CN103118379A - Node cooperation degree evaluation method facing mobile ad hoc network - Google Patents

Abstract

The invention discloses a node cooperation degree evaluation method facing a mobile ad hoc network and is used for resolving the technical problem of vast sample requirement of existing node cooperation degree evaluation methods. The technical scheme is that: firstly modeling node network actions, and dividing network actions depicting nodes into four types of index sets including network communication characters, wireless signal characters, moving characters and n spatial characters; then performing information acquiring, initial evaluation algorithm and recommendation comment exchange; and at last performing gray cluster evaluation and taste cluster decision. The method solves the problems of 'few data uncertainty' of 'few sample information sources and incomplete information' under a moving ad hoc network environment, adopts the gray cluster evaluation method, solves the problem of vast sample requirement of existing evaluation methods, and is simple in algorithm and easy for engineering to achieve.

Description

Node cooperation degree appraisal procedure towards mobile ad hoc network

Technical field

The present invention relates to a kind of node cooperation degree appraisal procedure, particularly relate to a kind of node cooperation degree appraisal procedure towards mobile ad hoc network.

Background technology

Mobile ad hoc network is a kind of multi-hop, without the wireless network of center, self-organizing.Along with the fast development of wireless communication technology and computer technology, mobile self-networking technology is more and more paid attention to.Compare with legacy network, mobile ad hoc network needs work in concert (in network, each node is not only terminal, is also a router) of network node.In legacy network, route is mainly completed by firmwares such as router, gateways, is that to belong to equipment reliable; And in mobile ad hoc network, each user node is equivalent to one " router ", these nodes be all can " be free to come and go ", " moving freely ", the data retransmission between node relies on the cooperation behavior of intermediate node fully.But for the terminal node of participation network operation, tend to the demand according to number one, selectively network is made contributions, such as only " enjoyment " not " devotion " of the node of selfishness; Some malicious nodes are carried out black hole, grey hole, worm hole etc. routing attack targetedly for another example, all can cause the sharply decline of network performance, even the result of paralysis.Thereby, the cooperation of the how behavior of standard node, excitation node, prevent that the Node selfishness behavior from becoming the critical problem of needing at present solution badly.

On for this problem, two kinds of solutions are arranged at present, a class is to take credit mechanism (credit-based models), and is another kind of based on Fame Mechanism (reputation-based models).Wherein, individual A refers to that to the trust of individual B the individual B of individual A expectation is the subjective possibility of A service (namely carrying out the action that the interests of A rely on).Reputation is based on the observation of certain individual historical behavior or evaluation information and the expectation to this individuality future behaviour that draws, and is presented as the cooperation degree of this node in mobile ad hoc network.

The basic principle of credit-based mechanism is, node will obtain the service of network, must defrayment, i.e. credit (concrete form is chip or ideal money), credit can obtain by service (such as route, forwarding data bag etc. are provided) is provided for network.This class scheme is if be applied in reality, need to provide a hardware module with anti-tamper function for each node, perhaps need to realize the role of a bank in network, as trusted third party (trusted-third party), provide the services such as credit storage, payment for node.The problem of this class scheme ubiquity poor expandability.

Basic thought based on the evaluation mechanism of reputation is: for any one node N, the evaluation to its behavior of the network behavior of carrying out according to N and other nodes, be that this node distributes a cooperation desired value, make other nodes to make the decision-making whether network behavior occurs with N according to the cooperation desired value of N.This scheme need not hardware supports in realization, extensibility is strong.

The key problem that need to solve based on the evaluation mechanism of reputation can be summarized as 4 points:

1) which network behavior sample can be used to assess cooperation degree (reputation), and wherein the sample evidence source can be divided into direct sample and indirect sample;

2) how to generate, obtain and assemble the information of relevant these network behavior samples;

3) can evaluation mechanism resist the individual attack of malice;

4) how the cooperation degree that calculates (being reputation) is used.

Comprehensive existing scheme, the following problem of its ubiquity:

1, the quality of sample.Most schemes are taked monitoring mechanism, are promiscuous mode with Node configuration, and neighbors is monitored, and whether observe has abnormal behaviour to occur, and form sample with this.Here existing problem is, can't decision node extremely by " subject intent "---attack or selfish causing, or by " odjective cause "---network congestion, link failure etc. causes, thereby collected sample must have partial distortion, thereby affects follow-up assessment algorithm.

2, assessment algorithm.As described in problem 1, in MANET, node obtains the means scarcity of sample, and sample size and of low quality if in the process of information fusion, is not considered this situation, will cause result of calculation not accurate enough.At present existing scheme majority is taked the account form of Based on Probability statistics, but probability statistics having relatively high expectations to sample size and quality; Another part adopts the account form based on fuzzy theory, but fuzzy theory requires have enough " prior information " to carry out sample training, and in mobile ad hoc network, network topology structure changes frequent, node is free to come and go, and does not have enough " prior information " to carry out sample training.Therefore, need to redesign assessment algorithm.

3, the use of indirect sample.Although some schemes are not by adopting indirect sample to reduce the complexity of agreement, the sample information that is used for the decision node behavior has reduced, and must have influence on the accuracy of assessment algorithm.Although other schemes adopt collateral information, the problem for existing malice to estimate does not provide effective solution.

Summary of the invention

The deficiency that needs great amount of samples in order to overcome existing node cooperation degree appraisal procedure the invention provides a kind of node cooperation degree appraisal procedure towards mobile ad hoc network.The method is had relatively high expectations to the quality and quantity of sample for existing appraisal procedure, and some algorithm also needs " prior information " to carry out sample training.And in mobile ad hoc network, it is frequent that network topology structure changes, and node is free to come and go, and can not meet these requirements.The present invention is directed to " minority is according to the uncertainty " problem of " the sample information source is few; INFORMATION OF INCOMPLETE " under the mobile ad hoc network environment, adopt grey Cluster Evaluation method, can avoid existing appraisal procedure to need the problem of great amount of samples, and algorithm is simple, is easy to Project Realization.

The technical solution adopted for the present invention to solve the technical problems is: a kind of node cooperation degree appraisal procedure towards mobile ad hoc network is characterized in comprising the following steps:

(1) to the meshed network behavior modeling, the network behavior of portraying node is divided into four class index sets: network service characteristic, wireless signal characteristic, mobility and spatial character.

(2) acquisition of information is sampled according to index set by node.Node is intercepted this neighbors and whether is forwarded this packet sending to the packet of neighbors to be placed in buffering area.If forwarded this packet at the in setting time neighbors, and be complementary with packet in buffering area, proof successfully forwards, the packet in the buffer release district; If do not mate, illustrate that neighbors revised this packet, abnormal behaviour is arranged.If interior nodes does not all listen to neighbors and forwards this packet at the appointed time, wrong generation is described, or being that this node is uncooperative give not to forward, or is that channel confliction is arranged, two kinds of reasons all be summed up as forward unsuccessful.All obtain by MAC layer or physical layer as for channel problems such as the signal strength signal intensity of neighbors, network congestion, communication contention awares.

After the acquired information sample, initial evaluation calculates scoring by the method for probability statistics, the number of times that route data in a period of time is forwarded success and failure is added up, and calculating successful probability by formula is p, makes initial evaluation with this data retransmission index to neighbors.

After as node A, its neighbours being collected Neighbor (A) and making initial evaluation, combine by Event triggered or with periodic refreshing, broadcast in two hop distances, can obtain with the neighbours that guarantee all nodes in Neighbor (A).

After node A receives the evaluation of other nodes to neighbors B, carry out preliminary treatment by the flavor swarm algorithm, reduce malice and estimate the impact that degree of belief is calculated.Selected core loop evaluation is as reference, common evaluation will compare with the core evaluation, exceed certain limit if depart from, and think that this evaluation is invalid, when the invalid evaluation of this entity node surpasses certain threshold value, will assert that this entity node is malicious node.Otherwise, if estimate effectively, will accept evaluation, if this evaluation simultaneously near the core evaluation, and when reaching certain number of times, the evaluation weight with this entity node will increase so.

(3) assessment initiate assessment sample that node will receive distinguish the flavor of assemblage classification with choose, right cluster data are carried out grey clustering analysis, draw final conclusion by the decision-making of flavor cluster at last.

1. node T ₀The broadcasting of initiation assessment request is with property set A={a _h| h=1,2 ..., e} tells each node, communication details is determined by assessment communication protocol.

2. each assessment node carries out sample to issue T after safety verification ₀, by T ₀The cluster of distinguishing the flavor of is selected, and cluster T obtains distinguishing the flavor of ₁, T ₂..., T _q

3. to each flavor cluster T _jThe sample matrix that obtains carries out the conversion of index polarity; Be the multiattribute index polarity disunity problem in the reply mobile ad hoc network, adopt the cost residual method to carry out the polarity conversion.Cost residual method processing procedure is as follows: establish certain attribute a _i=ξ _i, and 0≤a _i≤ max (a _i), this attribute is the smaller the better to the side of assessment, is minimum polarity, makes a _i ^*=max (a _i)-ξ _i, a _i ^*To the side of assessment, be changed to and be the bigger the better, be maximum polarity.

4. attribute quantification unit is unified processes, and adopts just value method of sample.Processing procedure is as follows: be provided with node d ₁Attribute sample D ₁={ ξ ₁₁, ξ ₁₂, ξ ₁₃And d ₂Attribute sample D ₂={ ξ ₂₁, ξ ₂₂, ξ ₂₃, appoint and get node d ₁, make that its sample is D ₂ ^*={ ξ ₁₁/ ξ ₁₁, ξ ₁₂/ ξ ₁₂, ξ ₁₃/ ξ ₁₃,={ 1,1,1}, D ₂ ^*={ ξ ₂₁/ ξ ₁₁, ξ ₂₂/ ξ ₁₂, ξ ₂₃/ ξ ₁₃,, cancellation all units.After just value of sample, corresponding white function also carries out just value according to the node of choosing.

In conjunction with the communication requirement of mobile ad hoc network, define the white function of each attribute, then T ₀For each flavor cluster T _jThe sample matrix that obtains carries out grey Cluster Evaluation, obtains evaluated node d _i(d _i∈ D, the grey class g of 1≤i≤m) _ji ^*

The invention has the beneficial effects as follows: because the method is had relatively high expectations to the quality and quantity of sample for existing appraisal procedure, some algorithm also needs " prior information " to carry out sample training.And in mobile ad hoc network, it is frequent that network topology structure changes, and node is free to come and go, and can not meet these requirements.The present invention is directed to " minority is according to the uncertainty " problem of " the sample information source is few; INFORMATION OF INCOMPLETE " under the mobile ad hoc network environment, adopt grey Cluster Evaluation method, avoided existing appraisal procedure to need the problem of great amount of samples, and algorithm is simple, is easy to Project Realization.

Below in conjunction with drawings and Examples, the present invention is elaborated.

Description of drawings

Fig. 1 is that the present invention is towards the frame diagram of the node cooperation degree appraisal procedure of mobile ad hoc network.

Fig. 2 is that the present invention is towards the evaluation index definition schematic diagram of the node cooperation degree appraisal procedure of mobile ad hoc network.

Embodiment

With reference to Fig. 1～2.The inventive method concrete steps are as follows:

(1) determinant attribute definition.

Before the behavior risk of determining a node, need the definition determinant attribute, namely determine evaluation index.These determinant attributes must embody the nodes in MANET behavioural characteristic, with objective assessment and the processing of effectively carrying out the behavior risk.At first the determinant attribute definition will meet the design feature of mobile ad hoc network.That mobile ad hoc network has is fully distributed, self-organizing and dynamic topology characteristics, node in network can move freely, the service behaviour of whole network depends on the mutual cooperation degree between node, the node assessment of general networking, majority is only observed communication session or the success or not of routing forwarding, then processes in conjunction with collateral information.But for the nodes in MANET risk assessment, also should consider self physical attribute of this node except network service characteristic (as: packet loss), comprise mobility, wireless signal characteristic and spatial character etc., each specificity analysis is as follows: the network service characteristic refers to node for the characteristic of network message transmitting-receiving, is the central factor of reaction node risk level.The network service characteristic comprises transmission speed, packet loss etc., and transmission speed has embodied the service behaviour of node, and transmission speed is too low is unfavorable for network service, if it is improved as the routing node degree of risk; Packet loss is the key index of evaluation node functional reliability, and packet loss is also attacked with some specific mobile ad hoc networks and is correlated with except with the stability of self working, and as black hole attack, worm hole attack etc., is the indispensable part of assessment.Aspect the physical attribute of node, mobility major concern node movement velocity, movement velocity is faster, and packet loss can improve, and also can cause route frequent variations degree higher, is unfavorable for stable communication; The wireless signal characteristic comprises the factor of two aspects, one is the strong and weak level of signal, and signal is stronger, works more reliable, another one is the signal stabilization degree, be change rate signal, signal intensity is too fast, may be both that node motion is too fast, the interference that may be also the unstable and environment of node self causes, even node itself is an attack node, and therefore, the signal stabilization degree has also been reacted the risk class of communication; Spatial character is mainly the distance between node, and node is nearer, and transmission reliability is higher, and distance is also the factor that the communication risk need to be considered.Determinant attribute definition also needs to satisfy can monitor and the characteristics of objectivity.Can monitor and refer to that this ATTRIBUTE INDEX need possess the means of measurement in mobile ad hoc network, can be by the method for monitoring as forwarding speed and packet loss, network interface is set to listening mode, can complete the monitoring of data, and the method is equally applicable to the detection of wireless signal; Objectivity refers to that index measurement is objective, should avoid malice to forge index value, as the energy content of battery index of evaluated node, although the risk that how much can affect communication of the energy content of battery, but need node self to detect and report to the assessment node, be easier to forge.According to the requirement that can monitor, the measurement of mobility and spatial character needs special equipment, as range radar, GPS positioning equipment etc., this is difficult to satisfy in the mobile ad hoc network environment of shortage of resources, but can be according to the transformational relation of signal strength signal intensity and distance, mobile speed is reflected in the variation of wireless signal, and the distance of distance is reacted in the intensity of signal, can assess and process it equally.

In sum, the determinant attribute collection of mobile ad hoc network risk assessment can be defined as A={ packet loss a ₁, transmission speed a ₂, signal strength signal intensity a ₃, change rate signal a ₄, wherein, the availability of transmission speed and signal strength signal intensity reflection node, packet loss and change rate signal reflect communication reliability.Whole definition procedure as shown in Figure 2.it is to be noted, these attributes unit on the one hand are different, the attribute extreme value is different, the dimensionless number certificate as packet loss, be worth the smaller the better, but transmission speed unit is the kilobytes per seconds, value is the bigger the better, in addition on the one hand, each node performance characteristic can't be fully definite, as low in node A packet loss, but transmission speed is slow, the Node B packet loss is high, but signal strength signal intensity is considerable, simultaneously due to the node dynamic, the sample that gathers can not be a lot, therefore can't directly assess by numerical statistic or the method for probability, adopt grey clustering analysis can more efficiently carry out the cluster judgement.

(2) definition of grey class set.

According to purpose of appraisals, define grey class set G ₁={ g ₁, g ₂, g ₃, g wherein ₁The cooperation degree that represents this node participation network is very high, i.e. " high cooperation degree ", g ₂Expression " general cooperation degree ", g ₃Expression " low cooperation degree ".

(3) white function definition.

For each evaluation index, demand and network index requirement according to communication define its white function.Suppose and adopt the 802.11b wireless device to set up mobile ad hoc network, its desirable transmission speed is as shown in table 1 with distance,

Table 180211b distance and length velocity relation table

Sequence number	Distance/m	Speed/Mbps
			1	160	11.0
2	270	5.5
			3	400	2.0
4	550	1.0

In actual applications, usually have half " raw velocity " be grouped load, verification and, framing bit, the mistake information of recovering data and other " useless " takies, even do not consider spread scope and barrier weakening the influence to performance, it is even lower that actual throughput rates also only can reach half of maximum transmission rate, be about 0.5Mbps～5.5Mbps, in actual working environment, the outdoor efficient working range of 802.11b is 300m, surpasses 550m and thinks that transfer of data is unreliable.According to this principle, simultaneously associative list 1 as can be known, for transmission rate, be unreliable less than 0.5Mbps (63KB/s), is that reliability is high greater than 5.5Mbps (688KB/s), and about 2Mbps (250KB/s) for generally reliable; According to the general networking evaluation index, packet loss thinks that less than 10% the degree of reliability is high, and 30% left and right is normal packet loss, thinks that greater than 70% reliability is low; For signal strength signal intensity, the transmitting power of the 802.11b network equipment can not surpass 20dBm according to national radio management stipulation, be 100mW, therefore establishing network equipment transmitting power is 100mW, according to the free space model, calculates in signal power corresponding to different distance.In the free space model, transmission range and power attenuation are closed and are

$\mathrm{FSPL}\left(\mathrm{dB}\right)={10\log }_{10}\left({\left(\frac{4\mathrm{\π}}{c}\mathrm{df}\right)}^2\right)$

$={20\log }_{10}\left(d\right)+{20\log }_{10}\left(f\right)+{20\log }_{10}\left(\frac{4\mathrm{\π}}{c}\right)$

$={20\log }_{10}\left(d\right)+{20\log }_{10}\left(f\right)-147.56,$

Wherein, the power of FSPL (dB) expression decay is take dBm as unit, and d represents distance, and c represents the light velocity, and f represents frequency, and 802.11b uses the 2.4GHz frequency.

According to above-mentioned formula and operating distance, as can be known at 160m, the decay intensity at 300m and 550m place is respectively 84.74dB, 90.21dBm and 95.47dBm, and the network interface card transmitting power is the 20dBm left and right, therefore, when the signal strength signal intensity that detects greater than-64.74dBm (3.36 * 10 ^-7MW) time, can think that this node signal is more reliable ,-70.21dBm (0.953 * 10 ^-7MW) this node is for generally reliable left and right time ,-75.47dBm (0.284 * 10 ^-7MW) time, think that this node signal is unreliable.for change rate signal, judge in conjunction with rate travel, might as well be with general document transmission as judgement, suppose that the file that needs in mobile ad hoc network to transmit is 10MB to the maximum, at average 2Mbps(250KB/s) speed under need 40s, move away from mode with the straight line that has the greatest impact, if the node 550m that moved within this time, can think that the document can't complete transmission, at this moment, change rate signal is 95.47/40=2.39dBm/s, be that change rate signal is low-risk greater than 2.12dBm/s the time, 2.26dBm/s the left and right is average risk, establishing again the node maximum speed of motion is 20m/s, the signal intensity maximum is 3.47dBm/s.

Each index choosing unit is respectively: packet loss is taked percentage, dimensionless unit; Transmission speed adopts KB/s; Signal strength signal intensity adopts 10 ^-7MW, change rate signal adopts dBm/s.The white function for each index is:

1) the 1st grey class (upper grey class):

f ₁₁(0，c ₁₁)=f ₁₁(0，0.1)，

f ₂₁(c ₂₁，∞)=f ₂₁(688，∞)，

f ₃₁(c ₃₁，∞)=f ₃₁(3.36,∞)，

f ₄₁(0，c _t1)=f ₄₁(0，2.12)。

2) the 2nd grey class (middle grey class):

f ₁₂(-，c ₁₂，+)=f ₁₂(-，0.3，+)，

f ₂₂(-，c ₂₂，+)=f ₂₂(-，250,+)，

f ₃₂(-，c ₃₂，+)=f ₃₂(-，0.953,+)，

f ₄₂(-，c _t2，+)=f ₄₂(-，2.26，+)。

3) the 3rd grey class (lower grey class):

f ₁₃(c ₁₃，∞)=f ₁₃(0.7，∞)，

f ₂₃(0，c ₂₃)=f ₁₃(0，63)，

f ₃₃(0，c ₃₃)=f ₃₃(0，0.284)，

f ₄₃(0，c ₄₃)=f ₄₃(2.39,∞)。

Above-mentioned white function independent variable has different units for different attribute, needs to do unitized the processing before cluster analysis.The formulation of these indexs is simplification metrics-thresholds that propose for specific transmission requirement, can adopt corresponding index parameter for concrete mobile ad hoc network in reality.

(4) grey Cluster Evaluation and the decision-making of flavor cluster.

If through three flavor clusters are arranged after selecting, be respectively for the sample matrix of four evaluated nodes

${\mathrm{<figure-callout\; id="1"\; label="D"\; filenames="HDA00002824251700011.png"\; state="\{\{state\}\}">D</figure-callout>}}_1=\left[\begin{array}{cccc}0.23& 270& 1.10& 2.21\\ 0.81& 060& 0.30& 2.51\\ 0.12& 299& 1.46& 3.32\\ 0.35& 298& 3.41& 2.14\end{array}\right]$

${\mathrm{<figure-callout\; id="2"\; label="D"\; filenames="HDA00002824251700011.png"\; state="\{\{state\}\}">D</figure-callout>}}_2=\left[\begin{array}{cccc}0.31& 230& 1.01& 2.29\\ 0.75& 060& 0.26& 2.50\\ 0.07& 670& 0.26& 2.50\\ 0.12& 598& 3.34& 1.18\end{array}\right]$

$D_3=\left[\begin{array}{cccc}0.13& 657& 5.12& 1.09\\ 0.89& 030& 0.16& 2.30\\ 0.31& 272& 1.17& 2.25\\ 0.28& 240& 0.89& 2.29\end{array}\right]$

First to sample D ₁Carry out grey clustering analysis.

1. polarity is unified.

Consider that in sample attribute, packet loss is opposite with other attribute polarity, first carry out the polarity unified operation.Adopting the cost residual method is maximum polarity with the minimum reversal of packet loss and change in signal strength rate.

${{\mathrm{\&xi;}}_{i1}}^{*}=1-{\mathrm{\&xi;}}_{i1},\left\{\begin{array}{c}{{\mathrm{\&xi;}}_{11}}^{*}=1-{\mathrm{\&xi;}}_{11}=1-0.23=0.77,\\ {{\mathrm{\&xi;}}_{21}}^{*}=1-{\mathrm{\&xi;}}_{21}=1-0.81=0.19,\\ {{\mathrm{\&xi;}}_{31}}^{*}=1-{\mathrm{\&xi;}}_{31}=1-0.12=0.88,\\ {{\mathrm{\&xi;}}_{41}}^{*}=1-{\mathrm{\&xi;}}_{41}=1-0.35=0.65,\end{array}\right.$

${{\mathrm{\&xi;}}_{i4}}^{*}=3.74-{\mathrm{\&xi;}}_{i4},\left\{\begin{array}{c}{{\mathrm{\&xi;}}_{14}}^{*}=3.47-{\mathrm{\&xi;}}_{14}=1.26,\\ {{\mathrm{\&xi;}}_{24}}^{*}=3.47-{\mathrm{\&xi;}}_{24}=0.96,\\ {{\mathrm{\&xi;}}_{34}}^{*}=3.47-{\mathrm{\&xi;}}_{34}=1.5,\\ {{\mathrm{\&xi;}}_{44}}^{*}=3.47-{\mathrm{\&xi;}}_{44}=1.33;\end{array}\right.$

Polarity sample matrix after reunification is

${\mathrm{<figure-callout\; id="1"\; label="D"\; filenames="HDA00002824251700011.png"\; state="\{\{state\}\}">D</figure-callout>}}_1^0=\left[\begin{array}{cccc}0.77& 270& 1.10& 1.26\\ 0.19& 299& 0.30& 0.96\\ 0.88& 299& 1.46& 0.15\\ 0.65& 298& 3.41& 1.33\end{array}\right]$

Polarity after reunification, for packet loss and change rate signal upper, in and lower grey class white function be transformed to

$\left\{\begin{array}{c}{f}_{11}(c_{11},\&infin;)=f_{11}(0.9,\&infin;),\\ f_{12}(-,c_{12},+)=f_{12}(-,0.7,+),\\ f_{13}(0,c_{13})=f_{13}\left(\mathrm{0,0.9}\right).\end{array}\right.$

$\left\{\begin{array}{c}{f}_{11}(c_{41},\&infin;)=f_{41}(1.35,\&infin;),\\ f_{42}(-,c_{42},\&infin;)=f_{41}(-,1.21,+),\\ f_{43}(0,c_{43})=f_{43}\left(\mathrm{0,1.08}\right).\end{array}\right.$

2. sample matrix initialization.

Be unified white function, sample matrix pressed just value of row (namely by project):

${\mathrm{<figure-callout\; id="1"\; label="D"\; filenames="HDA00002824251700011.png"\; state="\{\{state\}\}">D</figure-callout>}}_1^{*}=\left\{\begin{array}{cccc}{\mathrm{\&xi;}}_{11}/{\mathrm{\&xi;}}_{11}& {\mathrm{\&xi;}}_{12}/{\mathrm{\&xi;}}_{12}& {\mathrm{\&xi;}}_{13}/{\mathrm{\&xi;}}_{13}& {\mathrm{\&xi;}}_{14}/{\mathrm{\&xi;}}_{14},\\ {\mathrm{\&xi;}}_{21}/{\mathrm{\&xi;}}_{11}& {\mathrm{\&xi;}}_{22}/{\mathrm{\&xi;}}_{12}& {\mathrm{\&xi;}}_{23}/{\mathrm{\&xi;}}_{13}& {\mathrm{\&xi;}}_{24}/{\mathrm{\&xi;}}_{14},\\ {\mathrm{\&xi;}}_{31}/{\mathrm{\&xi;}}_{11}& {\mathrm{\&xi;}}_{32}/{\mathrm{\&xi;}}_{12}& {\mathrm{\&xi;}}_{33}/{\mathrm{\&xi;}}_{13}& {\mathrm{\&xi;}}_{34}/{\mathrm{\&xi;}}_{14,}\\ {\mathrm{\&xi;}}_{41}/{\mathrm{\&xi;}}_{11}& {\mathrm{\&xi;}}_{42}/{\mathrm{\&xi;}}_{12}& {\mathrm{\&xi;}}_{43}/{\mathrm{\&xi;}}_{13}& {\mathrm{\&xi;}}_{44}/{\mathrm{\&xi;}}_{14}\end{array}\right\}$

$=\left\{\begin{array}{cccc}1.000& 1.000& 1.000& 1.000\\ 0.247& 0.222& 0.273& 0.762\\ 1.143& 1.107& 1.327& 0.119\\ 0.844& 1.104& 3.100& 1.556\end{array}\right\}$

3. white function is unified.

Value matrix first according to above-mentioned sample, the white function unification is

The 1st grey class (upper grey class):

$\left\{\begin{array}{c}{f}_{11}^{*}(c_{11},\&infin;)=f_{11}^{*}(1.169,\&infin;),\\ f_{21}^{*}(c_{21},\&infin;)=f_{21}^{*}(2.548,\&infin;),\\ f_{31}^{*}(c_{31},\&infin;)=f_{31}^{*}(3.055,\&infin;),\\ f_{41}^{*}(c_{41},\&infin;)=f_{41}^{*}(1.071,\&infin;);\end{array}\right.$

The 2nd grey class (middle grey class):

The 3rd grey class (lower ash $\left\{\begin{array}{c}{f}_{12}^{*}(-,c_{12},+)=f_{12}^{*}(-,0.909,+),\\ f_{22}^{*}(-,c_{22},+)=f_{22}^{*}(-,0.926,+),\\ f_{32}^{*}(-,c_{32},+)=f_{32}^{*}(-,0.866,+),\\ f_{42}^{*}(-.c_{42},+)=f_{42}^{*}(-,0.960,+);\end{array}\right.$ Class):

$\left\{\begin{array}{c}{f}_{13}^{*}(0,c_{13})=f_{13}^{*}\left(\mathrm{0,0.390}\right),\\ f_{23}^{*}(0,c_{23})=f_{23}^{*}\left(\mathrm{0,0.233}\right),\\ f_{33}^{*}(0,c_{33})=f_{33}^{*}\left(\mathrm{0,0.258}\right),\\ f_{43}^{*}(0,c_{43})=f_{43}^{*}\left(\mathrm{0,0.857}\right);\end{array}\right.$

4. calculate the grey cluster weight.

According to grey clustering analysis computational process, grey cluster weight coefficient such as table 2.

Table 2 grey cluster weight coefficient table

According to grey cluster weight formula:

To node 1(i=1) grey class 1(k=1), have

${\mathrm{\&sigma;}}_{\mathrm{ik}}={\mathrm{\&sigma;}}_{11}=\underset{j=1}{\overset{e}{\mathrm{\&Sigma;}}}{f}_{\mathrm{jk}}\left({\mathrm{\&xi;}}_{\mathrm{ij}}\right){\mathrm{\&eta;}}_{\mathrm{jk}}$

$=f_{11}({\mathrm{\&xi;}}_{11}{\mathrm{\&eta;}}_{11}+f_{21}\left({\mathrm{\&xi;}}_{12}\right){\mathrm{\&eta;}}_{21}+f_{31}({\mathrm{\&xi;}}_{13}{\mathrm{\&eta;}}_{31}+f_{41}({\mathrm{\&xi;}}_{14}){\mathrm{\&eta;}}_{41}$

$=f_{11}\left(1\right)\&CenterDot;0.149+f_{21}\left(1\right)\&CenterDot;0.325+f_{31}\left(1\right)\&CenterDot;0.389+f_{41}\left(1\right)\&CenterDot;0.137$

$=0.510$

Similarly, obtain σ ₁₂=0.908, σ ₁₃=0.411.

In like manner, after obtaining other each grey cluster weight, form table 3.

Table 3 grey cluster weight table

5. grey clustering analysis.

For node 1, grey cluster sequence σ ₁:

The grey class 3 of the ash grey class 2 of class 1

${\mathrm{\&sigma;}}_1=\left(\begin{array}{ccc}{\mathrm{\&sigma;}}_{11},& {\mathrm{\&sigma;}}_{12},& {\mathrm{\&sigma;}}_{13}\\ 0.510& 0.908& 0.411\end{array}\right)$

σ _1k*＝max(σ ₁₁，σ ₁₂，σ ₁₃)

＝max(0.5100.9080.411)

=0.908=σ ₁₂

k ^*=2.

Show that node 1 belongs to grey class 2, i.e. " general cooperation degree node "; In like manner, learn that node 2 belongs to grey class 3, belong to " low cooperation degree node "; Node 3 belongs to grey class 2, belongs to " general cooperation degree node "; Node 4 belongs to " high cooperation degree node ", obtains assessment result collection G _1*={ g ₂, g ₃, g ₂, g ₁.

6. the cluster decision-making of distinguishing the flavor of.

According to the grey clustering analysis process, by sample matrix D ₂, D ₃, in like manner, get each assessment result collection: G ₂ ^*={ g ₂, g ₃, g ₁, g ₁, and G ₃ ^*={ g ₁, g ₃, g ₂, g ₂, form thus flavor cluster evaluation decision matrix, as follows:

$G^{*}=\left\{\begin{array}{cccc}{g}_2& g_3& g_2& g_1\\ g_2& g_3& g_1& g_1\\ g_1& g_3& g_2& {\mathrm{<figure-callout\; id="2"\; label="g"\; filenames="HDA00002824251700011.png"\; state="\{\{state\}\}">g</figure-callout>}}_2\end{array}\right\}$

According to decision making algorithm, for node 1, C (g ₂)=2 are maximum, corresponding grey class g ₂Node 2, C (g ₃)=3 are maximum, corresponding grey class g ₃Node 3, C (g ₂)=2 are maximum, corresponding grey class g ₂Node 4, C (g ₁)=2 are maximum, corresponding grey class g ₁So the last result of decision is G ^*={ g ₂, g ₃, g ₂, g ₁.

From above computational process and result as can be known, the present invention is directed to the characteristics of " the sample information source is deficient; INFORMATION OF INCOMPLETE " under the mobile ad hoc network environment, adopt grey clustering algorithm to assess, and use the flavor swarm algorithm to suppress malice evaluation, make assessment result have discreteness and discrimination preferably, algorithmic procedure is simply effective, is easy to Project Realization.

Claims (1)

1. node cooperation degree appraisal procedure towards mobile ad hoc network is characterized in that comprising the following steps:

(1) to the meshed network behavior modeling, the network behavior of portraying node is divided into four class index sets: network service characteristic, wireless signal characteristic, mobility and spatial character;

(2) acquisition of information is sampled according to index set by node; Node is intercepted this neighbors and whether is forwarded this packet sending to the packet of neighbors to be placed in buffering area; If forwarded this packet at the in setting time neighbors, and be complementary with packet in buffering area, proof successfully forwards, the packet in the buffer release district; If do not mate, illustrate that neighbors revised this packet, abnormal behaviour is arranged; If interior nodes does not all listen to neighbors and forwards this packet at the appointed time, wrong generation is described, or being that this node is uncooperative give not to forward, or is that channel confliction is arranged, two kinds of reasons all be summed up as forward unsuccessful; All obtain by MAC layer or physical layer as for channel problems such as the signal strength signal intensity of neighbors, network congestion, communication contention awares;

After the acquired information sample, initial evaluation calculates scoring by the method for probability statistics, the number of times that route data in a period of time is forwarded success and failure is added up, and calculating successful probability by formula is p, makes initial evaluation with this data retransmission index to neighbors;

After as node A, its neighbours being collected Neighbor (A) and making initial evaluation, combine by Event triggered or with periodic refreshing, broadcast in two hop distances, can obtain with the neighbours that guarantee all nodes in Neighbor (A);

After node A receives the evaluation of other nodes to neighbors B, carry out preliminary treatment by the flavor swarm algorithm, reduce malice and estimate the impact that degree of belief is calculated; Selected core loop evaluation is as reference, common evaluation will compare with the core evaluation, exceed certain limit if depart from, and think that this evaluation is invalid, when the invalid evaluation of this entity node surpasses certain threshold value, will assert that this entity node is malicious node; Otherwise, if estimate effectively, will accept evaluation, if this evaluation simultaneously near the core evaluation, and when reaching certain number of times, the evaluation weight with this entity node will increase so;

(3) assessment initiate assessment sample that node will receive distinguish the flavor of assemblage classification with choose, right cluster data are carried out grey clustering analysis, draw final conclusion by the decision-making of flavor cluster at last;

1. node T ₀The broadcasting of initiation assessment request is with property set A={a _h| h=1,2 ..., e} tells each node, communication details is determined by assessment communication protocol;

2. each assessment node carries out sample to issue T0 after safety verification, is selected by the T0 cluster of distinguishing the flavor of, and cluster T obtains distinguishing the flavor of ₁, T ₂..., T _q

3. to each flavor cluster T _jThe sample matrix that obtains carries out the conversion of index polarity; Be the multiattribute index polarity disunity problem in the reply mobile ad hoc network, adopt the cost residual method to carry out the polarity conversion; Cost residual method processing procedure is as follows: establish certain attribute a _i-ξ _i, and 0≤a _i≤ max (a _i), this attribute is the smaller the better to the side of assessment, is minimum polarity, makes a _i ^*=max (a _i)-ξ _i, a _i ^*To the side of assessment, be changed to and be the bigger the better, be maximum polarity;

4. attribute quantification unit is unified processes, and adopts just value method of sample; Processing procedure is as follows: be provided with node d ₁Attribute sample D ₁={ ξ ₁₁, ξ ₁₂, ξ ₁₃And d ₂Attribute sample D ₂={ ξ ₂₁, ξ ₂₂, ξ ₂₃, appoint and get node d ₁, make that its sample is D ₂ ^*={ ξ ₁₁/ ξ ₁₁, ξ ₁₂/ ξ ₁₂, ξ ₁₃/ ξ ₁₃,={ 1,1,1}, D ₂ ^*={ ξ ₂₁/ ξ ₁₁, ξ ₂₂/ ξ ₁₂, ξ ₂₃/ ξ ₁₃,, cancellation all units; After just value of sample, corresponding white function also carries out just value according to the node of choosing;

In conjunction with the communication requirement of mobile ad hoc network, define the white function of each attribute, then T ₀For each flavor cluster T _jThe sample matrix that obtains carries out grey Cluster Evaluation, obtains evaluated node d _i(d _i∈ D, the grey class g of 1≤i≤m) _ji ^*

Images