IEEE3

Full text

1
Hierarchical Dynamic Source Routing: Passive
Forwarding Node Selection for Wireless Ad Hoc
Networks
Mohammed Tarique, Student Member, IEEE, Kemal E. Tepe, Member, IEEE, and Mohammad Naserian, Student
Member, IEEE
Abstract— In this paper, Hierarchical Dynamic Source Routing
(HDSR) protocol is introduced. HDSR is derived from Dynamic
Source Routing(DSR) protocol. In HDSR, there are two states
that a mobile node can be: Mobile Node (MN) and Forwarding
Node (FN). Network nodes make distributed decisions on whether
to forward or not to forward traffic for others. FN routes
the packets and MN hosts the applications. FN selections are
made using local information and nodes solely rely on ”on
demand” routing discovery and maintenance traffic to determine
whether to act as MN or FN. Such provisions, significantly reduce
number of control messages (route request and route reply) in
compare to DSR. HDSR is implemented by a network simulator
(Network Simulator-2 of University of California). It was shown
via computer simulations that HDSR improves average network
throughput and packet delivery ratio compared to regular DSR
and provides energy efficiency.
Index Terms— Ad hoc networks, sensor networks, routing,
hierarchical routing, hierarchical networks.
I. INTRODUCTION
ROUTING protocols are an essential parts of a network
providing self-organizing capability and it is the most
widely studied element for ad hoc networks. Broadly those
protocols can be classified as: (1) Proactive routing protocols,
and (2) on demand routing protocols. In proactive routing,
routing information is periodically exchanged among network
nodes, like Dynamic Sequence Distance Vector (DSDV) [1].
Because of those periodic routing information updates and
exchanges, the network consumes large portion of the useful
bandwidth for routing overhead (i.e., route maintenance and
update packets). That is why proactive protocols do not scale
well. On the other hand, in on demand routing, routes are
discovered when they are needed. Such provision eliminates
periodic routing updates, hence allows protocols to operate
more efficiently (i.e. less routing overhead) than proactive
routing protocols. That is why most of the recent routing
protocols for ad hoc networks fall under ’on demand’ category,
like Ad hoc On Demand Distance Vector (AODV) [2] and Dy-
namic Source Routing (DSR) [3]. Although on demand routing
protocols offer lower routing overhead than proactive routing
protocols, the routing overhead during the route discovery
phase of the protocol can overwhelm the network. During that
This work was supported in part by the NSERC (Natural Sciences and
Engineering Research Council of Canada).
Authors are with the department of Electrical and Computer Engineering,
University of Windsor, Windsor, Ontario, Canada Tel: +1-519-253-3000 ext
3426, Fax: +1-519-971-3695 (Emails: {tarique, ktepe, naseria}@uwindsor.ca)
phase, the source node floods the network with route request
messages and all the neighboring nodes of the source node are
obligated to rebroadcast these requests to their neighbors until
the destination node replies back to the source with routing
information. Flooding can work for small networks but as the
number of nodes increases, it causes significant degradation
on the performance of the network.
There are proposals to improve routing overhead and scal-
ability. One of those proposals is clustering and cluster based
routing [6], [13], and [14]. Clustering (active or passive)
can be described as grouping nodes. A representative of
each group is called as cluster head, and other members
are called cluster members. There are proposals to provide
efficient formation of clusters, selection of the cluster heads
and member association [13]. In order to form and maintain
a cluster, nodes need to cooperate and exchange information
with each other, which can increase overall overhead. Passive
clustering [14] has recently been proposed to exploit on going
traffic to propagate cluster related information. But it is not
efficient to use passive clustering with on demand routing,
because in on demand routing there may not be on going data
traffic and routes are discovered after the data is generated. In
on demand routing, mobile nodes only need to decide whether
they should be forwarding node or not, and searching through
the four states of passive clustering can increase decision time,
hence further lengthens the route discovery phase.
Other proposals that improve scalability involve systemat-
ically reducing the number of messages generated and trans-
mitted during the flooding. Those schemes can be loosely
classified as probabilistic schemes [6], [8], [9], and [10]; and
location based schemes [11], and [12]. The major problem of
probabilistic schemes is that nodes should acquire and main-
tain statistical data to determine the probability at which a node
will rebroadcast a message. Unfortunately, the probability at
which a node should rebroadcast is not universal, but specific
to each topology, there is no analytical formula to obtain that
probability. In location based schemes the topology informa-
tion is used to avoid unnecessary rebroadcasts. Obtaining and
distributing location information require additional hardware
and protocols.
Our approach to improve the performance of on demand
routing protocols, particularly in DSR, is to introduce hierar-
chy. The new protocol derived from DSR, called Hierarchical
Dynamic Source Routing (HDSR), limits the number of nodes
that participate in the route discovery phase of the protocol,
0-7803-9182-9/05/$20.00 ©2005 IEEE.
- 73 -

which in turn reduces overhead in compare to DSR. There
are two states in which a mobile node can be: (1) Mobile
(regular) node (MN), and (2) forwarding node (FN). In MN
state, a node acts as either source or destination. In FN state,
a node forwards packets for other nodes. Only FNs participate
in the route discovery. FNs are selected based on an FN
selection algorithm described in the following sections. Our
FN selection algorithm is similar to SPAN described in [18].
But, in SPAN, control messages, like ’HELLO message’, are
used. In our proposed protocol FNs selection process does
not require additional control messages, it solely relies on
route discovery messages. With those modifications, HDSR
reduce the routing overhead significantly because the proposed
algorithm works in the route discovery phase of the protocol
that reduces flooding. HDSR is superior to probabilistic and
location based schemes, because it does not need to acquire
and maintain any statistical information about the neighbors
and locations of the nodes.
Rest of the paper is organized as follows. In next section,
we will explain DSR and HDSR in details, which highlights
differences between those two routing algorithms. In the
following sections, we will provide simulation model and
results. Finally we summarize the main results and conclude
the paper.
II. DSR PROTOCOL
The DSR protocol is based on source routing. Network
nodes cooperate to forward packets for each other to allow
communication over multiple hops between nodes those are
not directly within the transmission range of one another. The
originator of each packet determines an ordered list of nodes
through which the packet will travel to reach a destination.
The DSR protocol is composed of two main mechanisms: (1)
Route discovery, and (2) route maintenance.
Route discovery is the mechanism by which a source node
finds a route to a destination. When the source node has some
packets to send to a destination, it will search its cache to find
a route. If it cannot find a route in the cache, it will initiate
a route discovery to find a route to the destination. To initiate
the route discovery, source node transmits a route request as a
local broadcast packet. Each route request contains the source,
and the destination addresses; and a unique identification
number (id). When a node receives this route request, it checks
whether it is the destination or not. If it is the destination,
it sends a route reply to the source node after copying the
accumulated route from the route request packet in the route
reply packet. When the source node receives the route reply
packet, it records the new source route in its cache and send
packets using this new route that has been just learned. If the
node receiving the route request determines that it is not the
destination, it appends its own address to the route record in
the route request packet and propagates it as a broadcast packet
with the same request id.
Route maintenance is the mechanism by which a node is
able to detect any changes in the network topology such that
it cannot send a packet using a route because a link along the
route is broken. In DSR, each node transmitting the packet
Node 1
Node 2
T1
T2
Time
Remaining Time
MN
M
N
SD
2
1
Source Destination
Fig. 1. Transition from MN to FN
S
FN
F
N
FN
D
23
1
Source
Destination
Fig. 2. Transition from FN to MN
is responsible for confirming that data can flow over the link
from that node to the next hop. An acknowledgment provides
that confirmation. If it does not receive any acknowledgment,
the transmitting node treats the link to this next hop is broken.
It will mark all the routes in the route cache which uses that
link as ’invalid’. It will return a route error to each node that
has sent a packet over that broken link so that all those nodes
can update their own route cache as well.
In DSR, a node can learn and cache multiple routes to a
destination by means of a single route discovery. This support
for multiple routes allow the node to react rapidly to the
topology changes. In order to support that multiple routing
strategy, each node is obligated to reply when it receives
a route request. The ultimate outcome of these replies is
”flooding” or ”storming” of overhead packets.
Although some measures have been adopted in DSR proto-
col to reduce flooding due to route request and route reply
(for example limiting the rate of route discovery by using
exponential back-off algorithm, delaying reply from route
cache for random period of time, imposing shorter route
request hop count etc), flooding problem is still severe in
DSR. Network performance in terms of capacity, bandwidth
and energy efficiency is affected by these excessive overhead
packets.
In the introduced protocol, HDSR, we will try to limit the
overhead due to flooding by limiting number of nodes that
participate in the route discovery phase of the protocol.
III. HDSR PROTOCOL
HDSR adaptively selects ’FNs’ from all nodes in the net-
work. In HDSR, a node switches states from being an MN
to FN or being an FN to MN. HDSR achieves three main
goals: (1) it ensures that FNs are selected so that every source
node can reach the destination, (2) it attempts to minimize
the number of nodes to respond to request messages, (3) FN
selection is based on distributed algorithm. HDSR consists of
two main activities - FN Selection and FN De-selection.The
following sections describe those activities.
- 2 -

A. FN Selection
An MN determines if it should become an FN or not when it
receives a request broadcast from a source node. The following
FN eligibility rule in HDSR ensures that a source can find a
route to the destination:
•FN eligibility rule: MN should become FN if it discov-
ers, using only information gathered from local request
messages, that the source can not reach the destination
without the packet forwarding assistance from this node.
That FN selection algorithm does not minimize the number
of FNs, it roughly ensures that there are enough FNs in the
network so that a source can eventually send packets to a
destination. When an MN hears a route request first time and
does not hear an immediate rebroadcast of this request, it
goes into a random back-off period. If it does not hear any
rebroadcast of the same request until end of that period, it
switches to FN and rebroadcast the request by itself, and flags
the status as processed for this source and request id pair. But
if it hears a rebroadcast during the back-off period, it does not
switch to FN and stores the remaining time of the back-off
period, which will be used in the future back-off periods, and
flags this request as unprocessed for this source and request id
pair. This random back-off period resolves the conflicts that
may arise and eliminate the chances that all the MNs decide to
switch to become FN. The flowchart of the algorithm is given
in Figure 3 for further explanation. Each node chooses a delay
value randomly, and this can be expressed in the following
equation:
Delay =R·K, (1)
where Runiformly distributed random variable between
[0,1], and Kis a multiplying delay factor. Here Kis an
important factor in the stability and efficiency of the algorithm
as in the case of IEEE 802.11 contention window. It should
be function of number of neighbors or number of repeated
rebroadcast message. But, in our simulations, we did not
optimize or apply an adaptive algorithm. By experiments, we
found that the acceptable value is around 10 millisecond for
the network sizes that we simulated. We are working on an
algorithm that adaptively finds this value for every network
scenario.
Figure 1 illustrates an example how the FN selection al-
gorithm works. The source node Sand destination node D
are out of the radio range of each other. The source node
can only reach the destination node through either MN 1or
MN 2. As soon as MN 1and MN 2hear the route request,
they independently employ FN selection algorithm and they
randomly select back-off time, which are labeled as T1and
T2, respectively. Clearly the delay time of MN 1is greater
than that of MN 2.SoMN2’s timer will expire sooner and
it will become FN then rebroadcast the request message to
the destination node. As soon as MN 1hears the rebroadcast
request package, it stops its timer, and stores the remaining
back-off delay for future also stays as MN. After receiving
the request message, the destination node replies back to the
source node through MN 2, now it is actually an FN (FN 2).
received request
packet
request
already
no
processed ?
no
start timer
yes
yes
if
?
delay==0
select random
back−off time
set timer to delay
if
there’s rebroad−
cast ?
no
Timer expire
switch to FN
flag this req.
set delay to reamaining
time and flagged req.
return
yes
Fig. 3. FN Selection Algorithm
The source node starts sending data packets to the destination
node using the route S−2−D.
B. FN De-selection
FN automatically change state from FN to MN by over-
hearing ongoing traffic. The following rules in HDSR ensures
that there should not be more FN than required for the proper
network operation.
•FN De-selection rule: FN should become MN if it is the
source or destination or it discovers that its role as a FN
is redundant
For this operation, a node is able to overhear a packet
carrying routing information by operating its network interface
in promiscuous receive mode. If the FN discovers that it is no
longer needed or if there are routes shorter than it belongs to,
it switches to MN. This procedure is depicted in the Figure 2.
In that scenario, source node Sand destination node Dare
the MNs. When the source initiates route discovery to find
a route to the destination, it finds FNs 1,2and 3. Assume
that all of them forwards the rebroadcasts request message for
that source. As a result the source node discovers two source
routes S−1−Dand S−2−3−D. Since the source node
chooses the shortest path, route S−1−D, and starts sending
packets using that route, FN 2overhears the ongoing traffic
between Sand FN 1, then switches back to MN. Similarly
FN 3applies the same procedure and switches to MN.
IV. PERFORMANCE EVA L UAT I O N
We used Network Simulator 2 (NS-2) [4] to implement and
test the performance of the new proposed protocol. In NS-2,
the effective transmission range of wireless radio is 250 meters
and the medium access control (MAC) protocol is based
on IEEE 802.11 with 2 Megabits per second raw capacity.
The 802.11 distributed coordination function uses Request-To-
Send (RTS) and Clear-To-Send (CTS) control packets [5] for
- 3 -

80 85 90 95 100 105 110 115 120
0
5
10
15
20
25
30
Number of nodes
request broadcast per data pkt.
DSR
HDSR
Fig. 4. Route Request comparison of DSR and HDSR
80 85 90 95 100 105 110 115 120
0.2
0.25
0.3
0.35
0.4
0.45
Number of nodes
Delivery ratio
DSR
HDSR
Fig. 5. Delivery ratio of DSR and HDSR
unicast data transmission to a neighboring node. The RTS/CTS
exchanges precede the data packet transmission, and imple-
ment a form of virtual carrier sensing and channel reservation
to reduce the impact of the hidden terminal problem. Data
packet transmission is followed by an ACK. To compare the
performance of the proposed protocol and the traditional DSR
protocol, we focused on the following performance metrics:
•Scalability: It is a measure how the network protocols
perform under increasing number of network nodes,
traffic load and size.
•Throughput: It is measured as the total number of useful
data (in kilo bits per second) received at the destinations,
reported values are average throughput for the duration
of the simulation time.
•Delivery ratio: It is measured as the ratio of the number
of successfully received packets to number of generated
packets sources.
•Routing overhead: It is measured as number of routing
packets per one data packet successfully received at
destination.
•Energy Efficiency: It is measured as number of data
packets successfully reached the destination node with
the same amount of initial energy.
Source Destination
Intermediate Nodes
range of source
range of destination
Fig. 6. Scenario with number of neighbors.
A. Scalability
In order to evaluate the performance of HDSR in different
network topologies, we randomly distributed mobile nodes
over a flat area according to uniform distribution. In this pa-
pers, all the simulations carried while the network nodes were
static. We kept node densities constant when we increased the
number of MNs in the network. For example, if the area is
1000m by 1000m when the number of MNs is 80, the area is
1000m by 1250 m when the number of MNs is 100. Traffic
sources were Constant Bit Rate (CBR) with 128 bytes per
packet, and we varied the packet generation of the source
when we were measuring the throughput performance of the
network. The source-destination pairs were placed randomly
over the network but the number of pairs were constant during
each simulation scenario, there were 20 pairs communicating
in each simulation. Each CBR starts in 10 seconds interval
and each simulation run for 200 seconds. In order to increase
the reliability of the simulations, each scenario is simulated 5
5 10 15 20 25 30
0
5
10
15
20
25
30
35
no. of neighbors
request broadcast
DSR
HDSR
Fig. 7. Number of Route requests per number neighbors.
- 4 -

1 1.5 2 2.5 3 3.5 4 4.5 5
25
30
35
40
45
50
55
60
65
70
Throughput(KBPS)
Packet generation rate(pkt/sec)
DSR
HDSR
Fig. 8. Network throughput under varying load condition
different times with randomly selected node topologies, and
reported results are average of these 5 simulations. Figure 4
shows how the number of request rebroadcast messages1in
both protocols differs. From that figure, we observe that HDSR
reduced the number of request messages in the network in
compare to DSR. In Figure 5, we observe that delivery ratio
of HDSR is better than that of DSR. We project that the higher
delivery ratios provided by HDSR are originated from having
lower overhead in the network. As the number of nodes in the
network increases, the overhead occupies a significant portion
of the bandwidth in DSR, as a result the delivery ratios and
throughput drop.
As a second scenario, we wanted to see how HDSR differs
from DSR if we increase number of intermediate nodes. In
that scenario, depicted in Figure 6, we adjusted number of
intermediate nodes (nodes that are neighbors to both source
and destination; so all can forward packets for the source to
the destination.) between 5 and 30. Figure 7 shows that the
overhead with respect to number of increasing intermediate
nodes of DSR is increasing at a steeper rate than that of HDSR.
For those scenarios the overhead of DSR is roughly 3 times
larger than HDSR.
B. Network Throughput
Another parameter that we would like to measure is network
throughput. We measured the throughputs by varying traffic
generation rates of the source nodes. Figure 8 shows a typical
simulation result. That figure was obtained from a network
of 100 MNs spread over an area of 1000m by 1250m. We
varied the traffic generation of sources between 1 packet per
second and 5 packets per second. As the figure reveals, the
throughput performance of HDSR is greater than that of DSR,
and network responds to load increases gracefully and be able
carry the load. On the other hand, network with DSR can not
respond to load increase and the network gets congested after
2 packet per second CBR traffic.
1Number of route reply messages are negligible compare to number of
request messages, that’s why they are not included in the figure.
2 3 4 5 6 7 8
0
200
400
600
800
1000
1200
1400
1600
1800
Initial Energy(Joules)
Data pkt. reached destination
DSR
HDSR
Fig. 9. Energy Utilization of HDSR
C. Energy Efficiency
The proposed HDSR protocol is more energy efficient than
DSR protocol. Figure 9 shows how many data packets reach
their destination with given initial energy to the nodes. The
results were obtained from simulation of 80 nodes distributed
in an area of 1000m x1000m area. We used the energy model
bundled in NS-2. In that model, each packet, including route
request and replies, transmission and reception of a packet
consumes energy depending on their lengths. We varied the
initial energies of nodes to 2 Joules, 4 Joules, 6 Joules and 8
Joules. We ran the simulations for 200 sec. From the figure, it
is observed that HDSR can transmit two to three times more
data packets with same energy. One can conclude that the
HDSR improves the life-time of a network at least twice.
V. C ONCLUSIONS
This paper presented HDSR protocol, where forwarding
nodes are selected on demand and distributively. The protocol
does not require any additional control messaging and solely
rely on routing discovery messages of on demand routing pro-
tocols . Our initial experiments reveal that HDSR significantly
improve throughput performance of the network by reducing
the control overhead. HDSR increased the energy efficiency of
the network, which is always desirable in mobile environment
where main source of energy is battery. However, HDSR has
a limitations under very low node densities. If there are only
a single path between the source and destination, delay may
increase if all the intermediate nodes between the source and
destination wait for random back-off delay before forwarding
the request packet. This kind of delay may increase network
latency, as a future work, such limitations need to be over-
come. In this paper, we presented limited number of simulation
results. Those showed that HDSR offered improvements over
DSR but in order to make more quantitative comparisons, more
simulations need to be performed, as a future work, we will
try to find out how much improvement HDSR can offer over
DSR in what conditions? Also we need to devise an adaptive
algorithm to find multiplying delay factor, Kin Equation 1.
- 5 -

REFERENCES
[1] C.E. Perkins and P. Bhagwat, “Highly dynamic destination-sequence
distance-vector routing (DSDV) for mobile computers”,in Proc. of ACM
SIGCOMM 94, London, UK, August 1994, pp 234-244.
[2] C.E. Perkins, “Ad Hoc On Demand Distance Vector (AODV) routing”,
IETF Internet-Draft, draft-ietf-manet-aodv-00.txt, November 1997 (in
progress)
[3] J. Broch, D. B. Johnson, and D. A. Maltz, “The Dynamic Source Routing
Protocol for Mobile Ad Hoc Networks”, IETF Internet-Draft, draft-ietf-
manet-dsr-00.txt, March 1998 (in progress)
[4] K. Fall and K. Varadhan, “Ns Notes and Documentation Technical
Report”, University of California Berkeley, LBL, USC/ISI, and Xeron
PARC, 2003
[5] V. Bharghavan, A. Demers, S. Shenker, and L. Zhang “MACAW: A media
access protocol for wireless LANs”, in Proc. ACM SIGCOMM 1994,
London, UK, August 1994, pp. 212-225
[6] S. Ni, Y. Tseng, Y. Chen, and J. Sheu, “The Broadcast Storm Problem
in a Mobile Ad Hoc Networks”, in Proc. MOBICOM 1999, Seattle,
Washington, August 15- 20 1999, pp. 151-162
[7] W. Peng and X. Lu, “On the reduction of broadcast redundancy in
mobile ad hoc networks”, In First Annual Workshop on Mobile Ad Hoc
Networking and Computing, August 2000, pp. 129-130
[8] B. Krishnamachari , S.B. Wicker, and R. Bejar, “Phase transition phe-
nomenon in wireless ad-hoc networks”, in Proc. GLOBECOM 2001,
Vol.5, San Antonio, Texas, November 2001, pp. 2921-2925
[9] Y. Sasson, D. Cavin and A. Schiper , “Probabilistic Broadcast for
flooding in Wireless Mobile Ad Hoc Networks”, Swiss Federal Institute
of Technology, Switzerland, Technical Report IC/2002/54.
[10] Z.J. Haas, J.Y. Halpern, and L. Li, “Gossip Based ad hoc routing”, in
Proc. of IEEE INFOCOM 2002, Vol. 3, New York, June 23-27, 2002,pp.
1707-1716.
[11] H. Lim and C. Kim, “Multicast tree construction and flooding in
wireless ad hoc networks”,in Proc. 3rd ACM International Workshop
on Modeling, Analysis and Simulation of Wireless and Mobile Systems,
Boston, Massachusetts, August 20, 2000.
[12] H. Lim and C. Kim, “Flooding in wireless ad hoc networks”, Proc. ACM
MSWiM Workshop at MOBICOM, August 2000.
[13] C.R. Lin and M. Gerla , “Adaptive Clustering for Mobile Wireless
Network”, IEEE Journal on Selected Area in Communication,Vol. 15,
No. 7, September 1997, pp 1265-1275.
[14] Y. Yi, T.J. Kown and M. Gerla, “Passive Clustering (PC) in Ad Hoc
Networks”, Internet Draft,draft-ietf-yi-manet- pac-00.txt, November 2001
(in progress)
[15] N. Zhaou, H. Wu, A. A. Abouzeid , “Recative Routing Overhead in
Networks with Unreilable Nodes”, in Proc. MOBICOM 2003, September
14-19, 2003, San Diego, California, pp. 147-160.
[16] J. Sucec and I. Marsic, “ Hierarchical Routing Overhead in Mobile Ad
Hoc Networks”, IEEE Transactions on Mobile Computing, Vol. 3, No. 1,
Januray-February 2004, pp. 46-56
[17] P. Gupta and P. R. Kumar, “The capacity of wireless networks”, IEEE
Transactions on Information Theory, Vol. 46, No. 2, March 2000, pp.
338-404
[18] B. Chen, K. Jamieson, H. Balakrishnan, and R. Morris,” Span: An
Energy-Efficient Coordination Algorithm for Topology Maintenance in
Ad Hoc Wireless Networks”,in Proc. MOBICOM 2001, Rome,Italy, July
2001, pp. 85-96
Mohammed Tarique is a PHD candidate in the Electrical and Computer
Engineering department, University of Windsor, Ontario. He did B.Sc. in
Electrical and Electronic Engineering from Bangladesh University of Engi-
neering and Technology in 1992. He served as a Senior Engineering officer in
BEXIMCO,Bangladesh. He did his Master of Engineering degree from Lamar
University, Texas, USA. His research interests are wireless communication,
wireless ad hoc network, sensor networks, network security.
Kemal E. Tepe (S’97-00, M’01) is an assistant professor in Electrical and
Computer Engineering Department of University of Windsor. He received
the B.S. degree in Electrical Engineering from Hacettepe University, Ankara,
Turkey in 1992 and the M.S. and Ph.D.degrees in Electrical Engineering from
Rensselaer Polytechnic Institute, Troy, NY in 1996 and 2000, respectively.
After receiving his Ph.D., He was employed as a research scientist at Telcordia
Technologies Applied Research, Next Generation Wireless Networks research
group in Red Bank, NJ. In December 2001, He worked as a post doctorate
associate to Wireless Information Networks Laboratory (WINLAB), Rutgers
University, Piscataway, NJ. His research interest is in wireless communication
systems, particularly channel coding, wireless ad hoc network protocols,
sensor networks and network security.
Mohammad Naserian (S’04) received B.S. from University of Tehran and
M.S. degree from Iran University of Science and Technology(IUST) both in
Electrical Engineering in 1996 and 1999 respectively. In September 2002
he joined the Department of Electrical and Computer Engineering in the
University of Windsor where he is pursuing his Ph.D. His research interest is
in wireless mobile communication systems, wireless ad hoc network protocols,
sensor networks and network security.
- 6 -