant-based trust                             …       …             …           …            …
evidence discovery protocol. Our ant-based scheme uses the                               TEn     Pn1           Pn2         …            pnm
swarm intelligence paradigm [23]. The swarm intelligence                                    Figure 2. Trust Evidence Table (TET)
paradigm is inspired from artificial ant colonies techniques to
solve combinatorial optimization problems [24]. The main                       Each row in TET corresponds to trust evidence of a node.
principle behind the interaction in a swarm is called stigmergy            For each trust evidence TEn and for each neighbor node i, the
– indirect communication through the environment. An                       probability value p ni expresses the probability of choosing
example of stigmergy is pheromone laying on the trails
                                                                           node i as the next hop when searching for trust evidence n and
followed by ants. Ants are attracted to pheromones and thereby
                                                                           is calculated by the formula (2):
they tend to follow the trails that have high pheromone
concentrations.




                                                                     254                                      http://sites.google.com/site/ijcsis/
                                                                                                              ISSN 1947-5500
                                                          (IJCSIS) International Journal of Computer Science and Information Security,
                                                          Vol. 8, No. 4, July 2010



            p ni
                            if i ∈ N                                      Route Reply message to the initiator node on the reverse path.
     Pni =  j∑ p nj
                                                                          Each intermediate node that forwards the Route Reply message
              ∈N                                         (2)               also checks the list of nodes contained in the route record
                                                                          (excluding the source and destination nodes) and appends
           0
                               otherwise                                  relevant trust evidences by referring to its own TER.
                                                                           Intermediate nodes avoid appending repetitive trust evidences
    In (2), N is the neighbor node set of current node and p ni
                                                                           and also replace recorded evidences if they have more recent
is the amount of pheromone on the link between current node                versions of those evidences in their TER. Also a node
and node i for trust evidence n. During the trust evidence                 forwarding the Route Reply message adds useful trust
discovery process, p ni is updated using the following formula:            evidences to its own TER.
                                                                              Explicit mode: In situations where a source node has
    p ni = (1 − α ). p ni + ∆p nj                         (3)              multiple routes to a given target of communication, but due to
                                                                           the lack of adequate trust evidences for the nodes along the
    Where 0 < α < 1 is the pheromone evaporation parameter,                routes cannot effectively evaluate the reliability of available
                                                                           routes, it follows the following procedure:
 ∆p ni is the increment amount of p ni and is determined by
information contained in the received Ba and is calculated                    1) The source node creates a forward ant Fa and broadcasts
using the following formula:                                                     it to its neighbors.
                                                                              2) Each neighbor node receiving the Fa searches its first
    ∆p ni = r − m h − n                                   (4)                    hand trust evidence storage. If a relevant piece of
                                                                                 evidence is found, a backward ant Ba containing the
   In the above formula, r is the recency of the trust evidence                  discovered trust evidence will be generated and will
contained in the received Ba, h is the hop count the ant have                    retrace the path of the Fa back to the source. As the Ba
passed by from its source to the current node. m and n are                       moves on its path, the intermediate nodes will update
parameters which determine the relative importance of trust                      their TET using the formula (4) and will store the
evidence recency versus hop count.                                               evidence in their TER.
    To improve the performance of the trust evidence discovery                3) After decreasing the TTL value of the received Fa, if it is
protocol, discovered evidences are cached in trust evidence                      still greater than zero, then the current node will unicast
repository (TER) of every node on the path of backward ants.                     the Fa to the neighbor with the highest probability by
Therefore after a period of adaptation, the request overhead                     consulting its TET. If there is no preference to the
will be drastically reduced, since probability of obtaining                      neighbors, i.e. there is no entry in the TET for this
required evidences from neighbors would increase. The                            evidence, the Fa will be broadcasted to all neighbors.
replication procedure assures the availability of trust evidences,               This happens either when no path to the requested trust
even when some origins may be out of reach. Upon receiving                       evidence has been explored or the information of the
fresher trust evidences, cached evidences are updated. The                       node is outdated. Nodes discard repetitive Fa packets by
cached trust evidence TE i , j provided by node i about node j                   checking the sequence number of received packets.
will be deleted from the TER of the current node if a more                    4) The requester node waits for a predefined period of time
recent evidence is not received from node i about node j in a                    in order to get relevant trust evidences from other nodes.
fixed time interval ∆t .                                                         Once the requested evidences are received, the requester
                                                                                 applies the reputation computation model to combine
    In the proposed trust establishment framework as it applies                  evidences related to each node to form an overall
to the data packet delivery functionality of DSR protocol,                       reputation score for each node.
relevant evidences are collected by the requester nodes in two
following modes:                                                              The next section discusses the details of the reputation
                                                                           computation model.
    Implicit mode: In this mode, the trust evidence discovery
process is incorporated into the route discovery mechanism of              C. Reputation Computation Model
DSR protocol. Here, in addition to standard fields, each Route
                                                                               In the proposed framework, nodes utilize the trust evidence
Reply message contains a field (trust evidence record) specially
                                                                           discovery protocol discussed in previous section to obtain
considered for recording relevant trust evidences. Before
                                                                           relevant trust evidences from other nodes. The obtained trust
receiving a Route Request message at the target node, the route
                                                                           evidences are combined to form an overall reputation score for
discovery process is performed according to the standard
                                                                           each node. Combination of trust evidences from other nodes
specifications of DSR protocol. When the target node received
                                                                           that can vary in their reliability or trustworthiness is a
the Route Request message, it checks the list of all intermediate
                                                                           challenging task and has a significant impact on the overall
nodes contained in the route record of the received message,
                                                                           effectiveness of trust establishment framework. Previous
searches its TER, extract evidences related to those nodes and
                                                                           approaches have used simplistic combination techniques such
in addition to the standard route record, it appends these
                                                                           as averaging or majority voting [18,19]. Here, we employ the
evidences to the trust evidence record of the Route Reply
                                                                           Dempster-Shafer evidence theory which offers an alternative to
message that it creates. The target node then sends back the



                                                                     255                              http://sites.google.com/site/ijcsis/
                                                                                                      ISSN 1947-5500
                                                             (IJCSIS) International Journal of Computer Science and Information Security,
                                                             Vol. 8, No. 4, July 2010




traditional probabilistic theory for the               mathematical                The plausibility function constitutes the upper bound of the
representation of uncertainty and is well suited      to our context           interval and represents the weight of evidence that doesn’t
where doubt and uncertainty are inherent. The         theory and its           refute A.
applicability to reputation computation are            discussed in
subsequent subsections.                                                          2) Dempster’s Rule for Combination

  1) Dempster-Shafer Theory of Evidence                                            Suppose m1(A) and m2(A) are the basic probability
                                                                               assignments from two independent observers (in the same
    Dmpster-Shafer Theory (DST) is a mathematical theory of                    frame of discernment). The combination (called the joint m12)
evidence. The seminal work on the subject is [25], which is an                 is calculated from the aggregation of two bpa’s m1 and m2 in
expansion of [26]. The theory’s practical appeal is largely due                the following manner:
to Dempster’s rule for combining beliefs based on independent
pieces of evidence. In a finite discrete space, Dempster-Shafer                                    ∑      m 1 (B ) m 2 (C )
                                                                                                                                                           (9)
theory can be interpreted as a generalization of probability                    m12 ( A ) =    B ∩C = A
theory where probabilities are assigned to sets as opposed to                                              1− k
mutually exclusive singletons. Let X be the universal set: the
set of all states under consideration. The power set, P ( X ) , is the         Where
set of all possible sub-sets of X , including the empty set. Any
hypothesis A will refer to a subset of power set for which                      K =       ∑        m1 (B ) m 2 (C )                                       (10)
observers can present evidence.                                                          B ∩C =∅

        There are three important functions in Dempster-Shafer
                                                                                   The denominator in Dempster’s rule is a normalization
theory: the basic probability assignment function (bpa or m),
                                                                               factor and represents the basic probability mass associated with
the Belief function (Bel), and the Plausibility function (Pl). The
                                                                               conflict.
bpa, represented by m, defines a mapping of each subset of the
power set to the interval between 0 and 1. Formally,                             3) Dempster’s Rule                    for    Combination      Applied       to
 m : P (X ) → [0,1] where it verifies two axioms. First, the mass              Reputation Computation
of the empty set is zero:
    m (∅ ) = 0                                                (5)                  We apply the Dempster’s rule to combine multiple pieces
                                                                               of independent trust evidences collected from other nodes. In
   Second, the summation of the bpas of all the subsets of the                 our context, the power set has three focal elements: hypothesis
power set is 1:                                                                 H ={ } that characterizes the trust degree of a given node,
                                                                                     T
                                                                               hypothesis H = { } that characterizes the distrust degree of a
                                                                                                 T
      ∑
    A ∈P ( X )
                 m (A ) = 1                                   (6)              given node and universe hypothesis U ={ ,T } that   T
                                                                               characterizes the degree of belief that a given node is either
                                                                               trusted or distrusted.
    The value of the bpa for a given set A (represented as
m(A)), expresses the proportion of all relevant and available                      For a simple illustration of how trust evidences are
evidence that supports the claim that a particular element of                  combined using Dempster’s rule, consider that nodes A and B
 X (the universal set) belongs to the set A but to no particular               are offering trust evidences on node S. Assume that node A
subset of A. From the basic probability assignment, the upper                  claims that trust and distrust values for S are 0.8 and 0.2
and lower bounds of an interval can be defined. This interval                  respectively and B claims that these values are 0.2 and 0.8
contains the precise probability of a set of interest (in the                  respectively (according to its own observations or maliciously).
classical sense) and is bounded by two nonadditive continuous                  These two pieces of trust evidence are formalized as follows:
measures called Belief and Plausibility. The Belief function
                                                                                   m A (T ) = 0.8
(Bel) maps a hypothesis A to a value between 0 and 1and is
defined as follows.                                                                m A (T ) = 0.2                                                    (11)
    Bel ( A ) =        ∑
                      B |B ⊆ A
                                 m (B )                       (7)
                                                                                   m A (U ) = 1 − (m A (T ) + m A (T )) = 0


                                                                                   m B (T ) = 0.2
    The belief function constitutes the lower bound of the                         m B (T ) = 0.8
interval and represents the weight of evidence supporting A’s                                                                                        (12)
provability. The plausibility function maps each hypothesis A                      m (U ) = 1 − (m B (T ) + m (T )) = 0
                                                                                     B                             B
to a value pls(A) between 0 and 1and is defined as follows.
                                                                                  And the combination is computed as follows:
    pl ( A ) =         ∑
                    B | B ∩ A ≠∅
                                   m (B )                     (8)




                                                                         256                                       http://sites.google.com/site/ijcsis/
                                                                                                                   ISSN 1947-5500
                                                                                  (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                  Vol. 8, No. 4, July 2010




                     mA (T )mB (T ) + mA (T )mB (U ) +mA (U )mB (T ) 0.16           (13)         and a pause time of 50 seconds. Simulations run for 900
mA (T ) ⊕mB (T ) =                                                  =      =0.5                  seconds.
                          1− m (T )m (T ) + m (T )m (T )
                                                           
                                                                      0.32
                                A     B        A      B    
                                                                                                 B. Simulation Results
                                                                                                     We use the following metrics for evaluation of the proposed
    Dempster’s rule for combination is a commutative and                                         trust establishment framework:
associative rule and therefore for any arbitrary number of bpas
we can compute the combination by first combining any pair of                                       Success rate: the percentage of requests for which the
bpas and then combining the result with the remaining bpas in                                    requester successfully obtains the relevant evidence(s). In
the same way.                                                                                    simulation, it is the number of unique forward ants sent by the
                                                                                                 requester nodes over the total number of corresponding
    Even though in this paper we assume that all nodes are                                       backward ants received by those nodes.
completely reliable with respect to offering accurate trust
evidences, a significant advantage of utilizing Dempster’s rule                                      Throughput: In our context, throughput is defined as the
for combination is its ability to effectively discount the impact                                ratio of the number of packets received by the application layer
of evidences obtained from unreliable sources in the computed                                    of destination nodes to the number of packets sent by the
reputation score.                                                                                application layer of source nodes.
    An honesty coefficient with a value between 0 and 1 for the                                      Average Latency: this metric reflects the overhead
collected evidences can be utilized for this purpose. A value of                                 imposed by the proposed scheme. The metric is defined as the
0 for a given node represents its complete dishonesty and                                        mean time (in seconds) taken by the packets to reach their
completely neutralizes the impact of the trust evidence                                          respective destinations.
provided by that node in the combination rule. Conversely, a                                          Fig. 3 presents the success rate performance results. As it
value of 1 for a given node represents its absolute honesty and                                  can be observed, except at the beginning of the simulation that
maximizes the impact of the trust evidence provided by that                                      still trust evidences are not available and entries in trust
node in the combination rule. To exemplify this, suppose that                                    evidence tables of participating nodes aren’t accurately
in the previous example, instead of absolute honesty, the                                        updated, the success rate for requested trust evidences is low.
honesty coefficient of node B was 0.8. So, we would have:                                        However as the simulation proceeds a fast convergence is
      m (T ) = 0.8 × 0.2 = 0.16                                                                  achieved at the cost of using broadcast requests for finding
         B
                                                                                                 desirable trust evidences.
      m B (T ) = 08. × 0.8 = 0.64                                                 (14)
      m B (U ) = 1 − (m 2 (T ) + m 2 (T )) = 0.2


     And the combination rule would yield:
         (0.8 × 0.16) + (0.8 × 0.2)    0.288
                                     =       = 0.63                               (15)
             (
      1 − (0.8 × 0.64) + (0.16 × 0.2) 0.465 )
    As it can be easily verified, the impact of B’s trust evidence
in the combination rule has been weakened and A’s evidence
has been more influential in the gained result.

              IV. SIMULATION AND EVALUATION                                                                    Figure 3. Success rate performance results

   The Performance of the proposed framework has been                                                Notice that in fig. 3 the success rate of the proposed scheme
evaluated using some simulations. The simulation model and                                       increases and decreases repeatedly by a small amount. The
gained results are discussed in following subsections.                                           reason is the mobility of nodes hosting requested trust
                                                                                                 evidences and abolishment of trust evidences with the passage
A. Simulation Model                                                                              of time.
    To evaluate the effectiveness and efficiency of the proposed
                                                                                                     To investigate the effectiveness of the proposed framework
framework, we have conducted some simulations according to
multiple scenarios. We have used NS-2 for simulation purpose.                                    as it applies to the data packet delivery functionality of DSR
                                                                                                 protocol; we use the throughput metric. For this purpose, a
All simulations are in an ad hoc network consisting of 50 nodes
spread uniformly through a 1000×1000 meter square area.                                          varying number of selfish nodes that drop their received data
                                                                                                 packets destined to other nodes with a probability between 60
Nodes are equipped with an IEEE 802.11 radio network
interface, operating at 11Mbps data rate with a 250m                                             and 100 percent are implemented in simulations.
transmission range. There exist a total of 30 CBR connections                                       Simulations have been conducted for two different
and sending nodes send data packets of size 512b at 4pk/s rate.                                  scenarios. In each scenario the ratio of selfish nodes ranges
Nodes move according to the Random Waypoint mobility                                             from 0 to 50 percent. The two scenarios are: 1) when standard
model with speed uniformly distributed between 0 and 10m/s



                                                                                           257                                http://sites.google.com/site/ijcsis/
                                                                                                                              ISSN 1947-5500
                                                         (IJCSIS) International Journal of Computer Science and Information Security,
                                                         Vol. 8, No. 4, July 2010




DSR protocol is used for route selection decisions and                                           V. CONCLUSION
communications among nodes. 2) When the proposed scheme                      We presented a trust establishment framework that utilizes
is employed to improve route selection decisions.                        both first and second-hand observational data. An on-demand
   Fig. 4 presents the throughput performance results of the             ant-based scheme was introduced for efficient distribution of
two scenarios.                                                           trust evidences. The flexibility of the formula used for choosing
                                                                         the best next hop for obtaining the requested trust evidences
                                                                         provides the possibility of embedding more complicated
                                                                         metrics such as node mobility, provider’s trustworthiness and
                                                                         security related items.
                                                                                 Dempster's rule for combination which offers an
                                                                         effective and robust mechanism for combination of trust
                                                                         evidences collected from other nodes and quantification of
                                                                         reputation was used. We used experimental simulations to
                                                                         demonstrate the effectiveness of the proposed framework as it
                                                                         applies to the data packet delivery functionality of DSR
                                                                         protocol. However, we believe that the proposed framework is
                                                                         a general framework that may be adapted for a variety of
                                                                         scenarios where nodes have to rely on unreliable nodes to
                                                                         accomplish a cooperative service.

                                                                                           ACKNOWLEDGMENT
              Figure 4. Throughput performance results
                                                                             This research   has been      supported by Iran
    The performance results reveal that in the presence of a             Telecommunication Research Center (ITRC) under contract
varying percentage of selfish nodes, the proposed scheme                 19230/500 and herein authors are willing to express their
results in a better throughput. The main reason is that in               gratitude.
standard DSR protocol, by default shorter paths are preferred
for communication. However, in the proposed scheme,                                                     REFERENCES
whenever a node needs to communicate with another node, it
                                                                         [1]  Chlamtac, I., Conti, M. and Liu, J.N. (2003). Mobile ad hoc networking:
first evaluates the reliability of each available path and                    imperatives and challenges. Ad Hoc Networks, 1, 13-64.
consequently chooses the most reliable one. Path reliability is          [2] Wu, B., Chen, J., Wu, J. and Cardei, M. (2007). A Survey of Attacks and
defined as the probability that a packet won’t be dropped by the              Countermeasures in Mobile Ad Hoc Networks. In: Xiao, Shen and Du
nodes along the route and will be safely delivered to its                     (Eds), Wireless Network Security, 103-136.
destination.                                                             [3] Li, J., Li, R. and Kato, J. (2008). Future Trust Management Framework.
                                                                              Communications Magazine IEEE, 46(4), 108-114.
  The comparison results for average latency of standard                 [4] Buchegger, S. and Le Boudec, J. Y. (2002). Performance Analysis of the
DSR and the proposed scheme are shown in Fig 5.                               CONFIDANT Protocol (Cooperation Of Nodes Fairness in Dynamic
                                                                              Ad-hoc NeTworks). Proc. ACM MobiHoc 2002, Atlanta, GA.
                                                                         [5] Buchegger, S. and Le Boudec, J. Y. (2004). A Robust Reputation
                                                                              System for P2P and Mobile Ad-Hoc Networks. Proc. P2PEcon 2004,
                                                                              Harvard Univ., Cambridge, MA.
                                                                         [6] S. Marti et al., “Mitigating Routing Misbehavior in Mobile Ad Hoc
                                                                              Networks,” Proc. MobiCom 2000, Aug.2000, pp. 255–65.
                                                                         [7] Sun, Y., Yang, Y. (2006). A Trust Evaluation Framework in Distributed
                                                                              Networks: Vulnerability Analysis and Defense Against Attacks. Proc.
                                                                              IEEE INFOCOM 2006, Barcelona, Spain.
                                                                         [8] Theodorakopoulos, G. and Baras, S. (2006). On Trust Models and Trust
                                                                              Evaluation Metrics for Ad Hoc Networks. Selected Areas in
                                                                              Communications, IEEE, 24(2), 318-328.
                                                                         [9] Zouridaki, C., Mark, B.L., Hejmo, M., Thomas, R.K. (2007). Hermes: a
                                                                              quantitative trust establishment framework for reliable data packet
                                                                              delivery in MANETs. Journal of Computer Security, 15 (1), 3–38.
                                                                         [10] Ganeriwal, S. and Srivastava, M. (2004). Reputation-based Framework
                                                                              for High Integrity Sensor Networks. Proc. ACM Wksp. Sec. Ad Hoc and
                  Figure 5. Average latency results                           Sensor Networks, Washington, DC.
                                                                         [11] Zouridaki, C., Mark, B.L., Hejmo, M. and Thomas, R.K. (2005). A
   As expected, considering the fact that the trusted paths                   Quantitative Trust establishment Framework for Reliable Data Packet
chosen by the proposed scheme are not necessarily optimal in                  Delivery in MANETs. Proc. 3rd ACM Wksp. Sec. Ad Hoc and Sensor
                                                                              Networks, 1-10.
terms of the number of hops, compared to the standard DSR
protocol the average latency has been increased. However, the            [12] Eschenauer, L., Gligor, V.D., Baras, J. (2002). On trust establishment in
                                                                              mobile ad-hoc networks. proc. Wksp. Security, vol. 2845, LNCS, 47–66.
achieved amelioration of throughput compensates for the
imposed overhead.



                                                                   258                                    http://sites.google.com/site/ijcsis/
                                                                                                          ISSN 1947-5500
                                                                  (IJCSIS) International Journal of Computer Science and Information Security,
                                                                  Vol. 8, No. 4, July 2010



[13] Pirzada, A.A., McDonald, C. (2006). Establishing trust in pure ad-hoc
     networks. Wireless Personal Communications, 37, 139-163.
[14] Theodorakopoulos, G., Baras, J.S. (2004). Trust evaluation in ad-hoc
     networks. proc. ACM Wksp. Wireless Security (WiSe’04), 1–10.
[15] Buttyan, L., Hubaux, J.P. (2003). Stimulating cooperation in self-
     organizing mobile ad hoc networks. Mobile Networks and Applications,
     8 (5), 579–592.
[16] Huo, H., Gao, D., Niu, Y. and Gao, S. (2007). ASDP: An Action-Based
     Service Discovery Protocol Using Ant Colony Algorithm in Wireless
     Sensor Networks, Lecture Notes in Computer Science 4864, 338-349.
[17] Jiang, T. and Baras, J.S. (2004). Ant-based Adaptive Trust Evidence
     Distribution in MANET. Proc. 24th International Conference on
     Distributed Computing Systems Workshops (ICDCSW’04), 588-593.
[18] Kargl, F., Klenk, A., Weber, M. and Schlott, S. (2004). Sensors for
     Detection of Misbehaving Nodes in MANETs. Proc. Detection of
     Intrusion and Malware and Vulnerability Assessment.
[19] Zhang, Y. and Lee, W. (2000). Intrusion Detection in Wireless Ad-Hoc
     Networks. Proc. 6th Ann. ACM Int’l Conf. Mobile Computing and
     Networking, ACM Press, 275–283.
[20] Johnson, D.B., Maltz, D.A. and Hu, Y. (2003). The Dynamic Source
     Routing Protocol for Mobile Ad-Hoc Networks (DSR), IETF MANET,
     Internet Draft (Work in Progress).
[21] Clarke, I., Sandberg, O., Wiley, B. and Hong, T.W. (2000). Freenet: A
     distributed Anonymous Information Storage and Retrieval System. In
     Proc. ICSI Workshop on Design Issues in Anonymity and
     Unobservability, Berkeley, CA.
[22] Babaoglu, O., Meling, H. and Montresor, A. (2002). Anthill: A
     Framework for the Development of Agent-Based Peer-to-Peer Systems.
     In Proc. 22nd ICDCS, Vienna, Austria.
[23] Bonabeau, E., Dorigo, M. and Theraulaz, G. (1999). Swarm Intelligence
     – From Natural to Artificial Systems. New York: Oxford University
     Press.
[24] Caro, G. D. and Dorigo, M. (1998). AntNet: Distributed Stigmergetic
     Control for Communications Networks, Journal of Artificial Intelligence
     Research, 9, 317-365.
[25] G. Shafer, A. (1976). Mathematical Theory of Evidence. Princeton:
     Princeton Univ. Press.
[26] Dempster, A. (1967). Upper and Lower Probabilities Induced by a
     Multivalued Mapping. Ann.Mathematical Statistics, 38(2), 325–339.




                                                                               259                           http://sites.google.com/site/ijcsis/
                                                                                                             ISSN 1947-5500

				
DOCUMENT INFO
Description: The International Journal of Computer Science and Information Security is a monthly periodical on research articles in general computer science and information security which provides a distinctive technical perspective on novel technical research work, whether theoretical, applicable, or related to implementation. Target Audience: IT academics, university IT faculties; and business people concerned with computer science and security; industry IT departments; government departments; the financial industry; the mobile industry and the computing industry. Coverage includes: security infrastructures, network security: Internet security, content protection, cryptography, steganography and formal methods in information security; multimedia systems, software, information systems, intelligent systems, web services, data mining, wireless communication, networking and technologies, innovation technology and management. Thanks for your contributions in July 2010 issue and we are grateful to the reviewers for providing valuable comments. IJCSIS July 2010 Issue (Vol. 8, No. 4) has an acceptance rate of 36 %.