GB2487073A - Credit balances for nodes in a network are amended based on confirmation messages received from the nodes and on reputation records - Google Patents

Abstract

A wireless ad-hoc network comprising a management unit arranged to calculate a reward for nodes that forward a message within the network or a punishment for nodes failing to do so. The reward or punishment is based on confirmation messages received from nodes that forward messages and on an indication of the nodes' reputation. The balance between a node's remaining energy and a nodeâ s remaining credits available for sending messages may be used in determining a reward a node should receive for forwarding a message.

Description

Wireless ad-hoc network and method of operating same

FIELD

Embodiments described herein generally relate to a wireless ad-hoc network, to components thereof and to methods of operating the same.

BACKGROUND

Successful operation of ad-hoc networks fundamentally depends on the willingness of individual devices to collaborate -to relay or forward the packets they receive from neighbours. Energy resources within each node, for instance in the case of battery powered mobile devices, is moreover limited and needs to be managed to ensure that nodes can send their own messages.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiment of the present invention will now be described by way of example only and with reference to the accompanying drawings, in which: Figure 1 shows an ad-hoc network in which an embodiment may be implemented; Figure 2a shows the architecture of a network node according to an embodiment; Figure 2b shows the architecture of the network management unit according to an embodiment; Figure 3 shows a flow chart illustrating the operation of a network node if the network node adheres to an operating principles of an embodiment; Figure 4 shows a flow chart of an embodiment; Figure 5 shows a payment scheme of an embodiment in case of successful message delivery; Figure 6 shows another payment scheme of the embodiment if a non-forwarding node is involved; Figures 7 shows another payment scheme of the embodiment if a non-forwarding and non-reporting node is involved; Figure 8 shows a payment scheme of the embodiment if nodes are colluding to deceive the network manager; Figure 9 shows the interruption of a transmission chain and the resulting pattern of submitted confirmation messages; Figure 10 shows the topology of the ad-hoc network that formed the basis for some of the computer simulations performed for evaluating embodiments; Figure II a shows simulation results for the entire network of Figure 10 and for one node, assuming operation according to an embodiment; Figure 11 b shows simulation results for the same network/node as Figure 11 a but assuming that a node does not forward packages; Figure 1 Ic shows simulation results for the same network/node as Figure 11 a but assuming that a node does not forward packages and does not provide confirmation messages; Figure 1 Id shows simulation results for the same network/node as Figure 1 Ia but assuming that a nodes collude to deceive the network manager; Figure 12 compares the packet success rates using an embodiment with that of a known energy efficient routing algorithm for the same network topology; Figure 13 shows packet success rates for different network topologies; Figure 14 shows a comparison between the packet success rates between a network with a fixed network topology and one with a mobile network topology; and Figure 15 shows an average packet success rate for differing weighting used when determining the costs of the transmission of a message during routing.

DETAILED DESCRIPTION

According to one embodiment there is provided a wireless ad-hoc network comprising a plurality of network nodes and a management apparatus.

The management apparatus comprises a memory, a processor and a receiver for receiving confirmation messages from network nodes. The network nodes comprise a policy for sending a said confirmation message to the management apparatus for each received message. The management apparatus is arranged to store a credit balance for each network node in the memory. The processor is arranged to update said credit balance. The management apparatus is further arranged to store and/or obtain an indication of the reliability of each of the network nodes. The processor updates said credit balance based on confirmation messages received from the network nodes in the data transmission route and based on.the indication of reliability of the said network nodes.

The network nodes may be arranged so that they can only send messages of their own if their credit balance comprises sufficient credit to cover the costs of sending such messages. The nodes in an ad-hoc network can be rational nodes act in a manner that maximises their benefit. Ln other words, the nodes are selfish. The nodes may thus not follow the transmission policy they store, if they determine that violating the policy will, or is likely to, provide a greater benefit to them than following the policy.

The combination of storing a credit balance for the node and an indication of the reliability of the node provides incentives for the nodes to operate in the manner defined by the policies. This is in particular so if the management unit, as it does in one embodiment, uses increases and decreases * in the credit balance and/or the indication of reliability in reaction to the manner in which the nodes operates within the network. If the node, for example operates in an intended manner, then the management unit can affect an increase in credit balance as well as reliability rating. If on the other hand the node violates network policies, then such behaviour can be penalised by a decrease in credit balance and/or reliability rating. Even a network node manages to identify a behaviour that is likely to lead to a beneficial increase in either of credit balance or reliability rating, penalties for behaviour that violate the transmission policies can be imposed upon the node by altering the other one of the credit balance and reliability indicator.

An attempt to determine the node that is likely responsible for a transmission failure may focus on nodes with a low reliability indicator. Any penalties for the transmission failure that may be attributed to nodes may then predominantly be attributed the nodes with low reliability rating, so that these nodes may be penalised by a decrease in their credit rating. The aim of rational nodes will be to maximise their benefits. This may mean a maximisation of their credit balance, giving rise to the potential of behaviour that irregularly maximises the credit balance, potentially at the cost of network reliability. A node that knows that it can be penalised based on its reliability rating, however, may seek to maintain a reliability rating that does not lead to penalties, thus causing the node to operate in the manner defined by the transmission policies.

While the reliability indicator may be stored within the management unit, the management unit may alternatively obtain the reliability indicator from a further entity, for example through an input, such as a database or network connection.

It will be appreciated that, while the embodiment adopts the use of virtual currency, the accounts dealing with the currency balance, the above mentioned credit balances, are managed by the management unit. While the nodes may have knowledge of their credit balance, they are not given authority to alter the credit balance. For this reason the need for tamper proof hardware, as may exist if the nodes had authority to update the.credit balance, may be eliminated.

The wireless ad-hoc network may be a wireless LAN.

In a further embodiment there is provided a network management apparatus comprising a processor, a memory and a receiver for receiving confirmation messages from network nodes. The processor is arranged to store in and retrieve from the memory credit balances for individual nodes in the network and to amend the credit balance based on confirmation messages received from the nodes and on reputation records of individual nodes.

Following notification of a transmission failure, the processor may determine, based on the reputation records, a likelihood for a node. The processor may then retrieve the credit balance of the node from the memory and to alter the credit balance of the node based on the determined likelihood.

S The altered credit balance may then be stored in the memory. The likelihood is a likelihood of the node being responsible, or partially responsible, for the transmission failure.

The processor may be operative to, following notification of a transmission failure, identify the last network nodes along a known transmission route from which a confirmation message has been received and/or the first network node along the transmission route from which no confirmation message has been received The credit balance of one or both of the identified nodes may then be altered based on the reputation record of the two nodes. The embodiment may consider the reputation records of the nodes in isolation from each other. Alternatively or additionally, the embodiment may seek to identify past behavioural patterns in which both of the nodes in questions have been involved in. Doing so may help identify a situation where one of the two nodes has obtained a low reliability record because of the behaviour of a neighbouring node.

The processor may be operative to, upon notification of a failure to transmit a message, detect colluding non-forwarding behaviour of one or more of nodes within a data transmission route based on confirmation messages received from network nodes along a known intended data transmission route of the message.

The processor may be operative to conclude that some or all nodes along the route are involved in colluding non-forwarding behaviour if it has received confirmation messages from all nodes along the route but not from an intended destination node of the message.

The processor may be operative to generate and update the reputation record or reputation records of a said node by noting, in case of a transmission failure, if the node was the last node from which a confirmation message has been received at the receiver and/or the first node from which no confirmation message has been received at the receiver.

According to another embodiment there is provided an apparatus comprising a transmitter for receiving data packets and a receiver for transmitting data packets and a processor. The apparatus of the embodiment is operative to forward a received data packet to another apparatus in an ad-hoc network in accordance with received routing instructions. The processor is arranged to determine, during a routing phase, a cost for forwarding a message to another apparatus in an ad-hoc network based on a current credit balance and on a remaining energy reserve.

According to another embodiment there is provided a method of operating an ad-hoc network comprising sending confirmation messages from network nodes along a transmission route for a message to a network management apparatus. The method further comprises, within the network management apparatus, updating a stored credit balance for network nodes along the transmission route based on received confirmation messages and, if a notification of the failure of the transmission has been received, also based on a reputation records of the network nodes along the transmission route.

The method may further comprise, during a routing discovery phase and within a said network node, determining a cost of forwarding a data packet to another network node based on a current credit balance of the said network node and based on a remaining energy reserve of the said network node.

Figure 1 shows an ad-hoc network 300 in which an embodiment may be practiced. The ad-hoc network 300 comprises a plurality of network nodes 310 and 330-350 and a network management unit 320. As indicated by the connection between network nodes 310, the network nodes 310 can establish signal connections with each other and, via these signal connections exchange messages. A sending node 330 may wish to send a message to a destination node 340. Such a message transmission may be conducted via a transmission path involving a number of intermediate nodes 350, as indicated by the thick line in Figure 1. The network management unit can enter into communicative contact with the individual network nodes 310 and 330-350.

Figure 2a) shows the architecture of a node 310/320/330 according to an embodiment. The comprises a receiver 360 for receiving signals/messages, a transmitter 370 for transmitting signals/messages, a memory 380 for storing operating instructions for execution by a processor 390. Such operating instructions may include policies for participation in the network, software that renders the node sufficiently intelligent to make decision that increase the node's benefit etc. The components of the node are connected via a bus in the illustrated embodiment.

Figure 2b) illustrates the architecture of the network management unit 320. The network management unit 320 also comprises a receiver 400 for receiving messages from the nodes 310, as well as a transmitter 410 for transmitting to the nodes 310. The management unit 320 moreover comprises a processor 420 and memory 430. The processor 420 is arranged to operate based on instructions stored in the memory 420. The memory 420 may further be used to store the credit balances of the nodes 310 in the network.

Reputation records for the nodes may further be stored in the memory 420.

However, it is also envisaged that the reputation records can be stored outside of the management unit 320 and may be obtained by the management unit 320, for example via a data request. The components of the network management unit 320 are connected via a bus in the illustrated embodiment.

Figure 3 illustrates the process based on which the network nodes in the ad-hoc network should or might operate. A network node receives a data packet in step 100 and either fails to forward the data packet (110), be that deliberately or because the network conditions prevent such forwarding, or forwards the data packet (130). The network node generates a receipt from the data packet.

This does not need to expose the content of the message. The receipt size can moreover be very small to minimise the amount of storage space within the network node for temporarily storing the receipt and also minimises the amount of energy consumed in transmitting the receipt onward. If the data packet is not forwarded, then the receipt is also not forwarded (120) if the network node operates as it is intended to. If the network node forwards the data packet, then the receipt is forwarded (140) to a management unit/central clearance server (CCS). If all nodes operate in the intended fashion, then the submission of receipt to the management unit means that the previous node has forwarded S the message successfully. In step 160 the management unit (referred to in this application also as a central clearance service, CCS) collects the receipts sent by the network nodes and determines, amongst other things as discussed below, an increase of the credit balance for the nodes involved in the transmission. The method thus eliminates the possibility of individual nodes lying about the service they provide.

Figure 4 illustrate the function of the network in one embodiment. A node intending to send data/a message sends a routing request that comprises source and destination addresses to other network nodes in preparation for the data transmission. Known routing routines (such as reactive routing protocols, for example DSR) can be used for determining the transmission route for the message. Once a network node has received (170) a routing request, the network node determines (180) whether or not it is on the transmission path and, if not, determines the amount of remaining energy/battery power, the remaining credits the node has for sending its own messages and the transmission power required for forwarding the message. The node then rebroadcasts the route request together with this information.

it will be appreciated that in this manner a number of routing request may be forwarded through the network. Some or all of these routing requests will end up with the destination node. The routing requests received by the destination node comprise the information appended to it by the intervening nodes that have rebroadcasted the routing request. When these routing requests reach the destination node, the destination node determines (190) the total price of each path based on the appended information, and picks the path with the lowest price. The destination node can then reply back to the source node using the reverse of the chosen path. The manner in which the price is determined for individual network nodes in one embodiment is discussed in more detail below. If an individual node is not on the chosen transmission path (200), then the network node returns to awaiting the next transmission request, as indicated in Figure 4.

Individual network nodes, being rational and selfish, will be keen to maximise their credit balance. One way of achieving this is to append information to the routing request that causes the declaration node to determine a high price for forwarding the message. However, if the determined price is so high that it increases the cumulative price of the route comprising the highly quoting node to exceed that of another route, then the route will not be chosen for data transmission and the selfish node will not be able to obtain any benefit.

The embodiment thus provides motivation for nodes to provide truthful information that allows calculation of the real costs of forwarding messages.

Steps 100 to 160 shown in Figure 4 correspond to those discussed above with reference to Figure 1 and are those steps performed by the nodes on the transmission path. These steps will not be discussed again in detail at this stage.

Once the management unit has received all receipts, the management unit is in a position to determine (210) whether or not the data transmission has been successful. Should this be the case, then the payment due to each node can be calculated in a straightforward fashion and awarded to the nodes along the transmission path. At the same time the sending node can be charged appropriatey.

Should the transmission have been unsuccessful, then the method of the embodiment checks (220) if there is a likelihood that the nodes have been colluding to obtain credit reward without adequate contribution to data transmission. The manner in which this check can be performed is discussed in further detail below. Should a likelihood of collusion be determined, then the method proceeds to calculate the payment for the nodes. It will be appreciated that, in the case of collusion this should involve a decrease in the credit balance of the colluding nodes.

If the collusion check 220 determines that there is no, or only a low likelihood of collusion, then the reputation records of some or all of the nodes along the transmission path are checked and the node likely responsible for the transmission failure is identified. The nodes can then be adequately remunerated or punished in accordance with the determination made. The relevant remuneration is transferred to the node's accounts in step 250 and the reputation records of the nodes are updated.

Turning now to determining the costs of the transmission, from an individual user's perspective, maximizing message transmission opportunity is key. Therefore, routing algorithms should not only enable individual nodes to increase transmission opportunities by gaining credits, but also help lengthen the network life time by saving energy in critical paths in the network. The transmission network can be expressed as a set of nodes: (1) The routing algorithm of one embodiment addresses the above requirement. Assume node c sends a packet to node t If Vj includes the transmitting power used in the packet header, and 5 can determine the minimum received power required, then the minimum transmitting power of -can be calculated. p71fl.. is the real consumed energy in the transmission, depending on the distance between v and 4. In the embodiment, when a node broadcasts the routing request, the receiving node advertises the lowest power level that can be used for transmitting the message in the packet header, as mentioned above.

If a node has very low level of remaining battery power, naturally it is not willing to forward for other users. To account for this the manner in which the embodiment determines the costs of forwarding to other nodes can include a variable that describes the willingness of a node to be cooperative. This variable includes the remaining energy. lf E is the power left in a node, then the unwillingness could, in one embodiment, be described as E. This metric can be the price demanded by individual network nodes, as reported in the routing phase. Thus, if remaining energy is low, the system needs to pay more for its service. If data transmission is successful, a node with low remaining power should be compensated more because of the relatively large amount of remaining energy it had to sacrifice for the data transmission.

Maximizing credits is, however, not the only goal for users/network nodes. A metric NE may be used to represent the number of packets allowed to be sent by a remaining energy level of a network node. A metric NC may be used to represent the number of packets allowed to be sent based on a credit balance of the network node. If the remaining energy levels permit the sending of more packets than the remaining credits (NE > NC), then the node may be willing to cooperate by forwarding messages from other nodes to gain some more credits, so that the energy that, in the absence of such new credits could not be used for the transmission of packets, can find use in packet transmission.

NC

The metric W may thus be used to model a degree of willingness of the node to participate in data relaying. An increase in the metric is associated with decreasing willingness of the network node to participate in the relaying of messages. The amount of compensation/payment required for motivating the node to participate in data transmission increases with increases in this metric.

The unwillingness (also representing the price) of a node 17E to forward messages can be expressed by: (2) Where coefficients a' and b' are two constants which are weights that could be tuned for different purposes of system performance. By making a' larger than b', for example, equation (2) can be weighted towards emphasising a nodes unwillingness to forward packages based on the remaining power in the node. By making b' larger than a' greater emphasis can be placed on the balance between available credits and available energy. The cumulative price for the transmission of data along a route comprising M network nodes (including the sender) can thus be: = 1=11 (3) The path with the lowest cost can be chosen as the path for data transmission. This solution provides a good compromise by accommodating communication demands while satisfying the concerns of individual users.

As discussed above) the ad-hoc network relies on a management unit for maintaining the credit balance of the individual nodes as well as an indication of the reliability of the nodes. This management unit is referred to in the figures as CCS, a Central Clearance Server. The CCS does not need to possess any credits itself. The energy consumption of the nodes related to control and communication for CCS activities can be assumed to be negligible when compared to that of data transmission. Alternatively each node can demand, as pad of the cost of forwarding a message, an overhead cost associated with the CCS operations/communications.

The payment made to each node is an increase in the credit balance stored within the CCS. Non-cooperative nodes will be punished either by a refusal on part of the CCS to increase the node's credit balance, or by a decrease in the node's credit balance.

In the embodiment the transmitter is required to pay for the transmission of the data. Charging the transmitter provides a motivation to the transmitter to transmit data efficiently. This motivation would not be present if it were the receiver that had to pay for the transmission, although an embodiment in which the receiver pays is also envisaged. The CCS "pays" the individual nodes based on confirmation message received from the nodes. A forwarding node hereby relies on the receiver or next hop node to send a report to the CCS upon receipt of packets.

Figure 5 illustrates the process of an embodiment, wherein node 1 sends a packet to node 6. In Figure 5 as well as in the following description relating to Figures 6 to 9, Ci is the price node i has agreed with the system at the routing phase as appropriate payment for forwarding a data packet. The price C2, for example is the price node 2 has agreed with the system. Because all users have submitted the receipts, it means the packet has been successfully received by the destination node. Due to the contribution of all intermediate nodes, every one obtains the credits they needed. And in exchange of services received from others, node 1 is charged for the sum of payments to each node as -(C2+C3÷C4+C5)', where -denotes charge or punishment which will decrease the affected node's credits. Since node 6 is the beneficiary of the data transmission no credits will be given to it.

Three kinds of misbehaviours could exist in the ad-hoc network, if no mechanism is used for countering such misbehaviours. The nodes can refuse to forward packets, not reporting receipts and collusion.

Not forwarding the packet Forwarding and submitting the receipts are what the system expects all users/nodes to do. Some node may want to preserve energy by dropping packets. Such behaviour would be typical selfish behaviour.

Not reporting the receipt Nodes that dropped a packet may lie about the receipts to avoid punishment for not forwarding the packet. Nodes could, for example, pretend not to have received the message, thereby seeking to put the responsibility for a failure to forward the message onto the previous node.

Collusion A number or all nodes could collude to deceive the CCS. The nodes could, for example, fail to forward the message while still sending receipts to the CCS, thus expecting to be paid without real contribution. The destination node would detect and flag up this situation by not submitting a receipt.

To counter the three misbehaviours mentioned above, the embodiment is designed to motive the forwarding of messages as well a the reporting of receipts.

Motivation for Forwarding Messages Assume node 3 drops the packet. The payment should be made as shown in Figure 6. Since node 4 did not receive the message, only two receipts are submitted, those of nodes 2 and 3. As node 3 has failed to forward the message it is punished by decreasing the credit balance of node 3 by the cost CS the node itself had originally indicated it would charge for forwarding the message. The same amount is awarded to the respective credit balances of the sending node, node 1, and intermediate node 2, in this embodiment in equal parts. This allocation effectively works as a compensation payment for the loss of energy of node I and node 2 resulting from the non-cooperating of node 3.

This punishment encourages nodes to forwarding data in the intended manner.

The credit balance of node 2 is additionally increased by the amount C2.

This is the amount determined during the routing phase as being an appropriate reward for node 2 for forwarding the message. Nodes 4 and 5 are not affected because they have not been part of the process yet.

Motivation for Reporting the Receipts Figure 7 relates to a situation where node 4 fails to forward a packet and also omits sending a receipt of the CCS. It will be appreciated that in this situation the receipts received by the CCS are the same as those received in the scenario discussed above with regard to Figure 6, where node 3 violated the forwarding rules. To distinguish between these two scenarios the embodiment uses an indication of the reliability of the individual nodes (discussed in more detail below). It this indication of the reliability of the nodes indicates that node 3 is more reliable than node 4, then it.is reasonable to assume that node 4 is the node refusing to cooperate and that, consequently the situation shown in Figure 7 is more likely to have occurred than the situation shown in Figure 6. If of course the indication of the reliability of the nodes shows that node 3 is the less reliable node, then it is likely that the situation shown in Figure 6 has occurred and that consequently node 3 should be penalised instead of node 4.

If node 4 is determined to be at fault, then the credit amount C4 (this being the originally agreed payment for the assistance of node 4 in forwarding the message) will be deducted from the credit balance of node 4 and distributed (in equal parts in the embodiment) to the preceding node in the transmission route: The two preceding nodes, nodes 2 and 3, intermediate the sending node, node 1, and the defaulting node, node 4, are again additionally rewarded for their participation in the data transmission process and the CCS increases their credit balance by the originally agreed amounts C2 and C3 respectively.

By basing the above discussed punishment mechanism on a reliability/reputation indicator nodes misbehaving by not reporting receipts can be identified and adequately punished, thereby discouraging nodes from dropping packets and trying to disguise such misbehaviour by refusing to report the required receipt.

Preventing Collusion If a number of nodes collude to defraud the transmission system of the embodiment the destination node, node 6, may not submit a receipt, while all other nodes submit a receipt. Consequently, even in a situation where all other nodes submit receipts (thereby claiming that they have received and forwarded the data) it is stilt possible that all nodes are colluding to defraud the system at the cost of the original transmitting node, node 1.

Whether colluding or not, since node 5 has not forwarded the message, it will be charged. However for other nodes, because it is still possible that only node 5 is the non-cooperative user penalizing them is not fair. In cases where nodes are colluding no one will be rewarded. Collusion thus does not create income for the colluding nodes and leads to a net loss for the nodes, as they cannot cover the costs of forwarding the receipts. CS is paid to the sender, node 1.

Should the reputation record maintained by the CCS indicate that nodes within the transmission path have good reputation, then this indication may be used as a basis for calculating payment and penalties even in situations where it is likely that nodes are colluding. Nodes with a good reputation may be assumed to be operating in the intended manner by forwarding data and providing adequate receipts. Any fault/misbehaviour may therefore be assumed to occur downstream of such an honest' node. As a consequence the CCS may award the agreed payment to any nodes upstream of the honest' node as well as to the determined honest' node itself. Any nodes further downstream may be treated in the manner indicated in Figure 8, namely by refusing the payment of any reward to them, with the final node in the transmission chain, node 5 in Figure 8 being penalised in the manner shown in Figure 8.

Maintaining a reputation record Different ways of establishing a reputation record are envisaged. In one embodiment the CCS records the last node to submit a receipt and the first node not to submit a receipt each time a data transmission is unsuccessful (as, for example, shown in Figure 9). Such a record can provide useful information regarding the behaviour of the node. If, for example, a particular node, say node 3, is always the last one to report a receipt, this node may be one that drops packages, yet submits receipts that indicate that the packet has been sent on to the next node along the route. Consequently the CCS may be arranged to punish this particular node in a situation where the node is again the last one to report a receipt. If, however, a further particular node (say node 4, if node 3 is taken to be the last node submitting a receipt) often is the first one not to submit a receipt when operating with a particular node (in this example node 3), then the probability that this particular node, node 4, has received the packet and simply refuses to forward it and to submit a receipt is high. Consequently it is node 4, rather than node 3 that will be penalised by the CCS.

In one embodiment two threshold values for reporting frequencies can be used for each node. For the purpose of this explanation these threshold values are referred to as A and p. A represents a threshold value for the frequency of a node being the last node reporting a receipt. p is a threshold value for the frequency of the node being the first one not to report a receipt.

Table I provides an example scheme for awarding payments and penalties to nodes 3 and 4 of Figure 9. If the frequencies Fl of each of node 3 and node 4 of being the last one to report is below the threshold value A, and if the frequency F2 of each of nodes 3 and 4 of being the first one not to report a receipt is below the threshold value p, then no decision as to which one of nodes 3 and 4 bear responsibility for the loss of the packet can be made. The packet loss may in any have come about without any irregular actions having been taken by a node. The packet may, for example, have been lost due to a change in transmission condition, such as a change brought about by an alteration in the network s topography which can make transmission more difficult or even impossible. As a consequence neither node 3, nor node 4 receives payment or punishment.

If node 3 is known to often be the last node to report a receipt (Fl > A as shown in the second line of Table 1) and node 4 is not notorious for being the first node not to report receipts (F2 <p in the same line of Table 1), then node 3 can reasonably safely be assumed to be the node that has dropped the package and that lies about forwarding it by submitting a receipt. Consequently the situation illustrated in Figure 6 can be assumed to exist and the payment scheme can thus be that shown in Figure 6.

Conversely, if node 3 is not known for being the last node to report a receipt (Fl <A as shown in the third line of Table 1) and node 4 is notorious for being the first node not to report receipts (F2> p in the same line of Table 1), then node 3 can reasonably safely be assumed to have forwarded the packet.

Node 3 should consequently be paid in the agreed fashion, by receiving C3 credits, as originally agreed. Node 4 in contrast will be assumed to have dropped the packet as this node is known to be the first node not to report any receipts. Consequently the situation shown in Figure 7 can be assumed to exist and node 4 should be punished accordingly.

Finally, if both nodes are known to exceed the threshold values (Fl > A, F2> p, as shown in the last line of Table 1), it can be assumed that, in situations where data transmission was unsuccessful, they are involved in irregular behaviour. In this case both nodes can be decided to deserve punishment for the failure of data transmission in accordance with the last line

of Table 1.

RePutation Payment Scheme J Record Node 3 Node 4 -F1<k, F2cp 0 0 F1>A, F2<p -C3 0 F1'cA, F2>p C3 --C4 F1>A, F2>p -C3 -C4 Table 1: Payment Scheme according to the Reputation Record It is appreciated the above described way in which the CCS arrives at payment and punishment decision is one that is based on an estimation of the probability of a node being at fault for the failure of the entire transmission path delivering a packet. Situations in which a node is incorrectly assumed to be at fault may thus occur. However, each node can increase its reputation (for example by reducing the values Fl and F2) by operating in the fashion intended by the system designers, so that the node itself can work towards reducing the likelihood of incorrect punishment. The manner in which the CCS monitors the reputation of the nodes thus provides a powerful incentive to nodes to operate in the desired fashion.

In the above description reference is made to a frequency' at which nodes are the last ones in a transmission path to report receipts or the first ones not to report receipts. In one embodiment this frequency is expressed in the form of a percentage of the total packet transmission operations a node has been involved in, rather than in the form of an absolute value. In this way nodes that are at the periphery of an ad-hoc network (and would consequently neither be involved in as many packet transmissions as nodes located less towards the periphery of the network, nor would they have the opportunity to drop packets as often) are treated in a similar fashion as nodes closer to the centre of the network. Alternatively, the frequencies at which the nodes are the last ones to report a receipt or the first ones not to may be stored in the form of a time average.

The COS may further take the overall network conditions into account. If, for example, the CCS receives information that that connectivity within the network is poor (indicated, for example, by the need for a large amount of minimum power for the nodes to achieve the desired forwarding of the packets or if the overall packet drop rate is higher than normal/average), then a failure in transmission may not lead to a punishment of the nodes within the transmission chain1 even if one of the above described scenarios applies. In this case the CCS may simply override the decision/punishment mechanism.

The CCS may further take the topology of the network into consideration, in particular in a less mobile topology. If, for example, a node is non-cooperative and always drops packets, its neighbour who is in the location that always receives the message from it, may also have a bad reputation for not reporting the receipts.

Examples

Figure 10 illustrates a 100 m x 100 m field with 10 nodes that has been used as the input for a computer simulation of the above described ad-hoc network and CCS. The topology is fixed in the manner shown in Figure 10 for all of the simulations discussed below.

The network life-time and packet success rate are used as the two main metrics to benchmark the performance of the simulated ad-hoc network.

Network lifetime has been considered in the art as the amount of time that the network can satisfy its coverage objective. For the purpose of the present disclosure, however, this definition simplified and the lifetime measured here is the time period between the simulation start time and the time when the first node stops working. The simulation is stopped at this point. A node stops working if its remaining energy is insufficient to continue transmitting further packets.

All nodes are assumed to have the same amount of energy stored at the beginning of the simulation. The amount of energy required for transmitting packages is dependent on the distance between nodes. The amount of energy required for transmitting the receipts to the CCS is small when compared to the amount of energy required for forwarding data packets and has consequently not been included in the simulations.

In the present description the packet success rate refers to the percentage of packets successfully received out of the total number of packets sent by all the nodes in one second. Overall, longer network life time and higher packet success rate are the main measures or indicators of the performance of the proposed solution.

Figures 9a) to d) illustrate the average packets success rates for the entire network (including node 4) and for node 4 in isolation in a different situations. Figure II a) illustrates a situation where the nodes operate in accordance with the above described embodiment. Figure 11 b) illustrates a situation where node 4 refuses to forward messages. Figure 1 Ic) relates to a network in which node 4 refuses to report receipts. Figure 6d) shows simulation results for a network in which node 4 reports a receipt despite colluding with other nodes not to forward the packet.

A comparison of Figures 9b) to d) with Figure II a) shows that the selfish behaviour of node 4 leads to a remarkable reduction in packet success rates.

Conversely, the same comparison shows that, if the embodiment is used, the behaviour of node 4 improves1 thereby improving the overall success rate of the network. This is because the embodiment detects selfish behaviours and punishes such behaviour by decreasing the credits of the nodes at fault.

Rational nodes will choose a strategy they benefit from most. In the embodiment selfish behaviour is no longer attractive, as it does not bring about any benefits to the nodes. Rational nodes will thus choose to behave in the manner they are intended to.

Figures ba) to c) show a comparison (Figure I2c)) between a known energy efficient routing algorithm (Sheetalkumar Doshi, Shweta Bhandare, and Timothy X Brown, "An On-demand Minimum Energy Routing Protocol for a wireless Ad Hoc Network", Mobile Computing and Communications Review, vol 6, 2002; Figure 12a)) and the above described embodiment (Figure I2b)). As can be seen from these figures, the scheme of the embodiment outperforms the conventional credit-based solution in terms of packet success rate. The new scheme improves the system lifetime (i.e. from 600 s to 3000 s) because the credits and energy combined utility optimization prevents users from dying quickly due to over-used resources. Further, the new solution not only achieves a success rate over 60% during the first 500 s, but also continues to operate with over 40% success rate during the first 1000 s while the energy efficient routing scheme reduced to below 50% at the beginning of 50 s. The higher success rate is due to the balance credit distribution among the nodes which is achieved by the above discussed combined credit and reputation metric.

Figures hand 12 compare the data shown in Figure 12b) with further simulations of the above described embodiment. These figures show that topology has a direct impact on the network performance and hence on cooperation.

As is shown in Figure 13 the performance in the case of a uniform topology (i.e. uniform distribution of nodes) outperforms the non-uniform case (i.e. random distribution of the nodes). This is understandable since in the uniform case the number of randomly located users (which are not able to participate in the network or collaborate with other nodes easily), such as those in a corner or very far away from other nodes, is reduced.

Figure 14 is based on a simulation that is based on the assumption that all nodes move once every 15 minutes, so that during the simulation time of 50 minutes, four different topologies were present. As can be seen from this figure, the performance of the network is better for mobile users.

As increase in successful message transmission leads to more energy consumption and thus inevitably to shorter network life time. Consequently packet success rate and network life time cannot be maximised at the same time as they are two contradictory goals. However, a level of compromise between these two goals could be controlled through the co-efficient a' in equation (2). Figure 15 shows to the results of simulations for different values of this coefficient a', as indicated in the figure.

In Figure 15, the life time of the network is approximately 440secs, I 200secs and more than 3000secs for the value of a' being 200, 400, 600 respectively. However, due to the trade-off between the packet success rate and network life time, larger a' causes reduction of success rate.

The above described credit-reputation hybrid approach may be able to successfully stimulate node cooperation and more importantly actively discourages three kinds of selfish behaviours. System overhead may be reduced because of the balanced payment scheme. Longer network life time S and higher packet success transmission rate may be achieved, thereby improving node level and network level performance by evenly distributing the energy consumption and credit balance of all users.

While certain embodiments have been described, the embodiments have been presented by way of example only, an area not intended to lim the scope of the inventions. Indeed, the novel methods, apparatus and systems described herein may be embodied in a variety of other forms; furthermore, various omissions, substitutions and changes in the form of the methods and systems described herein may be made without departing from the spirit of the inventions. The accompanying claims and their equivalents are intended to cover such forms or modifications as would fall within the scope and spirit of the inventions.

Claims (10)

1. Claims: 1. A wireless ad-hoc network comprising: a plurality of network nodes; and a management apparatus comprising a memory, a processor and a receiver for receiving confirmation messages from network nodes; wherein the network nodes comprise a policy for sending a said confirmation message to the management apparatus for each received message; wherein the management apparatus is arranged to store a credit balance for each network node in the memory, said processor arranged to update said credit balance; wherein the management apparatus is further arranged to store and/or obtain an indication of the reliability of each of the network nodes; wherein said processor is operative to update said credit balance based on confirmation messages received from the network nodes in the data transmission route and based on the indication of reliability of the said network nodes.
2. 2. A network management apparatus comprising a processor, a memory and a receiver for receiving confirmation messages from network nodes, wherein the processor is arranged to store in and retrieve from the memory credit balances for individual nodes in the network and to amend the credit balance based on confirmation messages received from the nodes and on reputation records of individual nodes.
3. 3. A network management apparatus according to Claim 2, wherein, following notification of a transmission failure, the processor determines, based on said reputation records, a likelihood for a node, to retrieve the credit balance of the node from the memory, to alter the credit balance of the node based on the determined likelihood and to store the altered credit balance in the memory; wherein the likelihood is a likelihood of the node being responsible, or partially responsible for the transmission failure.
4. 4. An apparatus according to claim 2 or 3, wherein the processor is operative to, following notification of a transmission failure, identify the last network nodes along a known transmission route from which a confirmation message has been received and/or the first network node along the transmission route from which no confirmation message has been received and to alter the credit balance of one or both of the identified nodes based on the reputation record of the two nodes.
5. 5. An apparatus according to claim 2, 3 or 4, wherein the processor is operative to, upon notification of a failure to transmit a message, detect colluding non-forwarding behaviour of one or more of nodes within a data transmission route based on confirmation messages received from network nodes along a known intended data transmission route of the message.
6. 6. An apparatus according to claim 5, wherein said processor is operative to conclude that some or all nodes along the route are involved in colluding non-forwarding behaviour if it has received confirmation messages from all nodes along the route but not from an intended destination node of the message.
7. 7. An apparatus according to any of claims 2 to 6, wherein said processor is operative to generate and update said reputation record or reputation records of a said node by noting, in case of a transmission failure, if the node was the last node from which a confirmation message has been received at the receiver and/or the first node from which no confirmation message has been received at the receiver.
8. 8. An apparatus comprising a transmitter for receiving data packets, a receiver for transmitting data packets and a processor, the apparatus operative to forward a received data packet to another apparatus in an ad-hoc network in accordance with received routing instructions; wherein the processor is arranged to determine, during a routing phase, a cost for forwarding a message to another apparatus in an ad-hoc network S based on a current credit balance and on a remaining energy reserve.
9. 9. A method of operating an ad-hoc network comprising: sending confirmation messages from network nodes along a transmission route for a message to a network management apparatus; within the network management apparatus updating a stored credit balance for network nodes along the transmission route based on received confirmation messages and, if a notification of the failure of the transmission has been received, also based on a reputation records of the network nodes along the transmission route.
10. 10. A method according to claim 9, further comprising during a routing discovery phase and within a said network node, determining a cost of forwarding a data packet to another network node based on a current credit balance of the said network node and based on a remaining energy reserve of the said network node.