Mobile ad hoc networks (MANET) are lately a hot spot for research. MANETs are a type of ad-hoc networks which are wireless networks and have a decentralized network structure.
MANETs contain mobile nodes connected by wireless links within the network range. Lately, the increase in the usage of laptops, cellular devices etc have demanded the enhancement of mobile ad hoc networks. Nodes present in MANET's are usually mobile and hence the links between the nodes keep changing frequently. Therefore, the network flexibility has to be very high in order to accommodate the mobility of nodes.
Communication in ad-hoc networks does not involve fixed routers as in wired networks where they are used solely for enabling communication between nodes. These networks are self-configuring and hence can be used in varied application domains where centralized nodes cannot be of help. Lack of pre-existing framework requires the nodes to participate in routing packets from source node to destination node by transferring them to the neighboring nodes.
Get your grade
or your money back
using our Essay Writing Service!
Communication in MANET's can be achieved by cooperation between the nodes involved in transfer of packets. As the nodes in MANET's are mobile, a fixed architecture cannot be implemented. The network topology keeps changing with the mobility of the nodes. The decision about which nodes need to participate in communication is made dynamically based on network traffic and statistics. The decision onto whether or not to participate in communication is based on their self interest. The presence of selfish nodes creates a problem as they do not take part in transfer of packets thus disrupting the communication process between nodes. Nodes behave selfishly sometimes to save their battery, errors etc. Some malicious nodes present in the network do so to invade private content, disable packet transfer between nodes.
Proper communication establishment in MANET's is a key field of study. It is proven that even selfish nodes cooperate when necessary. This can be attained by providing impressive incentives to the nodes to be cooperative. Therefore, mechanisms that enable communication need to be implemented at different layers. The incentives provided to the nodes have to be such that the nodes gain an edge over the other nodes in the network.
Reputation based incentives are one of the types of incentives that can be awarded to the nodes that wish to participate. Mechanisms that provide this kind of incentives evaluate reputation value or trust value of a node in the network. The nodes with adequate reputation, as they are supportive can use the networks resources. Those nodes with lesser reputation may be identified as malicious nodes and are not let to use network resources. Hence, implementing these principles many trust management models are proposed which use the value of trust to identify the presence of malicious nodes in the network.
Fig 1: Example of Mobile Ad-hoc Network
MANET's are greatly in use these days as they provide uninterrupted network service even though the node is moving around. Because of this advantage, MANET's are very helpful in emergency situations. Therefore, the reliability of the network has to be paid attention to. Reliability can be increased if the nodes that involve in communication are genuine nodes and not malicious nodes. Reliability can be retained when the malicious or detrimental nodes are identified and banished from affecting the network
Network performance is another aspect that has to be high in a network. Both network performance and reliability are interrelated. If the network performance is aggravated, network reliability would increase rapidly. Network performance can be improved if cooperation between nodes is achieved for communication.
To achieve reliability, performance, efficiency etc many reputation models have been proposed. Many of them involved calculation of reputation of nodes in the network. These values were used in identifying misbehaving nodes, avoiding Denial of Service (DoS) attacks by some models. Some models also proposed ways to encourage nodes to cooperate in communication. But, the models that are currently in use have certain drawbacks and there is no enhanced model that combines the advantages of the available models. All these conditions have persuaded in the introduction of a new upgraded model for trust management.
The main objective of trust models is calculating the trust of the nodes. Imbibing trust on each other between nodes would encourage in attaining secure communication. Generally, trust is defined as the magnitude of confidence or certainty stimulated by the nodes behavior in the network.
Always on Time
Marked to Standard
The trust value of a node is maintained if the conduct of the nodes behavior is appreciable. Nodes with inadequate value of trust are identified as non-cooperating nodes.
The trust value of a neighboring node can be obtained by monitoring the node to know about its practice in packet forwarding. At times, the trust value of a stranger node has to be known. The neighboring nodes of the unknown node share their trust information about the stranger node. This gives the trust by recommendation. In many models, trust value of a node can be determined from the trust attained by direct observations or from the referential trust from other neighboring nodes about the behavior of the node. The trust value of the node that is recommending is also taken into account and based on that the trust by recommendation value is evaluated.
Cooperation of Nodes 'Fairness in Dynamic Ad-hoc Networks (CONFIDANT):
CONFIDANT is one of the important trust based models available. The main purpose of this model is to identify and separate misbehaving nodes in the network.
In order to get the value of trust of the other nodes, a monitoring system is employed by nodes to detect the behavior of the nodes. A copy of packet is maintained at the sending node and the monitoring system of the node continually listens to the transmission of next node to ensure that the contents of packet are not altered while forwarding them.
Trust evaluation of any node is based on experience with other nodes, direct observations or reported information about the behavior from other nodes. In this model, the value of trust given to a node is high if the behavior is known by direct observations, is small if the observations are in the neighborhood, and is even more small value if the trust is through reported experience.
Each node maintains a table with trust values of other nodes. Trust values are changed only when there is enough proof about the behavior of the node. If the trust value thus obtained is less than a predefined threshold value then the node is identified as malicious node.
This model lacks a secure way of recommending other nodes. And also the nodes with trust value less than defined threshold do not have any mechanism to regain trust and participate in communication. This is very important because the reason for inappropriate behavior by some nodes may be to save their battery, errors etc.
Collaborative Reputation (CORE):
CORE is a reputation based model which is very effective to avoid misbehavior by nodes in the network. Actions such as Denial of Service (Dos) are best prevented. This model promotes cooperation among the nodes as it makes selfishness unattractive.
In this model, reputation of each node is evaluated. To evaluate the trust value of a node, three different types of trusts are considered. They are:
1. Subjective reputation: Is calculated from nodes observation.
2. Indirect reputation: Is calculated by considering the information provided by neighboring nodes.
3. Functional reputation: Is calculated by using different functions in calculating subjective reputation and indirect reputation.
The collaboration of the above mentioned reputations give the final reputation value of a node. If the reputation value goes below the defined value, the node is considered as a selfish node and is not allowed to use any of the network resources.
The referential reputation by other nodes can be forged by mutual cheating by nodes. Lack of provision for re-obtaining reputation for nodes.
Secure and Objective Reputation based incentive (SORI):
SORI is also a reputation model that is implemented to inspire effective packet forwarding. It helps in disciplining the behavior of the nodes.
This model involves the calculation of reputation of each node and is called as Local Evaluation Record. For this, a monitoring system is involved. One of the unique features of SORI is that the reputation of node is quantitatively obtained by objective measures. That is the reputation value is obtained by evaluating the ratio of 'number of packets sent to a node' to the 'number of packets that node has forwarded'.
This Essay is
a Student's Work
This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.Examples of our work
Apart from encouraging nodes to cooperate in communication, a mechanism to punish selfish nodes is also implemented in this model. Whenever a node (S) identifies another node (M) as a malicious node, then the node (S) spreads the reputation information about node (M) to the neighboring nodes so that all other nodes punish node (M) by dropping the packets forwarded by node (M). Hence, a malicious node is punished by all the nodes present in the network.
SORI lacks rationality in reputation evaluation as it involves ratio of packets received by a node to the packets forwarded by the node. And also a mechanism that would help non- malicious nodes to regain reputation is not defined.
Proposed Hybrid Model:
A hybrid model for trust evaluation is proposed to overcome majority of the drawbacks present in the models currently in use. In this model, trust value for each of the nodes present in the network is calculated. Based on the trust values, the nodes are allowed to access networks resources.
The trust values are divided into three ranges. The nodes that are trusted by majority of nodes have great trust value hence fall in the 'high' range of values. The nodes that are averagely trusted fall in the 'medium' range. And the least trusted nodes have very less value of trust and hence are in 'low' range. If the value of trust is in 'low' range that is below a defined threshold level then the node is considered as malicious node and is not allowed to participate in communication.
Each node maintains a table consisting of the values of trust of other nodes. The calculation of trust is to be done when the node newly enters the network or when the trust value of a node is changed and needs to be updated.
New node in the network:
Whenever a new node enters the network, the other nodes present in the network do not know whether or not to trust the new node as the trust value is unknown. Hence, the new node is assigned a 'medium' range trust which is greater than the threshold value.
A monitoring system is implemented by each node which would then continually monitor the new node to observe the behavior. As long as the behavior of the node seems acceptable, the value of trust keeps increasing towards 'high' range.
Change in trust value:
The value of trust need to be updated based on the behavior of the node. The recent behavior of nodes is taken into consideration. If any non-cooperative behavior is observed, the trust value is decreased periodically.
The changes are made only when there is enough evidence to prove the misbehavior. The trust value is taken into consideration before a request for packet forwarding is made to any node. Therefore, if the trust value of a node is below the threshold level, the node is isolated from the network.
Hybrid model trust evaluation:
Each node in this hybrid model maintains a table consisting of the trust values of other nodes present in the network. For having a rational methodology of calculating trust, the trust value is based on network stats. The overall trust value of node is obtained by the usage of trust values obtained in two different ways.
Trust by direct observations - Two nodes are neighbors if they are in the transmission range of each other. Every node monitors its neighboring nodes and based on direct observations, a trust value is assigned to the neighboring node. This is the trust value attained by direct observation. It is represented by (Td)
Trust by recommendation - If a node is not a neighboring node then the behavior of the node can be known by considering the recommendations provided by other nodes that are neighbors. The trust value of the node recommending is also taken into consideration. Trust by recommendation is represented by (Tr) and can be obtained by:
Tr = Number of packets forwarded to a node
Number of packets forwarded- Number of packets the
node is destination
The trust by recommendation is evaluated in an interval of time thus paying more attention to the recent behaviors of the node. This helps in knowing whether or not to trust a node.
Overall trust - The final trust value is obtained by using both the trust obtained by direct observations and the trust by recommendation. The overall trust thus obtained is represented by (T).
T = ? (Td * Tr)
The value of 'T' is stored in the table containing trust values of each node.
Methodology for trust by recommendation:
The mechanism used for considering trust by recommendation has to be accurate and forged recommendations should be avoided. Some rules need to be followed to obtain recommendations from other nodes properly.
- If nodes 'a' and 'b' are not neighbors to each other's but are neighbor to another node 'c'. Node 'a' has a trust value 'Tc' and node 'c' has a trust value 'Tb' on node 'b'. Then node 'a' has trust value 'TcTb' on node 'b'. The propagation of trust results in depreciation of trust. This rule is called trust depreciation rule.
- If trust through recommendations about a node is obtained through two different paths then the resulting trust indicates upgrading of trust.
If node 'a' has got trust by recommendations 'T1b' through one path and 'T2b' through other path, then node 'a' has '1-(1-T1b)(1-T2b)' trust on node 'b'. The propagation of trust results in upgrading of trust and hence called as trust upgrading rule.
- If a node 'a' wants to attain the trust of other node 'b' as shown
Here, the node 'b' is an intermediate node which is copied to form two distinct parallel paths.
The trust value of node 'b' is divided into equal parts to support the number of parallel paths. This is called division of trust rule. The recommendations of the other nodes in each of these paths are evaluated and the total trust by recommendations is obtained. Even if the recommendations are forged in one path, the impact on the overall trust is hence minimized.
1. Identify the number of input and output links of the intermediate nodes present.
2. Exclude the links which involve nodes containing trust value less than the minimum defined threshold value.
3. Obtain parallel paths using the nodes in between.
4. If a node can form more than one parallel path then apply division of trust rule.
5. Apply trust depreciation rule or trust upgrading rule through all the intermediate nodes.
6. Obtain the trust by recommendations from all the intermediate nodes using
Tr = ?(Tr1 * Tr2 * Tr3'.. Trn)
This algorithm would be effective to provide secure referential trust and to remain unaffected because of forged recommendations.
Trust Regaining methodology:
Trust needs to be regained by nodes to be a trustable node and to participate in communication in the network. A regaining methodology is a must to help the nodes regain trust.
This is most useful for the nodes that are considered as malicious nodes. Sometimes nodes do not cooperate to save their own resources like battery or lack of network coverage etc. So, though not malicious some nodes have trust value that fall in the 'low' range of trust values. A mechanism that provides an opportunity to regain trust is proposed.
For every stipulated interval of time, all the nodes that have trust values less than the threshold value are identified. And their trust values are increased to the minimum threshold value and the nodes are monitored. These nodes gain trust and hence can again start participating in communication.
If a nodes trust value repetitively decreases and falls in 'low' range then after a certain number of times, the node is completely identified as malicious node and is ignored. If the node maintains its trust value then the node's trust is periodically increased and is allowed to actively take part in communication.
Conclusion and Future Work: