Detection of suspected nodes in MANET

ACEEE Int. J. on Network Security , Vol. 03, No. 01, Jan 2012

Detection of suspected nodes in MANET
Suman S Chandanan1, Brajesh Patel2, Amit Kumar Chandanan3
1
Shri Ram Institute of Technology, Jabalpur, India
1
Email: suman_singh2008@yahoo.com
2
Shri Ram Institute of Technology, Jabalpur, India
2
Email: brajesh.patel@rediffmail.com
3
Hitkarini College of Engineering and Technology, Jabalpur, India
3
Email: amit.chandanan@hcet.hitkarini.com

Abstract—Mobile ad hoc network (MANETs) is an emerging The main contribution of this paper is concept of intrusion
area with practical applications. One such field concerns detection mechanism used to discourage selfish and
mobile ad hoc networks (MANETs) in which mobile nodes adversarial behavior in MANET.
organize themselves in a network without the help of any
predefined infrastructure. Securing MANETs is an important
II. PROBLEM STATEMENT
part of deploying and utilizing them, since them are often
used in critical applications where data and communications Intrusion Detection is an activity that determines whether
integrity in important. Existing solutions for wireless a process or user is attempting something unexpected. It
networks can be used to obtain a certain level of such security. works, as defined by [10], on the basis of examining activity
Nevertheless, these solutions may not always be sufficient, as
on a specific machine or network and deciding whether the
ad-hoc networks have their own vulnerabilities that cannot
be addressed by these solutions. To obtain an acceptable level activity is normal or suspicious. It can either compare current
of security in such a context, traditional security solutions activity to known attack patterns or simply raise an alarm
should be coupled with an intrusion detection mechanism. condition when specific measurements exceed preset
We propose using a quantitative method to detect intrusion in values.There have been many approaches intrusion detection
MANETS with mobile nodes. Our method is a behavioral in MANETs.
anomaly based system, which makes it dynamic, scalable, The initial classification is based on authentication based
configurable and robust. Finally, we verify our method by schemes. These rely on the identification of nodes by a unique
running ns2 simulations with mobile nodes using Ad-hoc on- identifier. Use of encryption keys fall into this category, and
demand Distance Vector (AODV) routing. It is observed that
they have been deeply studied. The second approach is
the malicious node detection rate is very good, and the false
positive detection rate is low. behavioral based algorithms where intrusion is defined based
on nodal activities, rather than its identifier. This, according
to us, is a better approach for the following reasons:
Keywords- MANET, Intrusion detection, AODV
Node identities can be easily stolen. Behavior is tougher to
replicate.
I. INTRODUCTION  Identity based behavior involves storage of Identifier
databases or logic
Misbehaving nodes in a MANET can adversely the
Each new node has to be given a unique identifier, making
availability of services in the network shown by the research
the process of deployment more expensive (time and cost).
[3].node misbehave either because they are broken, selfish
Thus, we limit our focus to intrusion detection based on
or malicious. Broken nodes are non functional node in the
behavior, since we think it is a more efficient, lightweight and
network. A node is agreed to forward traffic on the behalf of
easily scalable solution to Intrusion Detection in MANETs.
other node but it work as non functional node prior to it fulfill
Intrusion Detection Systems based on behavior can be
this agreement. Selfish node can agree to forward packet but
broadly classified into these categories: anomaly detection,
silently drop the packets in attempt to consume bandwidth
signature or misuse detection, and specification based
and energy of the channel. the decentralized nature, scalable
detection. We mention these as per the taxonomy proposed
setup and dynamically changing topology makes ad hoc
in [11].
networks ideal for a variety of applications ranging from front-
line zones(military and natural) to data collection as A. A NOMALY DETECTION
investigated in [4],[8], [16].Number of MANET secure routing In such systems, a baseline profile of normal system
schemes in the research literature, for the example [5], [6], [7], activity is created. Any system activity that deviates from
[10] do not mitigate against these misbehaviors. In this paper the baseline is treated as a possible intrusion. The problems
we present quantitative intrusion detection mechanism on- with this approach are:
demand multipath source routing protocol that effectively Anomalous activities that are not intrusive are flagged as
mitigate against selective packet dropping effect and other intrusive(false positives)
adversarial activities.

© 2012 ACEEE 34
DOI: 01.IJNS.03.01. 61


 Intrusive activities that behave in a non-anomalous manner detectors are said to have a greater accuracy in detecting
are not detected (false negatives) undefined behavior. They would at the very least be able to
Anomaly detection for mobile computing may demand qualify their decisions better. M. Alam, T Li et al, in [12],
that the normal profile be periodically updated and the proposes an IDS which uses a quantitative method of anomaly
deviations from the normal profile computed. The periodic definition based on transmission characteristics, but factors
calculations can impose a heavy load on some resource in historical transmission behavior of the node.
constrained mobile devices.
Zhang and Lee [15] propose distributed and cooperative III. PROBLEM DIFINATION
intrusion detection model in their pioneer work in this field
Joo B. D. Cabrera, Raman K. Mehra [19] defines an
based on statistical anomaly detection techniques. Every
“ensemble” method to detect, report and average anomaly-
node in the network participates, and runs an IDS agent runs
data in networks using clusters. Each node runs a”Local IDS”,
which performs local data collection and local detection,
and measures Anomaly-index measuring the deviation of
whereas cooperative detection and global intrusion response
measured data from the normal. This is reported to the cluster
can be triggered when a node reports an anomaly. The
head which then propagates a cluster level anomaly index to
authors consider two attack scenarios separately - abnormal
a manager, which performs all the decisions.
updates to routing tables, and detecting abnormal activities
It is a distributed Solution, but involves two levels of
in layers other than the routing layer; these formed the
central entities. The presence of central entities makes it a
definition of the anomaly.
central point of failure - the cluster may become dysfunctional
B. MISUSE DETECTION if an attacker targets the cluster head. Also, these central
In misuse detection, decisions are made on the basis of nodes are usually more resource intensive (also due to
the signature of an intrusive process, and the traces it leaves complex logic), and decrease survivability.
in the observed system. Legal behavior is defined and M. Alam, T Li et al in [12] suggests a non-centralized
observed behavior compared against it to recognize solution, but do not cater to mobile nodes or MANETs. Our
intrusions. Such a system tries to detect evidence of intrusive central challenge is to find a quantitative, distributed and
activity irrespective of any knowledge regarding the dynamic intrusive detection solution for MANETs that
background traffic (i.e., the historical behavior of the system). involve mobile nodes in a non-cluster based environment.
In an architecture proposed by P. Albers, O. Camp etc, [18], A. SPECIFIC NEEDS AND CHALLENGES
the authors suggest using nodes individually running misuse-
This section breaks up our research problem definition
detecting local IDS (or LIDS) agents. They define misuse/
into further detail. This will assist in proposing a solution
attack signatures using variables in SNMP Management
that addresses each of the challenges or problems faced in
Information Bases (MIB) variables. A prototype that defines
creating an intrusion detection system.
misuse as telnet access arising from outside the community
has been tested. B. INTRUSION DETECTION
C. SPECIFICATION BASED DETECTION Members of MANETS that display erroneous or
malevolent behavior are often termed “malicious” nodes; from
This defines a set of constraints that describe the correct
now on, all nodes that display any undefined or unexpected
operation of a program or protocol, and monitors the
behavior are referred to as “malicious nodes”. Thus, the first
execution of the program with respect to the defined
question to be answered is: How do we identify nodes
constraints. This technique may provide the capability to
displaying malicious behavior? In other words, how is the
detect previously unknown attacks, while exhibiting a low
anomaly described?
false positive rate.
Secondly, nodes moving in uncontrolled environments
Tseng, Balasubramanyam et al [13] propose IDS based
with relatively poor physical protection have a non-negligible
on this approach. Their approach uses finite state machines
probability of being compromised. Along with attacks from
to specify correct AODV routing behavior and distributed
the outside world, the possibility of attacks launched by
network monitors for detecting run-time violation of the
compromised nodes from within the network exists too. So,
specifications .Similar work for DSR has been done by P. Yi, Y
the second question to answer is: Is our solution time-
.Jiang at al [14].
continuous? Can a node that started as a legal node, but was
D. COMPOUND DETECTION compromised after some time is recognized?
An improvement over misuse and anomaly detection is
compound detection, which is misuse inspired system that IV. RESEARCH APPROACH
forms a compound decision based on both the normal The solution to research challenge is presented in this
behavior of the system and the intrusive behavior of the section. It is based on the quantitative intrusion detection
intruder. The detector operates by detecting the intrusion techniques in [24], but is applied to a MANET containing
against the historical, normal traffic in the system. These mobile nodes.

© 2012 ACEEE 35
DOI: 01.IJNS.03.01.61


A. RESEARCH APPROACH FEATURES transmission from a node to another. Each node calculates
The detailed set of points against which to measure the and maintains DTQ for each of its neighbor nodes.
effectiveness of proposed approach were mentioned in When DTQ value falls below set thresholds, the neighbor is
section 3.1.1, The answers to those questions are enumerated signaled as a malicious node.
here. Confirmation: The next step in the identification of such
Is our solution time-continuous? Can a node that started as nodes is to ascertain whether a reading made by one node is
a legal node, but was compromised by another malicious node correct. This is decided based on a group consensus approach
be recognized? Yes, our solution bases only on transmission every node in the network is sent a request to accept/reject
and response behavior alone. Our solution is also dynamic; this decision.
every node regularly (periodically) checks its neighbor Nodes receiving such a request can vote for or veto by
statistics to determine abnormal behavior. referring to its own DTQ readings for the node in question.
Is our proposed method truly distributed? Yes. Each node The vote initiating node then draws a consensus based on
in the MANET, or any number of nodes in the MANET can these replies. If more votes have been received approving of
be configured to assume the responsibility of detecting malicious behavior, the node is added into a Black-list that
abnormal behavior allows all nodes to refrain from further communication with
this node.
B. DETECTION OF SUSPECTED NODE
C. RECOGNITION OF MALICIOUS NODES
The first level of moving toward a secure ad hoc network
consists of identification of nodes within the network that Recognition of a node displaying malicious behavior is a
display unexpected behavior, or, in other words, may have continuous process followed by each node. The process of
turned malicious. Identifying malicious nodes consists of malicious node recognition is detailed by the flowchart in
two steps. fig.1.
The first is the recognition of nodes that may be classified
as displaying malicious behavior, and the second is to
ascertain whether that classification is correct.
Recognition: Detecting malicious nodes entails defining
the term malicious. The scope of the current research allows
definition of malicious nodes as those that have aberrations
in data exchange patterns. Dr. Alam, Tao Li et al, in [12],
propose a method in which nodes are expected to
acknowledge every message it receives. Every node measures
the number of acknowledgments it has received from the
neighbor nodes 1 it has tried to transmit to.
In other words, each node records the throughput of every
neighbor node it has attempted to communicate with. This
value is a measure of near-term behavior. This behavior
measured over a period of time determines the historical
quality of behavior of the neighbor node. This statistic is the Figure 1. Flow chart for Intrusion detection

stability of the nodal behavior, and will henceforth be referred D. CONFIRMATION OF DETECTION
to as”STB ()”. “Data transmission quality” (referred to as The next step is that of collectively deciding whether a
DTQ from now on) is defined as a function of STB (), node whose behavior is erratic is actually a malicious one.
probability of error in the channel (P ()), and the energy For example, node A has detected that node B’s DTQ has
needed to transmit data (E). fallen below a threshold.
DTQ = kx D × STB () (1) Node A now wants consensus on its suspicion, and
E × P () triggers a vote by sending a broadcast request for the same.
D = Power needed for transmitting the total data attempted to When MANET nodes receive such a request, they check the
be sent, DTQ values for node B in their tables, and reply with a positive
E = Energy needed to send 1 byte of data, and or negative vote. These votes are aggregated at node A to
k is a constant. decide node B’s status.
The current research is limited to non-cluster based
networks, and it varies here from the paper mentioned above. E. VOTING DETAILS
Also, in the current research, transmission is always atomic Its flow is defined through flowchart in fig.2. A few more
in terms of packets - a packet is either transmitted completely, details within the voting process are discussed below.
or not at all. We rely on measuring the effectiveness of

© 2012 ACEEE 36
DOI: 01.IJNS.03.01. 61


question improves, then no vote-requests are rescheduled.
If not, the node may genuinely be a malicious one, and it’s
time to ask for a consensus. This guarantees that there are
no premature, repeated vote requests.
E. ACKNOWLEDGMENT FOR MESSAGES
Acknowledgment messages every node sends an
acknowledgment of message receipt as soon as a it receives
a data message. The sender waits for acknowledgment for
some time.
(a) Acknowledgment arrival: If the acknowledgment arrives
on time, the statistics for the acknowledgment sender are
updated. If this is the end of a bucket, the DTQ is calculated
anew, and a comparison for DTQ versus threshold is made. If
Figure 2. flow chart for voting process
necessary, a vote-request is scheduled.
(a) Vote arrival: A vote-initiating node keeps a count of the (b) Acknowledgment timeout: The ACK-timeout is the time a
number of votes in receives. It also does not register more sender A waits for an acknowledgment from the intended
than 1 vote from the same neighbor, for a particular vote- recipient, node B .If the acknowledgment does not arrive on
request. Once it has received votes from all of its neighbors, time (i.e. arrives after ACK-timeout seconds), and if this is
it decides for or against the voted-upon node. For this the end of a block, then, again, the DTQ is recalculated and
implementation, we take “all” neighbors to mean the total the process of comparison repeats. Also, if the end of a block
number of node in the network less one, which is the maximum (bucket) is reached, the sender no longer accepts any more
expected neighbor count. acknowledgments for this block of sent data, i.e. the DTQ for
(b) Vote Request timeout: The situation where all neighbors this block of data is final.
respond is an ideal situation, in wireless networks and more
so in MANETs, where data packets may be lost in transit. V. SIMULATION AND RESULT ANALYSIS
Also, some nodes may decide not to vote. In such cases, the
vote-initiator cannot wait indeterminately. The vote request This section is dedicated to represent graphical simulated
time out solves this dilemma, and is set as soon as the vote- result and their analysis in NS-2[1]. The simulation is aim to
request is sent out. At the end of this time-out period, the show the performance of the routing protocols with present
vote request initiator aggregates all the votes it has received, of Detected malicious nodes in the mobile ad hoc network for
and makes a decision based on the counts. All votes received this two metrics are simulated : first, change in the mobility
after this timeout are useless. futures of the nodes and second, changes in the IDS settings
(c) Who votes? All nodes that receive a vote-request attempt of the nodes.
to vote. However, if the number of messages they receive A. CHANGES IN MOBILITY FEATURES
from the vote-initiator is not sufficient for them to decide,
This section aims to measure the functioning of our IDS when
they refrain from voting. We will discuss this sufficiency
changing features of mobility like speed.
number in sections that follow.
A.A . SIMULATION RESULTS
(d) Process after vote decision:
Here, in this section all simulated results shown with their
i. On blacklisting: Immediately after a node has been
parameters
blacklisted, as demonstrated in fig. 4, a message is sent out
A.A .A CHANGES IN THE MOBILITY FEATURES
to all nodes with this information. All nodes receiving this
This set of tests measure the functionality by varying speeds
message add the node to their blacklist details too. Once a
of all nodes uniformly, and then heterogeneously.
node is blacklisted, no communication from such nodes is
A.A .A.A. VARYING SPEED
responded to anymore.
ii. On being acquitted: If a node is acquitted after the vote TABLE I. SIMULATION PARAMETERS FOR MOBILITY
decision, all nodes treat it as usual. No information about the
acquittal is sent out. This raises the question as to whether
the vote-initiator, who now has a low DTQ values for this
node will repeatedly generate redundant vote-requests! In
short, the answer is no! The vote request is scheduled only
once every bucket.
So, if it fails in one bucket, then, the node waits for the
next bucket to occur before it can make a new vote request.
Within this bucket, if communication with the node in

© 2012 ACEEE 37
DOI: 01.IJNS.03.01.61


received/ acknowledged message counts for various nodes
using the output files .False positives occur due to one of
the following reasons:
A node, say, A, does not receive messages for an extended
period from a particular node, say B. The sending node B
evaluates the absence of acknowledgments from A as
malicious behavior, even though A is a legal node (i.e. based
on our simulation settings). However, this is good behavior,
since we have positively identified nodes based on their
transmission characteristics, and can identify innocent nodes
that have turned malicious after establishment of the network.
Vote-replies are lost. Why or where are messages lost?
Figure 3. Intrusion Detection with heterogeneous speeds Either in transit, or due to time-out due to losing
connectivity while being mobile. The former happens because
Fig.3 displays the count of nodes that are recognized as a
the routing, Mac and physical layers use (the AODV [17]
function of time.
based implementation) have a definite loss factor, which
A.A .B CHANGES IN IDS SETTINGS
increases with speed. This is documented by the AODV
We run the various test below with the below set of
implementation documentation - tests for the AODV
common configuration details. The sections then indicate
implementation were done for speeds not exceeding of 10
the changed details alone.
mps, and with not more than 25 nodes.
A.A .B.A VARYING HISTORY COUNT AND BUCKET COUNT
TABLE II. SETTINGS USED FOR IDS CONFIGURATION TESTS

The following fig.4 shows the effect of change in near-time
buckets
A.A .C CHANGES IN NETWORK SETUP
This section seeks to measure the behavior of our IDS Figure 4. Detection rate with varying near-term bucket size
when the topology of the network it is being used changes.
All malicious nodes are detected correctly. It is noticed
This may pertain to the number of malicious nodes
that when the near-time bucket size is low, the false positive
introduced, the number of nodes employed by the network
detection rate is high. This is expected, because a low near-
as a whole etc.
time-bucket value means that the behavior of the nodes is
A.A .C.A NUMBER OF MALICIOUS NODES IN THE NETWORK
measured based on very few transmissions (and the
This section aims to check the effectiveness of our IDS in
acknowledgments received for fewer transmissions). As the
tracing malicious nodes independent of the number of such
bucket size becomes more in tune to the network’s current
nodes present in the system. The tests are conducted by
settings of behavior, false positives become almost nil.
using a varying count of malicious nodes, perpetrating 20 to
The second effect is that of having a low history count
90 percent of the network (20, 40, 60, 80 and 90 percent) fig. 5
itself. This also displays the same behavior as that above.
captures the result.
This is because long-term-bucket measurements aim to
A.A .C.B NUMBER OF NODES IN THE NETWORK
capture the long term behavior of nodes. Say, historically, a
We have shown that the correct number of malicious
node has an 80 percent acknowledgment rate.Then, using
nodes, and the exact malicious nodes are pointed out,
near-term buckets, we measure if the node is consistent with
whatever be the configuration of the number of nodes in the
its”character” of 80 percent. If not, the activity is of interest
network. Thus, we are not repeating tests for this section.
and may be marked for a vote-request trigger. But, if the period
B. DISCUSSION over which history is measured is lowered (by reducing the
All malicious nodes are successfully detected. number of transmissions we monitor), it does not present a
There is possibility of false positives, as noticed. true measure of regular node behavior.The graph starts with
 False positives can be explained by analyzing the sent/ 20 percent perpetration and proceeds to 90 percent
perpetration. All malicious nodes are successfully detected.

© 2012 ACEEE 38
DOI: 01.IJNS.03.01. 61


REFERENCES
[1] Ns2 network simulator. http://www.isi.edu/nsnam/ns.
[2] A.Rajaram and S.Palaniswami. A trust based cross layer security
protocol for ad hocnetworks. International Journal of Computer
Science And Information Security, 6(1),2009
[3] Ashwin Perti ,Pradeep Sharma: Reliable AODV protocol for
wireless Ad hoc Networking , 2009 IEEE International Advance
Computing Conference, Patiala ,India(ICACC-2009,March 2009
[4] Amit Kumar Chandanan, Shailendra Kumar Shrivastava, “Secure
Mobile Network Routing Protocol Using PSR,” cicn, pp.289-295,
2010 International Conference on Computational Intelligence and
Communication Networks, 2010
[5] J. Binkley and W. Trost. Authenticated ad hoc routing at the
Figure 5. Intrusion Detection with varying malicious node link layer for mobile systems. Wireless Networks, 7(2):139–145,
cou nt 2001
[6] Y. Hu, A. Perrig, and D. Johnson. Sead: Secure efficient distance
With the settings used, no false positive identifications
vector routing for mobile wireless ad hoc networks. In Proceedings
happened, even though the simulation ran for a considerable of the 4th IEEE Workshop on Mobile Computing Systems and
amount of time (in some cases more than double the time) Applications (WMCSA’02), pages 3–13, June 2002.
after the actual malicious nodes were identified. Note that in [7] D. Johnson and D. Maltz. Dynamic source routing in ad-hoc
the right-most run 90 percent of the nodes in the network are wireless networks routing protocols. In Mobile Computing, pages
malicious. The graph shows 0 malicious nodes detected. This 153–181. Kluwer Academic Publishers, 1996.
is perfectly expected behavior, since the Voting system [8] B. Awerbuch, D. Holmer, C. Nita-Rotaru, and H. Rubens. An
proposed requires at least one neighbor node to vote. Here, on-demand secure routing protocol resilient to byzantine failures.
none are available, as only one of the ten nodes is a valid In Proceedings of the ACM workshop on Wireless security (WiSE
’02), pages 21–30, September 2000
node.
[9] F. Kargl, A. Klenk, S. Schlott, and M. Weber. Advanced detection
of selfish or malicious nodes in ad hoc networks. In Proceedings of
CONCLUSION the 1st European Workshop on Security in Ad-Hoc and Sensor
We aimed to determine a method to identify malicious or Networks (ESAS 2004), pages 152–165, August 200426
[10] Michael G Solomon and Mike Chappel. Information Security
compromised nodes in a MANET with mobile nodes based
Illuminated. Jones and Bartlett, 2004
on behavioral attributes. We proposed to use a system in [11] Amitabh Mishra, Ketan Nadkarni, and Animesh Patcha.
which aberrations of normal behavior (or anomalies in Intrusion detection in wireless ad hoc networks. IEEE wireless
behavior) are defined quantitatively by observing data communications, February 2004
exchange activity. We defined this anomaly in terms of these [12] Tao Li, Min Song, and Mansoor Alam. Compromized sensor
items: node detection: A quantitative approach. IEEE International
• Long term behavior of a node as measured by continuous Conference on Distributed Computing Systems,pages 352–357,
observation of its responses, mainly in the form of IDS logical 2008.
layer acknowledgments to transmissions [13] Chin-Yang Tseng, Poornima Balasubramanyam, Calvin K,
Rattapon Limprasittiporn, and Karl Levitt Jeff Rowe. A
• Short term response to transmissions we then selected NS-
specification-based intrusion detection system for aodv. pages 125–
2 as the simulator of choice to create an environment dubbing 134, 2003.
real-life mobile nodes. Where there are mobile nodes, [14] Ping Yi, Yichuan Jiang, Yiping Zhong, and Shiyong Zhang.
forwarding of data to the correct recipient cannot be done Distributed intrusion detection for mobile ad hoc networks.
without the use of a routing algorithm. We used an Symposium on Applications and the Internet Workshop, 2005.
implementation of AODV [17], an Ad hoc On-demand Distance [15] Y.Zhang and W.Lee. Intrusion detection in wireless ad hoc
Vector reactive routing protocol, to perform this function for networks. International Conference on Mobile Computing and
us. Networks, page 275283, August 2000.
The last phase involved measurement of all the data with [16] Brent A. Peacock. Connecting the edge :mobile ad-hoc networks
(manets) for network centric warfare. April 2007.
various simulation runs. A”Selective Forwarding Node” that
[17] C. Perkins and E. Royer. Adhoc On-demand Distance Vector
simulates a”black hole” or”selective forwarding” attack by Routing. IEEE workshop on Mobile Computing Systems and
not replying to any transmissions is created. Such node also Applications, 3(4):90–100, February 1999
acts as”flooding attack” nodes. Instances of these nodes [18] Patrick Albers, Olivier Camp, Jean-Marc Percher1, Bernard
masquerade as malicious nodes.in the simulation runs. The Jouga, Ludovic Me, and Ricardo Puttini. Security in ad hoc networks:
data collected has shown that our proposed system works a general intrusion detection architecture enhancing trust based
well. Our IDS can detect malicious nodes with almost 100 approaches. 2005
percent proficiency. The percentage of false positives is also [19] Joo B. D. Cabrera, Raman K. Mehra, and Carlos Gutirrez.
reasonable, and does not exceed 20 percent for most Ensemble methods for anomaly detection and distributed intrusion
detection in mobile ad-hoc networks. International Conference on
simulation cases.
Mobile Computing and Networks, 9(1), January 2008

© 2012 ACEEE 39
DOI: 01.IJNS.03.01.61

Detection of suspected nodes in MANET

More Related Content

What's hot (18)

Viewers also liked (20)

Similar to Detection of suspected nodes in MANET (20)

More from IDES Editor (20)

Recently uploaded (20)

Detection of suspected nodes in MANET