Learning Center
Plans & pricing Sign in
Sign Out



									                                                                                   International Journal of Computer Information Systems,
                                                                                                                       Vol. 2, No. 1, 2010

          An Intelligent AntNet-Based Algorithm for
        Efficient Secure Data Routing over Peer to Peer
                M. Nofal, S.F. El-Zoghdy                                                        M. Hadi
Prof., Comp. Eng. Dept, Assit. Prof., Comp. Science Dept.                                       Lecturer
    College of Computers & Information Technology                                        College of Engineering
                     Taif University,                                                        Taif University
             Taif, Kingdom of Saudi Arabia                                           Taif, Kingdom of Saudi Arabia,

Abstract— In this paper, we evaluate the performance of the            distributed and potentially heterogeneous platforms, and
AntNet routing algorithm in terms of efficiency and security in        provide high availability by eliminating the need for a single
peer-to-peer networks. Using the network simulator NS2, a              centralized component.
simulator is implemented for a network of 8-nodes which
simulates the ant colony intelligence in deciding the most                 Routing in computer networks refers to the process of
efficient and secured path from source to destination                  discovering, selecting, and employing paths from one node to
nodes. In the simulator, a variety of network scenarios is             another in the network. Routing involves two basic activities:
considered. The AntNet routing algorithm has the ability to            determining optimal routing paths and transporting information
detect faulty and malicious nodes and abandon them from the            packets through an inter-network. Routing protocols use
routing tables of the intermediate nodes. Hence, a healing is          metrics to evaluate what path will be the best for a packet to
established so that packets are efficiently and securely               travel. Routing determines the overall performance of a
transferred from source to destination. In addition, the               network in terms of throughput and transmission delay.
probabilistic nature of AntNet routing distributes the network         Routing process involves building forward tables, one for each
traffic of data packets over several optimal and suboptimal paths      node in the network, which tell incoming data which link to use
which lead to improve the network performance and minimize             to continue their travel towards the destination node [4]. While
the packet latency.                                                    directing traffic from source to destination, the goal of the
                                                                       routing algorithm is to maximize the network performance and
Keywords- peer to peer networks; Network Routing; AntNet-Based         minimize costs [4-8].
Algorithm; Performance Evaluation.
                                                                           A P2P system can take many forms [8]. Email, Internet
                                                                       Relay Chat and Napster are all examples of P2P systems.
                        I.   INTRODUCTION
                                                                       Routing on these networks is either centralized or statically
    The P2P network comprises a collection of nodes that can           configured and is therefore unproblematic. Another class of
cooperate and collaborate with each other to offer opportunities       P2P networks is the overlay network. Overlay networks build a
for real-time communication, collaboration and information             virtual topology on top of the physical links of the network.
sharing in a large-scale decentralized and distributed manner. A       Nodes leave and join this network dynamically and the average
node in a P2P network can access information present in the            uptime of individual nodes is relatively low. The topology of
network using peer discovery followed by a search and                  an overlay network may change all the time. Once a route is
retrieval phase. The most distinct characteristic of P2P               established, there is no guarantee of the length of time that it
computing is that there is a symmetric communication between           will be valid. Routing in these networks is therefore very
the peers; each peer has both a client and a server role. Both         problematic [2, 3].
parties have the same capabilities and both can initiate a                 Great interest was devoted to the routing problem and
connection. It is the opposite of the client/server model in the       several routing algorithms were proposed. Static routing
sense that there is no central entity that the other parties contact   determines the path taken by a packet on the basis of the source
but every single entity is able to initiate a connection directly      and destination without regard to the current network state
with all other entities [1, 2, 3].                                     [4,5]. This path is usually chosen as the shortest one according
    The advantages of the P2P systems are multi-dimensional.           to some cost criterion. Adaptive routing, on the other hand,
They improve scalability by enabling direct and real-time              adapts the routing policy to the varying traffic conditions. Open
sharing of services and information. In addition, these networks       shortest path first (OSPF) is one of the most widely used
enable knowledge sharing by aggregating information and                routing protocols. In OSPF, every node gathers information
resources from nodes that are located on geographically                about the entire network and calculates the best path to each
                                                                       destination. For every destination, the interface to this path is

       January Issue                                           Page 20 of 67                                      ISSN 2229 5208
                                                                                     International Journal of Computer Information Systems,
                                                                                                                         Vol. 2, No. 1, 2010
saved in the routing table. While OSPF minimizes the static                              III.   THE ANTNET ALGORITHM
link cost, it cannot react to the dynamic nature of the network.           Unlike traditional routing algorithms such as OSPF, AntNet
While OSPF protocol is a static, deterministic algorithm,              dropped the deterministic way of routing and introduces a
AntNet [4, 6, 9, 10] on the other hand introduces a dynamic,           probabilistic approach. An overview of the entire network is no
probabilistic approach. AntNet is a dynamic algorithm for              longer needed [1, 4, 6, 8]. The link state database of AntNet is
packet routing in communication networks, originally proposed          smaller than the database of OSPF as an AntNet node only
by M. Dorigo and G. Di Caro in 1997. In AntNet, a group of             needs to keep information about the links between itself and the
mobile agents (or artificial ants) build paths between pair of         adjacent node, and not about all the nodes in the network.
nodes; exploring the network concurrently and exchanging               Routing decisions are based on the basis of local and
obtained information to update the routing tables.                     approximate information about the current and the future
    In this discourse, we evaluate the performance of the              network states.
AntNet routing algorithm in terms of efficiency and security in            Each artificial ant builds a path from its source to its
peer-to-peer networks. Using the network simulator NS2 [11,            destination node. While building the path, it collects explicit
12], a simulator is implemented for a network of 8-nodes which         information about the time length of the path components and
simulates the ant colony intelligence in deciding the most             implicit information about the load status of the network [1,4,
efficient and secured path from source to destination. In the          6]. This information is then back propagated by another ant
simulator, a variety of network scenarios is considered.               moving in the opposite direction and is used to modify the
    This paper is organized as follows: Section II describes the       routing tables of the visited nodes. AntNet system comprises
Ant colony optimization methodology. Section III presents the          two sets of homogeneous mobile agents called forward and
AntNet routing algorithm. In Section IV, we explain the                backward ants [7, 8]. They possess the same structure but they
simulation model and assumptions. Section V describes the              can sense different inputs and they can produce different
results of numerical examination for the performance of the            independent outputs. Forward ants gather information. On a
AntNet routing algorithm. Finally, Section VI summarizes this          regular time base, every node sends one forward ant with a
paper.                                                                 random destination over the network. This forward ant is
                                                                       forward by some intermediate nodes to its final destination, in a
             II.   THE ANT COLONY OPTIMIZATION                         way that balances between exploitation of known good paths
                                                                       and the exploration of new, possibly better paths. Backward
    In ant colony, ants are able to find the shortest path between     ants are created out of forward ants once they have reached
a food source and their home colony [7]. Real ants                     their destination. The backward ant follows exactly the same
communicate via pheromones. The ant deposits a certain                 path as the forward ant but in the opposite direction resulting in
amount of pheromone when it walks. An ant tends to choose a            adapting the probabilities in the routing table of all
path positively correlated to the pheromone intensity of found         intermediate nodes.
trials. The pheromone trail evaporates over time [5, 6, 8]. If
many ants choose a certain path and lay down pheromones, the                   IV.      SIMULATION MODEL AND ASSUMPTIONS
intensity of the trail increases. Thus, this trail attracts more and
more ants; a process that results in ant highway following the             In order to gauge the behavior of AntNet routing algorithm
shortest pat. Ants also have the ability to adapt to the               in our P2P network, the instance of the network is mapped on a
environmental changes, for example, finding the new and                directed graph with N nodes. All the links in the network are
shortest path once the old one is no longer feasible due to a          considered bidirectional and specified by a transmission
new obstacle.                                                          capacity and a transmission delay. Each node is considered a
                                                                       communication end-point (host) and a forwarding unit (router).
    Ant colony optimization ACO mimics in software the                 Every node in the network maintains an input buffer composed
behavior of real ants in colony. In applying ACO in network            of a single queue and an output buffer composed of a high
routing, an artificial ant is typically realized as a simple           priority queue and a low priority queue for each neighbor or
program consisting of simple procedures that simulate the              outgoing link. The high priority queue is served before the low
laying and sensing of pheromone, and data structures that              priority queue. All the packets within the network can be
record trip times and the nodes that it passes [7]. Moving from        divided into two different classes:
one node to another, an artificial ant emulates laying of
pheromone by updating the corresponding entry in the routing                  Data packets: represent the information that the end-
table in a node which records, for example, the number of ants                 users exchange with each other. In ant-routing, data
that pass that node. In ant colony based algorithms, a set of                  packets do not maintain any routing information but
artificial ants move on the graph which represents the instance                use the information stored at routing tables for
of the network. While moving they build solutions and modify                   traveling from the source to the destination node.
the problem representation by adding collected information.                   Mobile agents (forward ants and backward ants): are
The most important application of ACO is network routing as it                 used to update the routing tables and distribute
is desired to transfer data packets from a source to a destination             information about the traffic load in the network.
in an efficient way.

       January Issue                                           Page 21 of 67                                        ISSN 2229 5208
                                                                                        International Journal of Computer Information Systems,
                                                                                                                            Vol. 2, No. 1, 2010
    Backward ant packets have a higher priority than the data
and forward ant packets and are thus stored in the high priority
queue [8, 13], while data and forward ant packets are stored in
the low priority queue. We assume that all the packets in the
low priority queue and the high priority queue in the output
buffer are served in a FIFO order. Further, the maximum
number of packets stored in the input buffer or output buffer is
limited by the size of the buffer. We have assumed that the
buffer size is sufficiently large to neglect buffer overflow.
When a node receives a packet from a neighbor, the packet is
first stored in the input buffer and then served in a FIFO order
or according to a different scheduling rule. After the packet has
been served, the packet is sent to the output buffer. Within the
output buffer, the packet goes to a particular queue for a
particular outgoing link based on the type of the packet and the
next node.
     We also assume that all links are of equal, unit cost. To                 Figure 2. Simulation snapshot of the referenced network scenario.
move from one node to a neighboring one, the packet suffers a
transfer delay that depends on its size as well as link
transmission characteristics. According to the information                  To calculate the average pheromone level, the simulation
content in the routing table, the packet follows its path toward       program is run many times. A sample of the routing tables at
its target node. When a link is available, it is reserved and the      the participating nodes as well as the average pheromone levels
transfer is set up.                                                    is listed in Table 1.
    A routing module was implemented to support AntNet
functionalities [14, 15]. Artificial ants were simulated by                TABLE 1: ROUTING TABLE AT NODE 4 IN CASE 1 ( REFERENCE MODEL)
implementing two types of ant packets to represent forward
and backward ants. The ants flow through the network
                                                                                               Dest      Next         Ph-average
according to the algorithm and keep a memory of the path
                                                                                                0         5            0.057067
traversed to discover optimal solution. A routing agent was
                                                                                                0         1           0.0165905
defined to implement route logic and perform the protocols
                                                                                                0         0           0.92634275
functionality. In order to evaluate the performance of the
                                                                                                1         5           0.0358515
current scenario of AntNet routing protocol, three network
models were simulated on NS2 simulation environment.                                            1         1           0.71592475
                                                                                                1         0           0.24822375
                                                                                                2         5           0.31052175
                   V.    RESULTS AND DISCUSSION                                                 2         1           0.46374025
                                                                                                2         0           0.22573825
A.      Case 1: Reference Model                                                                 3         5           0.3833685
    We first consider a network with 8 nodes as a reference to                                  3         1           0.4343045
gauge the capability of the protocol on deciding the shortest                                   3         0            0.182327
path. No malicious nodes are assumed. Fig.1 presents the                                        5         5           0.76226375
network hierarchy. Tcl scripts for these models were developed                                  5         1            0.054009
in the simulation and the agent AntNet was implemented. The                                     5         0           0.1837275
simulator was run for 5 minutes to reach steady state. An                                       6         5           0.43687625
animated snapshot is shown in Fig. 2.                                                           6         1            0.337457
                                                                                                6         0            0.225667
                                                                                                7         5           0.3687785
     0              1                      2               3
                                                                                                7         1            0.378741
                                                                                                7         0           0.25248075

                                                                       B. Case 2: Faulty Links
     4              5                     6                7               To check the healing capability of the AntNet algorithm on
                                                                       detecting broken routes, we consider a network of 8 nodes with
                                                                       faulty links as depicted in Fig. 3. The link between nodes 0-1 is
            Figure 1. A model of a network with 8 nodes.               faulty as well as the link between nodes 0-4. A Tcl script for
                                                                       this scenario is developed and the simulation is run. An
                                                                       animated snapshot is shown in Fig. 4 where faulty routes are
                                                                       detected and bypassed by the protocol. Moreover, the routing

         January Issue                                         Page 22 of 67                                               ISSN 2229 5208
                                                                                                    International Journal of Computer Information Systems,
                                                                                                                                        Vol. 2, No. 1, 2010
table at the participating nodes ignores the faulty links and                      TABLE II.             ROUTING TABLE AT NODE 7 IN CASE 2 (FAULTY LINKS)
generates alternative routes between source and destination                                          Dest.       Next        ph-avg
nodes. The average pheromone levels at network nodes are                                              0           6        0.5981755
observed to explore this capability. It can be concluded that
                                                                                                      0           3       0.15448775
although there are faulty links in the network graph, the AntNet
                                                                                                      0           2       0.24733675
protocol is operationally active as it determines the shortest
                                                                                                      1           6       0.06803075
path to the destination nodes.
                                                                                                      1           3       0.17078675
                                                                                                      1           2       0.76118175
                                                                                                      2           6       0.05941175
  0                1                    2                      3
                                                                                                      2           3         0.218449
                                                                                                      2           2         0.722139
                                                                                                      3           6        0.1162755
                                                                                                      3           3       0.86395675
                                                                                                      3           2       0.01976775
  4                5                    6                      7                                      4           6        0.5756385
                                                                                                      4           3        0.2070205
                                                                                                      4           2         0.217341
      Figure 3. A model of network of 8 nodes with faulty links.
                                                                                                      5           6         0.678631
                                                                                                      5           3         0.150995
                                                                                                      5           2         0.170374
                                                                                                      6           6         0.893601
                                                                                                      6           3         0.099643
                                                                                                      6           2         0.006756

                                                                                        0                    1                     2                         3

                                                                                        4                    5                     6                         7

                                                                                                Figure 5.    A model of network with a malicious node.

                                                                                    Once the AntNet algorithm detects that Node 1 is
                                                                                malicious, the node is deleted from the routing tables and nodes
                                                                                are reordered as shown in Fig. 6. In addition, all links related to
                                                                                Node 1 are omitted; these are the link between Nodes 0-1 as
                                                                                well as the link between Nodes 1-2. An animated snapshot of
                                                                                case 2 scenario is shown in Fig. 7. Investigating the routing
Figure 4. Snapshot of simulation scenario of a network with faulty links.       tables of the intermediate nodes as well as the average
                                                                                pheromone levels indicates that the network is still operable.
                         Figure 5.                                              This implies that the AntNet algorithm possess the efficiency
   A sample of the routing tables at the participating nodes as                 and security properties.
well as the average pheromone levels is listed in Table 2.

                                                                                        0                    1                     1                         2

C.   Case 3: Malicious Nodes
    Finally, we check the security capability of AntNet through
detecting malicious nodes. For an algorithm to be secure, it
should drop un-trusted malicious nodes so that traffic does not                         3                    4                     5                         6
pass through them. This results in keeping information secrete
during transmission until reaching its destination. We consider
one (Node 1) of the 8 nodes to act as a malicious node as
                                                                                            Figure 6. Reconfiguration of the network with malicious nodes.
depicted in Fig. 5.

        January Issue                                                   Page 23 of 67                                                  ISSN 2229 5208
                                                                                              International Journal of Computer Information Systems,
                                                                                                                                  Vol. 2, No. 1, 2010
                                                                                 [2]    L. Lertsuwanakul and H. Unger, “An Improved Greedy Routing
                                                                                        Algorithm for Grid using Pheromone-Based Landmarks”, World
                                                                                        Academy of Science, Engineering and Technology, vol 59, 2009, pp.
                                                                                 [3]    E. Stone, T. Czerniak, C. Ryan and R. McAdoo, “Peer
                                                                                        to     Peer    Routing”,    4ba2.05/
                                                                                 [4]    G. D. Caro and M. Dorigo, “Mobile Agents for Adaptive Routing,”
                                                                                        Tech. Report IRIDIA, 1997.
                                                                                 [5]     M. Strobbe, V. Verstraete and E. V. Breusegem, “Implementation
                                                                                        and Evaluation of AntNet, a Distributed Shortest-path Algorithm”,
                                                                                        Proc.       of     Advanced       Industrial       Conference       on
                                                                                        Telecommunications/Service Assurance with Partial and Intermittent
                                                                                        Resources Conference/ E-Learning on Telecommunications
                                                                                        Workshop. AICT/SAPIR/ELETE 2005. July 2005 pp. 320 – 325.
                                                                                 [6]    G. D. Caro and M. Dorigo, “AntNet: Distributed Stigmergetic Control
                                                                                        for Communication Networks,” J. Artif. Intell. Res., vol. 9, 1998, pp.
                                                                                 [7]    K. M. Sim and W. H. Sun, “Ant Colony Optimization for Routing and
                                                                                        Load-Balancing: Survey and New Directions,” IEEE Trans. Systems,
  Figure 7. Simulation Snapshot of network with one malicious node.                     Man, and Cybernetics-Part A: Systems and Humans, vol. 33, no. 5,
                                                                                        Sept. 2003, pp. 560-572.
                                                                                 [8]    S.S. Dhillon and P. Van Mieghem, “Performance analysis of the
   A sample of the routing tables at the participating nodes as                         AntNet algorithm,” Computer Networks, vol. 51, no. 8, 2007, pp.
well as the average pheromone levels is listed in Table 3.                              2104–2125.
                                                                                 [9]    M. Castro, P. Druschel, A. Ganesh, A, Rowstron and D. S. Wallach,
                                                                                        “Secure Routing for Structured Peer-to-Peer Overlay Networks”,
 TABLE III.      ROUTING TABLE AT NODE 0 IN CASE 3 (MALICIOUS NODES)                    Proc. of the 5th Usenix Symposium on Operating Systems Design and
                                                                                        Implementation, Boston, MA, December 2002.
                  Dest     Next      ph-avg                                      [10]   E. Michlmayr, A. Pany and S. Graf, “Applying Ant-based Multi-
                   1        3       0.280625                                            Agent Systems to Query Routing in Distributed Environments,” Proc.
                   1        4       0.719375                                            of 3rd Inter. IEEE Conf. on Intelligent Systems, Sept. 2006, London,
                   2        3       0.302257                                            pp. 36-41.
                   2        4       0.697743                                     [11]   T. Issariyakul and E. Hossain, Introduction to Network Simulator 2
                                                                                        (NS2), 2009, 1-18, DOI: 10.1007/978-0-387-71760-9_2
                   3        3       0.999937
                                                                                 [12]   K. Fall and K. Varadhan, “The network simulator ns-2”,
                   3        4       0.000063                                  
                   4        3       0.221786                                     [13]   M. Dorigo and G. Di Caro, New Ideas in Optimization: The Ant
                   4        4       0.778214                                            Colony Optimization, Meta-Heuristic, McGraw-Hill, 1999.
                   5        3       0.215656                                     [14]   E. Bonabeau, M. Dorigo, and G.Theraulaz, “Inspiration for
                   5        4       0.784344                                            Optimization from Social Insect Behaviour,” Nature, July 2000, vol.
                   6        4       0.149368                                            406, pp. 39–42.
                   6        4       0.850632                                     [15]   G. Di Caro, F. Ducatelle, and L. M. Gambardella, “AntHocNet: An
                                                                                        Ant-based Hybrid Routing Algorithm for Mobile and Ad Hoc
                                                                                        Networks,” Proc. of Parallel Problem Solving from Nature, Sept.
                                                                                        2004, Springer.
                          VI.     CONCLUSION
    In this paper, the performance of AntNet routing algorithm                                             AUTHORS PROFILE
is evaluated through simulation using NS-2 Simulation
package. A variety of network scenarios with 8 nodes is
considered. The protocol is proved to be robust in the sense that
                                                                                                     Prof. Mostafa Nofal Was born in El-Gharbia, Egypt,
faulty nodes are detected and bypassed. The broken links are                                         in 1958. He received the BSc degree in Electronic
abandoned from the routing tables and a healing is established                                       Engineering in 1981, and MSc degree for his work in
so that packets are routed from their sources to destinations. In                                    Electronic and Communications Engineering in 1986,
addition, the AntNet algorithm is securable so that malicious                                        all from the Faculty of Electronic Engineering,
nodes are ruled out to avoid leakage of transmitted information.                                     Menoufia University, Egypt. In 1991, he received his
                                                                                                     Ph. D. in Wireless Networks from Menoufia
The probabilistic nature of AntNet routing distributes the                    University, Egypt in collaboration with Dept. of Electronics and Computer
stream of data packets over several optimal and suboptimal                    Science, Southampton University, UK. From 1981 to 1986, he was a
paths which improves the network performance and minimize                     demonstrator at Faculty of Electronic Engineering, Menoufia University,
packet latency                                                                Egypt. From 1986 to 1989, he was an assistant lecturer. From 1989 to 1991,
                                                                              he was a Ph. D. candidate at Dept. of Electronics and Computer Science,
                                                                              Southampton University, UK, where he was conducting research on wireless
                             REFERENCES                                       networks. In 1994 and 1999 he was a visitor professor and consultant in
  [1]   C. Zuo, R. Li, and Z. Lu, “A Novel Approach to Improve the Security   Southampton University, UK. From 2002, he worked as a professor of mobile
        of P2P File-Sharing Systems, Int. J. Communications, Network and      networks, Menoufia University, Egypt. From 2005 until now, he is working as
        System Sciences, 2009, vol. 3, pp.229-236.                            a professor at the College of Computers and Information Technology, Taif

        January Issue                                                 Page 24 of 67                                               ISSN 2229 5208
                                                                                         International Journal of Computer Information Systems,
                                                                                                                             Vol. 2, No. 1, 2010
University, KSA. His research interests include wireless networks, data
security, satellite communications and computer networks.

                        Dr. Said Fathy El-Zoghdy Was born in El-Menoufia,
                        Egypt, in 1970. He received the BSc degree in pure
                        Mathematics and Computer Sciences in 1993, and
                        MSc degree for his work in computer science in 1997,
                        all from the Faculty of Science, Menoufia, Shebin El-
                        Koom, Egypt. In 2004, he received his Ph. D. in
                        Computer Science from the Institute of Information
Sciences and Electronics, University of Tsukuba, Japan. From 1994 to 1997,
he was a demonstrator of computer science at the Faculty of Science,
Menoufia University, Egypt. From December 1997 to March 2000, he was an
assistant lecturer of computer science at the same place. From April 2000 to
March 2004, he was a Ph. D. candidate at the Institute of Information Sciences
and Electronics, University of Tsukuba, Japan., where he was conducting
research on aspects of load balancing in distributed and parallel computer
systems. From April 2004 to 2007, he worked as a lecturer of computer
science, Faculty of Science, Menoufia University, Egypt. From 2007 until
now, he is working as an assistant professor of computer science at the
Faculty of Computers and Information Systems, Taif University, Kingdom of                      .
Saudi Arabia. His research interests are in load balancing in
distributed/parallel systems, Grid computing, performance evaluation,
network security and cryptography.

                        Eng. M. Hadi Was born in Cairo, Egypt, in 1960. He
                        received     BSc     degree   in    Electronic     and
                        Telecommunication in 1983 from Faculty of
                        Engineering, Helwan University, Egypt. In 2003, he
                        received his MSc in Computer Science from Al-Azhar
                        University, Egypt. Currently, he is conducting
                        research leading to Ph. D in Computer Networks. He
                        was a consultant engineer in KSA from 1984 to 2002.
In 2004, he was a lecturer in Derna University, Libya. From 2005 to 2009, he
worked as a lecturer of computer science in Taif Teacher College, KSA. From
2010 until now, he is working as a lecturer of computer engineering at College
of Engineering, Taif University, KSA. He is a consultant for Quality
Insurance. His research interests include computer networks and computer

        January Issue                                                    Page 25 of 67                                  ISSN 2229 5208

To top