Method of stabilising communication networks in conditions of disruptive external effects

FIELD: physics; communications.

SUBSTANCE: invention relates to inspection technology and can be used in telecommunication systems. Values of disruptive effects on a communication line are monitored, while simultaneously evaluating transmission capacity of each type of communication line. The obtained values are scaled relative maximum values for each class of parametres. The given values are used to train artificial neural networks with radial basic elements for approximating dependency of efficiency of each type of communication line on values of disruptive effects. Matrices of synaptic weights of trained neural networks are filled and further installed in accordance with a specific designed communication network for evaluating transmission capacity from forecast values of disruptive effects, obtained with time delay. Based on the forecast values of transmission capacity for each communication line, the available network resource is allocated between subscribers taking into account their priority categories.

EFFECT: wider functional capabilities, lying in timely rendering of information services to subscribers of different categories with external disruptive effects acting on a communication network.

11 dwg

 

The invention relates to measuring technique and allows you to:

1) to carry out constant monitoring of the electromagnetic environment (EME)that affect the performance of the distributed communication network;

2) to determine the form of the dependence of the throughput (productivity) of the communication network from the values of the electric field strength, train control system a distributed communication network, and control the current value of the electric field intensity to predict changes in network bandwidth over time;

3) the results predict the throughput of a distributed communication network to implement sustainability through selective (including categories of users of the communication network) user switching for redundant network connection.

There is a method, implemented in the analyzer parametric faults and failures (RF Patent No. 2270470 Analyzer parametric faults and failures", published 20.02.2006, bull. No. 5), which identifies the refusal of technical devices according to the specified number of failures for a selected time interval and containing an assessment of the duration of the stay option in the field of health when hit slice of the reference pulse corresponding to the maximum over the spine of the output parameter for border health, as well as comparing the estimates obtained with the minimum execution time of the task. At a lower value compared to the minimum execution time of the task, the time of the initial crossing the boundary parameter is recorded as a refusal, and the length of the considered reference pulse is included in the total duration of the failure. When crossing the boundary parameter in the direction of the field of health and achieving the minimum duration time of the task analyzer, you decide to return parameter in a workable state, at the same time is reset driver reference pulse. When the option is in the area of failure in a shorter period of time, the analyzer records the fact of occurrence and the duration of the failure. When the number of failures of the threshold, the analyzer records the fact of failure.

The disadvantage of this method is limited in use, which consists in not taking into account the destructive external influences on technical devices.

The terms "failure" and "failure" understand:

1) failure is an event consisting in the violation of the operating state of the object (GOST 27.002-89. Reliability in engineering. The basic concepts. Terms and definitions. M.: Publishing house of standards, 1990. - 36 S.);

2) fail - samo is cranaodes failure or acute failure, disposable little operator intervention (GOST 27.002-89. Reliability in engineering. The basic concepts. Terms and definitions. M.: Publishing house of standards, 1990. - 36 S.).

There is a method implemented in the device active protection and ensure technical readiness elements distributed local area network in the face of external destructive influences (Patent RF №2316810 "Device active protection and ensure technical readiness elements distributed local area network in the face of external destructive influences", published 10.02.2008 g)removing the above limitations, including the determination of the values of electric field intensity Emin- the minimum value of electric field intensity at which the faults and failures in distributed computer LAN does not occur, and Emax- the maximum value of electric field intensity at which the faults and failures of computers and LAN occur, the definition of Ekr- the threshold of electric field intensity at which the increase of electric field strength to a value of Emaxoccurs for tupreit- lead time - the time at which it is still possible to switch user from the affected computer in a distributed LAN on PC, not affected. ISM is the expansion of the electric field intensity at the point where the computer is continuously in addition, assessment of technical condition of the computer and projected time during which the computer will be in a healthy state. When the threshold value of electric field intensity Ekrat the point where the computer is the alarm indication on the probability of failure of computers and the minimum time maintaining the operability of the computer. This method is chosen as a prototype.

The disadvantage of this method is limited in application, expressed in consideration of only the destructive effects leading to loss of efficiency of elements of a distributed network (failure), while in the method are not taken into account those impacts that lead to a decline in the functioning of a distributed network while maintaining her health failed and / or denial of service subscribers).

Characteristic performance is a quantitative assessment of the quality of functioning of the system in a certain state after the execution of this task (the Method of selection of indicators for estimating the reliability of complex technical systems. M.: Publishing house of standards, 1977. - 43 S.).

The decline in functioning means reducing the level of quality functioning of the system due to partial or complete failure of the system (the Method of selection of indicators to assess the dinasti complex technical systems. M.: Publishing house of standards, 1977. - 43 S.).

Partial system failure is a failure of a system component, after which the system remains capable of performing its intended function (the Method of selection of indicators for estimating the reliability of complex technical systems. M.: Publishing house of standards, 1977. - 43 S.).

In addition, the disadvantage of the prototype is the lack of priority subscribers while providing active protection and technical readiness of the communication network that, in turn, affect the procedure of the allocation (distribution) of available reserves to maintain the desired stability of the functioning of a distributed communication network.

The technical result of the present invention is to enhance the functionality of the method, expressed in ensuring the timely provision of information services to subscribers of different categories of distributed communication network when exposed to an external destructive influences.

The technical result is achieved by controlling the values of the destructive effects on the lines of a telecommunication network, simultaneously evaluate the performance of each type of line, scale throughput at its maximum value for the communication network, and the values of a destructive effect on max the e value destructive impact, in which the performance of the high-speed communication lines in the network connection takes a value of zero, scalable values exercise training artificial neural networks (NS) with radial basis elements for the approximation of the dependency of the performance of each type of line values from the destructive influences of the matrix of synaptic weights of the trained neural networks are maintained, in accordance with the characteristics of a particular network connection, the network administrator performs communication installation neural networks with radial basis elements by the number included in the communication network of communication lines, control the values of the external destructive influences that will scale the data values, using a focused network of direct distribution with time delay predict the values of the destructive influences that these values served on installed neural network with radial basis elements, the outputs of which get projected values of the bandwidth for each communication line, based on these values, the administrator of the communication network performs the allocation of the network resource between users based on their categories of priority.

The forecasting performance of the system is the definition of a quantitative assessment of the productivity and system in a certain state for the next time interval.

The synaptic weight is a quantitative characteristic of the communication between neurons in the artificial neural network.

Installation of artificial neural network - construction of artificial neural networks for matrix values of synaptic weights obtained during the preliminary training of the artificial neural network.

The claimed method is illustrated by drawings on which is shown:

figure 1 - structure of the focused network of direct distribution with time delay (TLFN);

figure 2 is a block diagram illustrating the method resilience of communication networks in terms of external destructive influences;

figure 3 - dependence of the scaled values of the bandwidth of the communication line from the scaled values of the electric field strength for different ratios of protection of lines of communication from external destructive influences;

figure 4 - summary of data sets for training artificial neural networks with radial basis elements and check their operation on the data sets that are not participating in the learning process;

figure 5 - continued summary table of data sets for training artificial neural networks based on radial basis elements and check their operation on the data sets that are not participating in the learning process;

figures the 6 - an example of a study and survey of neural network model based on radial basis elements in MATLAB 7;

figure 7 - example of training and survey model focused network of direct distribution with time delay to the schedule of forecast errors in the MATLAB 7;

figure 8 - real-time graphics and projected growth of the scaled values of the electric field strength of the EMP;

figure 9 - forecasting model focused network of direct propagation delay time values of the electric field strength;

figure 10 - prediction of bandwidth models of neural networks based on radial basis elements;

figure 11 is a graphical display of the adequacy of the results obtained by the neural networks.

The availability of the communication channel is a measure of long-term use of the communication channel, formulated on an average annual basis, and shows the percentage of time during which the communication channel may be closed. At the closing of the channel is defined as a positive value of energy stores in decibels that meet existing requirements for the reliability of transmission. In digital communication reliability of the transmission depends on measuresnormalized signal-to-noise pickup is nick, defined in the formula as follows (Sklar, B. Digital communications. Theoretical basis and practical application. 2-nd edition.: TRANS. from English. - M.: Publishing house "Williams", 2003. - 1104 S.: ill.):

where S is the signal power;

N is the noise power;

W is the bandwidth of the signal;

R is the bit rate information.

From a consumer's perspective, the most significant variable is the bit rate information, so this value determines not only the appearance of digital services provided by the communication network, but the ability to unite a diverse group of messages into a single digital data flow (speeds). It is known that the transmission of telephony via digital communication channels requires transfer speed uncompressed 64 kbps, and the application of compression 8-32 kbit/s (Ubaidullaev P.P. Fiber-optic network. ): ECO-TRENDS, 2000. - 267 C.). For teleconferencing, the desired speed increases from 48-64 kbit/s when compressed to 128 kbps for transmission without compression. Implementation services Videophone in the transfer format MPEG-4 (H.320) without compression will require a transfer rate 3.04 from Mbit/s With the increase of services offered by digital data networks, and increase requirements to the speed of transmission implemented in digital networks.

In turn, consumers of communication services under adelayda in importance for them continuity of services provided to subscribers of different categories. So, for subscribers of the first category network provides absolute priority for services of communication compared with subscribers of the lower orders. Moreover, as increasing numbers of categories of the subscriber's priority is reduced. In case of simultaneous requests for service primarily serviced by the network subscribers with the highest priority, and the rest of the subscribers are served in order of decreasing priority to the full utilization of network resource.

There are ways to "suppress" computers, communications on the part of the attacker, in which computers, communications technical parameters are considered healthy, but its main purpose they cannot deliver or perform with some restrictions. One of such methods is the introduction of computers for communications and information of organisms. The information organism refers to a relatively short set of commands (up to 30 teams), which is transmitted to the management computer, which results in the capture of information and computing resources of the automated control systems (Grinev S.N. Intellectual opposition informational weapons. A series of "Informatization of Russia on the threshold of XXI century". - M.: SINTEG, 1999. - 232 S.). The introduction of this type of viruses in the computer may lead to the restriction on the volume of available RAM space on magnetic media, to limit processor time, which will directly affect the performance of the telecommunications network while maintaining the functional elements. Another way undesirable impact on the telecommunication network from a malicious user is creating them with the help of technical means of electromagnetic interference on the cable leading to the establishment of lines of communication frames-ghosts (Lipatnikov VA, Malyutin, VA, Starodubtsev SCI Information security of telecommunication systems. - SPb.: MAS, 2002. - 476 S.: ill.). This type of frames is perceived network adapters as frames, not having the normal start of the frame. A staff ghosts in the information flow greatly reduces the payload segment of the communication network, which affects the attacker. Consumer information services less interested in the health of the network as such, moreover, the scope of his interests is wholly focused on guaranteed performance and the timely provision of necessary services from a telecommunication network.

The implementation of the destructive effects on components of the telecommunications network in order to display its elements in an inoperable condition will require the attacker pain is their cost, both technical and energy. The most acceptable way destructive impact of the attacker, from the point of view of economic efficiency, is the impact on the telecommunications network with the aim of reducing its performance while maintaining health. Empirically it is possible to determine the degree of influence of external destructive influences on the performance of the communication network. While it is safe to say that with increasing values of the destructive impacts the performance of any communication lines will be reduced. The influence of destructive impacts on the performance of one or another line of communication will depend on the kind of communication which is based on this communication line. Depending on the design specifics of channels and systems of communication and used resources (protection of wave propagation, frequency range) communication line can be classified according to the following genera communication (GOST RV 52216-2004. The relationship of the military. Terms and definitions. M.: Izd-vo standards. 2004. - 11 S.):

1) the satellite communication line;

2) fiber-optic communication lines;

3) line tropospheric communication;

4) line wired;

5) radio communication lines;

6) radio relay communication lines.

When the same values of the destructive effects on one direction of the communication network among the communication line, the OST is indeed working in the technical aspect, will significantly reduce their performance, while others will keep their performance at the required level. In this case, the network Manager to ensure timeliness of services provided to subscribers based on their priority, you must have predictive information about the performance of a communication network as in all areas, and within them lines.

As a tool for carrying out an impact assessment of the values of the destructive effects on network performance, and the prediction changing the values of the destructive impacts for the lead time, the device will be used in artificial neural networks.

The possibility of using artificial neural networks for solving these problems is based on the fact that a neural network consisting of two layers and having a hidden layer of an arbitrary large number of nodes can approximate any function of real numbers with a given degree of accuracy (Proceedings of the V.I., Smirnov D.A. Neural networks and their applications in control systems and communications. - M.: Hot line - Telecom, 2002. - 94 C.).

Imagine a simplified model of the influence of the electric field intensity on the performance of the communication line of the first kind with variable names:

E - the electric field at a particular point in space;

Ri- roizvoditelnost communication line of the i-th kind.

Between E and Rithere are causal relations expressed by functions of the form:

where fi- functionality for the communication line of the first kind;

the value (-1) means decreasing the nature of changes in the performance of the i-th kind of communication lines with increasing electric field intensity.

In terms of artificial neural networks the problem of estimating an unknown function

fi(E)-1is the problem of approximation of functions. The learning sample is a pair of data inputs-outputs of the neural network obtained in single moments in timeAs a result of training the neural network on the next set of data required functionality will be represented by a matrix of synaptic weights of the artificial neural networkwhich can later be used to determine the performance values of the communication line of the i-th kind, controlled by the values of electric field intensity. As a neural network that implements this task, the selected artificial neural network with radial basis elements RBF (Radial Basis Function Network). The RBF network has an undeniable advantage over multi-layered networks of direct distribution, such as the ability to simulate arbitrary nonlinear functions using is this one of the intermediate layer, what saves us from having to decide on the number of layers (Callan R. Basic concepts of neural networks: TRANS. from English. - M.: Publishing house "Williams", 2003. - 288 S.: ill.). Each value of the training sample for the correct functioning of the RBF network, you must first scale. One method of scaling is the division of the characteristic value by the maximum value of this characteristic, resulting in the maximum value will be limited to a value of 1. The physical nature of the values used for training and further use of RBF networks, which are the values of the electric field strength E and the throughput of connection cables Rishows that there is no negative values, it is also possible to determine the maximum values for each parameter. So, for each line of one kind of communication bandwidth is defined on the intervalThe magnitude of the electric field strength maximum limit value for the communication line of the first kind defined as the value at which the bandwidth of the communication line of the i-th kind of takes a value of zero, and further possible to increase the strength of the electric field will not affect the ability of p is dostavlenija services communication and will be destructive impact on the technical elements of the communication line. Thus, for each kind of line of communication with regard to its technological aspects is determined by the maximum bandwidthas well as empirically determined maximum limit value of the electric field strength. Using the data values are scaled current value pairsfor each i-th kind of line to the scaled valuesfor k∈[1...N] expressions:

where- the current value of the electric field strength for the communication line of the first kind;

- maximum limit value of the electric field strength for the communication line of the first kind;

and

where- the current value of the bandwidth of the i-th kind;

- maximum bandwidth of the communication line of the i-th kind.

The only requirement for the training of RBF networks is the provision of the required number of training implementations to achieve the required accuracy and reliability of neural network training. When the unknown is the a priori probability of quantities is required implementations of N is determined by the expression (Ivanov E.V. Simulation tools and communication systems and automation. SPb.: YOU, 1992. - 206 S.):

where tα- valued argument function Laplace;

ε is the desired accuracy of the estimates.

To ensure the accuracy of the estimate ε=10-2and α=0.99, and which corresponds to the tabular value of tα=2,53, the required number of training implementation N for RBF neural network will be the value 16002, as for the values of the electric field strength and throughput of the communication line.

The problem of forecasting also successfully solved in the context of artificial neural networks. What training sample set of discrete samples, pre-scaled in accordance with the expression (3) values of the electric field strength {E*(t), E*(t-1), ..., E*(t-k)} at the appropriate time, and the output value at some future time (t+1) is the scaled value of E*(t+1). As a prediction network is proposed to use a focused network of direct distribution with time delay (focused time lagged feedforward network - TLFN) (Haykin S. Neural networks: a complete course, 2nd edition: TRANS. from English. - M.: Publishing house "Williams", 2006. - 1104 S.: ill.). This network includes a memory based delay lines with features the AMI order k and multilayer perceptron.

Multilayer perceptron multilayer neural network feedforward input with one input layer sensor element, at least one hidden layer of computational neurons and one output layer of neurons (Haykin S. Neural networks: a complete course, 2nd edition: TRANS. from English. - M.: Publishing house "Williams", 2006. - 1104 S.: ill.).

The structure of this artificial neural network presented in figure 1. Shown in the drawing, the elements designated by Arabic numerals, are:

1 - receiving element input values;

2 - the first element of the delay line input signal values in one step of time;

3 - the second element of the delay line input signal values in one step of time;

4 - k-th element of the delay line input signal values in one step of time;

5 the first sensor element of the input layer of the neural network;

6 - the second sensor element of the input layer of the neural network;

7 - the third sensor element of the input layer of the neural network;

8 - (k-1)-th sensor element of the input layer of the neural network;

9 - k-th sensor element of the input layer of the neural network;

10 - the first neuron in the hidden layer of the neural network;

11 - the second neuron in the hidden layer of the neural network;

12 - the third neuron in the hidden layer of the neural network;

13 - the i-th neuron in the hidden layer of the neural network;

<> 14 - neuron output layer of the neural network.

Structural elements are interconnected as follows: the information output of the receiving element of the input value information is input to the first element of the delay line input signal values for one cycle of time and the first sensor element of the input layer of the neural network, the information output of the first element of the delay line input signal values to a single quantum information is input to the second element of the delay line input signal values for one cycle of time and the second sensor element of the input layer of the neural network, the information output of the second element of the delay line input signal values to a single quantum information is input to the third sensor element of the input layer of the neural network and the third element of the delay line input signal values for one clock cycle time information output (k-1)-th element of the delay line input signal values in one step of time information is input to the (k-1)th sensor element of the input layer of the neural network and the k-th element of the delay line input signal values in one step of time, the information output of the k-th element of the delay line input signal values in one step of time information is input to the k-th and the th element of the input layer of the neural network, information outputs from the first to the k-th sensory elements of the input layer of the neural network are informational inputs from the first to the i-th neuron in the hidden layer of the neural network, the information outputs from the first to the i-th neuron in the hidden layer of the neural network are informational inputs of the neuron of the output layer of the neural network, with the information of the output neuron of the output layer of the neural network remove the desired signal values.

Artificial neural network TLFN used as one-step prediction system, in which E*(t+1) contains the desired response to an input signal consisting of a set of

For training, use the standard back-propagation algorithm. At time k the time the image is applied to the input layer of the network in the form of a vector scaled signal:

which can be viewed as describing the state of a nonlinear filter at time t. One epoch of training consists of a sequence of States, the number of which is determined by the order of memory k and the cardinality of the training examples n the order of the memory is selected by the user, to the forecasted value at time t taking into account its worst values at time (t+1) allowed editing the Torah to adopt and implement a management decision on the distribution of requests from subscribers of the communication network based on their category and needs for bandwidth communication channel.

Scaling is controlled by the values of the external destructive influences in our case, values of electric field intensity Ei(t), is necessary to achieve consistency in working together previously trained neural networks with radial basis elements and focused network of direct distribution with time delay.

The proposed method, a block diagram is shown in figure 2, allows not only to evaluate the impact of the destructive effects on the performance of heterogeneous lines of communication within the communication network, but also to provide the network administrator with the necessary information about available spare capacity at the time of the forecast. This information enables us to provide subscribers of the communication network information needs based on their category and total. Unlike the prototype method the proposed method is primarily aimed at ensuring the main purpose of communication networks exposed to destructive factors (in our example, the factor is the electric field), so must be pre-actions of the prototype method.

Explain the implementation of the method in the following example. Let the given communications network comprising at least two communication nodes and one communication line, connect the processes on the nodes. For some time Δt on the network affected by the destructive impact in the form of electric field intensity E(Δt). In communication network subscribers are three categories. The subscriber first category require bandwidth to ensure the call, i.e. the data transmission rate is 128 kbps, the subscriber of the second category for telephony required transmission rate without compression - 64 kbps, and the subscriber of the third category data rate of 32 kbit/s is Required to determine whether the available resource of the communication network to provide the required bandwidth to the subscribers of the three categories in the predicted value of the electric field strength?

Total demand subscribers of all three categories will be 224 kbit/s, With a 30% reserve bandwidth requirement of the communication network bandwidth will be 288 kbit/s This speed transmission is provided with a minimum of destructive influences in our case, Emin=0 V/m at the impact on the network connection values of the electric field strength(this value was chosen to clarify the essence of the method, the actual value requires clarification in the course of research for each specific kind of communication line) the speed of the network connection will be reduced to h is left values. Based on these provisions, the scaling is controlled values for the expressions (3) and (4) will be respectively equal toandThe scaling parameters of intensity and bandwidth will result in different values of the dimensionless values on the interval [0...1].

Taking into account expression (2) we can assume that the dependence of the bandwidth of the communication line from the increasing strength of the electric field is in the nature of a decreasing function. Adequate analytical account of the dependence of the scaled throughput R*ifrom the scaled values of the electric field strength E*iis:

where η is the coefficient reflecting the degree of protection of lines of communication from the destructive influence of an electric field.

Values of the coefficient η is defined on the set of integers ≥0, with higher values of this ratio, the greater the security has a communication line, and its throughput is slower changes in accordance with the values of the electric field strength. Figure 3 graphically presents the dependence of the scaled throughputfrom the scaled values nab is azenosti electric field for different values of the coefficient of security η. A curve consisting of short line segments, corresponds to the expression (7) with the value of the coefficient of security η=6. A curve consisting of a long line segments corresponds to the expression (7) with the value of the coefficient of security η=3. The solid curve line corresponds to the expression (7) with the value of the coefficient of security η=2. A curve consisting of line segments and a single point corresponds to the expression (7) with the value of the coefficient of security η=0,7. A curve consisting of line segments and double points, corresponds to the expression (7) with the value of the coefficient of security η=0,5.

Training of artificial neural networks with radial basis elements RBF feasible for dependency ratios of protection equal to 2 and 0.5.

Simulation of artificial neural networks with radial basis elements feasible in MATLAB 7 using Neural Networks Toolbox (neural network), which contains tools for design, modeling, training, and use of a variety of well-known paradigms of artificial neural networks.

For training the neural network model with radial basis elements of RBF using expressions (7) and the values of the coefficients of the protection of communication lines from destruc the positive effects were obtained training data sets, presented in the table on the figures 4 and 5. Model of artificial neural network with radial basis elements includes one input, one hidden layer with ten neurons and one output layer with one neuron, the activation function of the output neuron is linear. The values of radial basis functions for each neural network, depending on the factor of protection of lines of communication are presented in the table on the figures 4 and 5. Example training and testing of neural networks with radial basis elements are presented in figure 6, where the trained neural network with radial basis elements are presented in the upper left of the workspace named values of a and b class network. In the right part of figure 6 presents the values of W and Q reactions of trained neural networks a and b on input a set of scaled values of the electric field strength not involved in the learning process. Test the scaled values of the electric field strength and the values of the reactions of neural networks with radial basis elements are presented in the table on the figures 4 and 5. Comparison of the responses of neural networks with radial basis elements with scaled analytical throughput R*shows that an approximate capacity of neural networks shortcuts is just high and provides the necessary accuracy, especially with increasing scaled values of electric field intensity.

The model focused network of direct distribution with time delay achievable in MATLAB 7 using Neural Networks Toolbox. To find the training sets of scaled values of the electric field strength will use the dependence presented in the prototype (Patent RF №2316810 "Device active protection and ensure technical readiness elements distributed local area network in the face of external destructive influences", published 10.02.2008,). The graphical dependence of the scaled values of electric field intensity with time are presented in figure 8 as a solid line. The volume of the set values is a matrix of memory in which are stored the detainees on one (or more) quantum scaled values of the electric field strength. These sets of values are given in the input focused network of direct distribution with time delay and in accordance with the set of output values representing a sequence of values for the next cycle for each trainee series, using the least squares method, we define the weights of the neural network. An example of learning, to forecast focused CE and forward propagation delay time, as well as the schedule of forecast errors are presented in figure 7. Small values of the forecast errors model focused network of direct propagation delay time enable you to use a neural network model for solving the problem of forecasting. A plot of the scaled values of the electric field strength from the time, built to forecast the values represented by the dashed line in figure 8.

The figure 9 presents an example of a forecasting model focused network of direct distribution with time delay the next value of electric field intensity on five previous values. These values are presented in column 18 of the table in the lower window of figure 9. These values using the trained neural network's predicted value w=0,2685 provided in the upper window of figure 9. In turn, for the prediction of bandwidth per measure of time corresponding to the obtained predicted value of electric field intensity, it is necessary to submit this value to a previously trained neural network with radial basis elements a and b. An example of forecasting bandwidth trained neural networks with radial basis elements according to the forecast value of the electric field strength presented at the Figo is e 10. With the resulting scaled predictive value of bandwidth for the communication line with the factor of protection equal to 2, amounted to the value of 0,9279, and lines of communication with the factor of protection equal to 0.5, amounted to the value of 0,4829. These values are adequate analytical dependences, shown in figure 11.

The network administrator based on projected throughput values for a given kind of communication lines can make the following conclusions:

for lines of communication with the factor of protection equal to 2, the subscribers of all three categories at the time of the forecast can be provided with communication services, with potential throughput of this type of line will be 43,245 kbps, which will provide additional service to a subscriber of a third category for the data transmission speed of 32 kbps;

for lines of communication with the factor of protection equal to 0.5, the predicted throughput will be 139,075 kbit/s, which, in turn, can satisfy only the subscriber of the first category, and for subscribers of a lower category to this point it is necessary to provide a resource for communication with the necessary reserve capacity.

Thus, the integrated use of artificial neural networks RBF on the number of delivery lines, implemented. is established in a communication network, and artificial neural network TLFN allows to provide the required timely provision of information services to subscribers of a distributed communication network when exposed to an external destructive influences.

The advantage of artificial neural networks for the solution of these problems lies in their speed up to 54 billion switches per second at present neuropath (Proceedings of the V.I., Smirnov D.A. Neural networks and their applications in control systems and communications. - M.: Hot line - Telecom, 2002. - 94 C.), which is three orders of magnitude higher than the rate of increase of the electric field EMP (unit of change in a nanosecond). Another useful property of neural networks is their ability to retrain new parametric dependences on the basis of the new trainees values.

The analysis of existing methods has allowed to establish that the analogues, identical signs of the claimed technical solution is available, which indicates compliance of the claimed method the condition of patentability "novelty". Introduced distinctive feature is the integrated use of artificial neural networks RBF on the number of delivery lines that are implemented in the communication network, and allows to determine the parametric dependence of the influence of the electric field intensity on productivity is alnost these lines with artificial neural network TLFN, engaged in forecasting the strength of the electric field by the time the lead time, the analogues are not found. Therefore, the claimed method meets the criterion of "inventive step".

Industrial applicability of the method is the presence of a variety of neuro-computers - from specialized integrated circuits, in which is entered a predefined structure of neural networks to universal programmable coprocessors to computing machines on which you can implement any model of neural networks (Proceedings of the V.I., Smirnov D.A. Neural networks and their applications in control systems and communications. - M.: Hot line - Telecom, 2002. - 94 C.).

The only way to ensure stability of communication networks in terms of external destructive influences, which determine the threshold value destructive effects, which can switch a user exposed to electronic computing machines (computers) as part of a distributed local area network on the electronic computer, is not affected by this effect, constantly measuring the electric field strength at the point of location of the computer, evaluate the technical condition of the computer and predict the time during which the computer will be in raboto posebna condition, when reaching the destructive impact threshold at the point where the computer is the alarm indication on the probability of failure of computers and the minimum time to keep the computer, wherein the control value destructive impacts on the lines of a telecommunication network, at the same time appreciate the value of the throughput (productivity) of each kind of line, scale throughput at its maximum value for the communication network, and the values of the destructive effects of scale to the maximum destructive impact, where the performance of the high-speed communication lines in the network connection takes a value of zero, the scaled values exercise training artificial neural networks with radial basis elements for the approximation of the dependency of the performance of each type of line values from the destructive influences of the matrix of synaptic weights of the trained neural networks retain for subsequent installations, in accordance with the characteristics of a particular network communication carried out with the installation of neural networks with radial basis elements by the number included in the communication network delivery lines, a counter is leraut value destructive influences, scale the data values, using a focused network of direct distribution with time delay predicted values of destructive influences, these values serves installed on neural network with radial basis elements, the outputs of which get projected values of the bandwidth for each communication line, based on the data values distribute network resource between users based on their categories of priority.



 

Same patents:

FIELD: physics, computer technology.

SUBSTANCE: invention deals with techniques employed for protection of data in the course of biometrical identification and authentification and is highly relevant for e-transaction conclusion, e-trade and Internet-banking. In case of neuron network storage of the user personal key and biometrical parameters encryption of the neuron network description tables is done with segments of the key located within the neuron network proper.

EFFECT: neuron network container enhanced resistance to brute force attacks due to the Alien's error expansion effect.

3 cl, 2 dwg

FIELD: physics; control.

SUBSTANCE: invention relates to control systems. To predict value of parameter of the product being manufactured by means of neuronal network, it is suggested to account for the product manufacturing history in order to determine initial value of input neuron of neuronal network.

EFFECT: procurement of product parameters as uniform and well-defined, as possible, during its manufacture.

11 cl, 3 dwg

FIELD: computer engineering, in particular, modular neuro-computer devices, possible use for performing an operation for modulus multiplication of two polynomials.

SUBSTANCE: in accordance to the invention the effect is achieved due to application of neuron network basis and conduction of operations in polynomial system of residue classes of expanded Galois field GF(2V). The device contains eight neurons in first layer, where the first four neurons of first layer are meant for receiving binary code of first operand, and the second four neurons of first layer are meant for receiving binary code of second operand, sixteen neutrons in second layer, which realize the AND operation, four neurons in third layer, which realize the modulo 2 addition operation.

EFFECT: simplification of device, reduction of hardware costs.

1 dwg, 3 tbl

FIELD: medicine.

SUBSTANCE: the present innovation refers to new mutations, mutation combinations or those of mutation profiles of the genes of reverse transcriptase HIV-1 and/or protease that correlate with phenotypic resistance to medicinal preparations against HIV. To be more concrete, the present innovation deals with applying genotypic characteristics of HIV target population and subsequent correlation of the present information at phenotypic interpretation to establish the correlation of viral mutation profiles at resistance to medicinal preparations. The innovation in question enables to apply the data obtained in screening medicinal preparations applied against HIV-infection.

EFFECT: higher efficiency.

12 cl, 2 dwg, 14 ex, 13 tbl

FIELD: computer engineering, possible use in modular neuro-computers for fast division of numbers, represented in remainder class system.

SUBSTANCE: in accordance to the invention, neuron network contains input layer, neuron networks of end ring for determining remainder of denominator and neuron networks of end ring for determining quotient of division of two numbers.

EFFECT: increased speed of division operation, less equipment needed.

1 dwg, 1 tbl, 3 ex

FIELD: computer engineering, in particular, modular neuro-computer means, possible use for finding and correcting errors in modular codes of polynomial residual class system.

SUBSTANCE: in accordance to invention, polynomial residual class system is used, in which as system base minimal polynomials pi(z), i=1,2,...,7, are used, determined in extended Galois fields GF(25) and neuron network technologies, and also modified zeroing constants determined in current polynomial residual class system are used in parallel.

EFFECT: increased speed of detection and correction of errors in modular codes of polynomial residual class system.

2 dwg, 7 tbl

FIELD: computer engineering, possible use for controlling and correcting error occurring during transfer of information and for performing arithmetic operations in expanded Galois fields GF (2v).

SUBSTANCE: device contains memory block, adder, error syndrome computation block represented by two-layered neuron network, containing neurons forming first layer and second layer of neurons.

EFFECT: increased speed of detection and correction of errors.

2 dwg, 6 tbl

FIELD: secret communications engineering, possible use for transferring, detecting and changing cryptographic keys for users in spatially distributed system.

SUBSTANCE: for realization of method to key distribution center biometric images of user are delivered once and recorded, source artificial neural network is generated and user key as well, aforementioned neural network is taught so, that during its input to biometric image of user it spawned user key at outputs, after that neural network taught in such a way is delivered to user object via open communication channels, where it is recorded and utilized for forming key by feeding to its input by user of his biometric image, and if it is necessary to change the key new key is generated in center, source artificial neural network is taught again on new key and transformed neural network is delivered along open communication channel to user object, where it is recorded and utilized for receiving new key.

EFFECT: possible transmission to user of cryptographic keys, transformed using biometric parameters of this user to neural network, via open communication channel.

2 cl

The invention relates to computing and can be used to build automated data banks, information storage and artificial intelligence systems

The invention relates to the field of computer engineering and can be used to ensure reliable operation of steam turbines

FIELD: physics; computer engineering.

SUBSTANCE: invention relates to management of security of Windows family operating systems (including Microsoft Windows 2000 Professional, Microsoft Windows 2000 Server, Microsoft Windows XP Professional, Microsoft Windows Server 2003, Microsoft Windows Server 2008, Microsoft Windows Vista Ultimate, Enterprise and Business 32-x and 64-x versions of any localisation), and specifically to comparison of configuration characteristics of operating systems from the view point of assessing their security. The result is achieved owing to possibility of comparing configuration parametres, monitoring the behaviour of changes in status and detection from a given "standard" status in Windows family operating systems after proposal of a procedure for comparing security status and switching to the analysed complete set of configuration parameters of security of user layer resources.

EFFECT: increased efficiency of assessing system security.

3 cl, 1 dwg

FIELD: physics; communications.

SUBSTANCE: invention relates to the technology of protecting digital content, and specifically to playing back digital content using licenses. A chain comprises an end license associated with content at one end, and a root license at the other end and all intermediate licenses in between. The end license and all intermediate licenses in the chain are attached to neighbouring licenses in the chain towards the root license, and the root license is attached to the private key owner (PR-U). Each license in the chain is verified and confirmation is made of whether the license allows content playback. A decryption key is obtained from the end license based on application of (PR-U) to the root license. The obtained key is used to decrypt the encrypted content, and the decrypted content is played back.

EFFECT: provision of playback of encrypted digital content on a computer in accordance with a license chain, on which a request for playing back encrypted content is received and a license chain corresponding to that content is found.

5 cl, 5 dwg

FIELD: physics; computer engineering.

SUBSTANCE: invention relates to systems and methods for coordinating software components. Version management policy, which is included in the target component, shows how to access the target component, for instance either as a library component or as a platform component. A component can be designated a library component when it creates a version which is compatible at the binary code level. When other components request for such a component, they receive exactly the component version which they requested. On the other hand, a component can be designated a platform component when it creates a version which is compatible at the bit code level. When other components request for such a component, they receive the last updated version of the requested component instead. That way, access to the corresponding component version is provided (even a version which is different from the requested version). Other implementation versions include mechanisms for stratification of the component application field, based on different data processing levels.

EFFECT: improved version management.

23 cl, 8 dwg

FIELD: physics; computer engineering.

SUBSTANCE: invention relates to computer security. In the system, according to the invention, a basic operating system is used together with a highly reliable operating system. The basic operating system is at least a certain part of infrastructure of the highly reliable system. Occlusion of elements of the graphical user interface, related to the highly reliable operating system, is prevented. Also part of the secret information, which upon command can be displayed by elements of the graphical user interface, related to the highly reliable system, is stored. Coordination of defined components of images of all elements of the graphical user interface, related to the highly reliable operating system, also enables identification of valid elements. In the system for managing windows of the basic operating system there is public heading information for window identification, belonging to the process operating under control of the highly reliable operating system. Information of the secret heading, related to same window, is used only in the highly reliable operating system.

EFFECT: invention increases security of computer systems from hacking.

30 cl, 9 dwg

FIELD: physics; computer engineering.

SUBSTANCE: invention relates to architecture and a method of permitting display of digital content with the corresponding digital license, associated with a specific computer device. Transmitting and receiving computer devices are connected to each other over a network. A transmitting device transmits protected digital content to a receiving device such that, the receiving device can access that content even if the content is directly licensed to the transmitting device and not to the receiving device.

EFFECT: coordinated access to content between computer devices on a network.

20 cl, 6 dwg

FIELD: information technologies.

SUBSTANCE: inventions are related to computer systems and methods for provision of protected access to database. System comprises memory device for protection descriptors, which store information about protection, related to at least one line of database, besides database contains at least one table that includes at least one line and two columns, in one of columns there is a protection descriptor stored, being related to line, information stored in protection descriptor comprises data about which type of access and to which principal is permitted or prohibited; database processor that issues response to query of database, based at least partially on information about protection stored in protection descriptor, which is assessed on the basis of information stored in database, and context of user that makes query; query component that contains optimiser of queries, which defines optimal route for response provision to query.

EFFECT: improved protection of access to database.

20 cl, 9 dwg, 2 tbl

FIELD: information technologies.

SUBSTANCE: there chosen is domain identifier and connection of at least one user (P1, P2, …, PN1), at least one device (D1, D2, …, DM) and at least one information element (C1, C2, …, CN2) to Authorised Domain (AD) specified with domain identifier (Domain_ID). By means of that there have been obtained many checked devices (D1, D2, …, DM) and many checked personalities (P1, P2, …, PN1), which is authorised for access to information element of the above Authorised Domain (100). Thus, access of user who controls the device to information element of authorised domain is obtained either by checking the fact that information element and user are connected to one and the same domain or by checking the fact that device and information element are connected to one and the same domain.

EFFECT: ensuring method and system for providing Authorised Domain structure based both on personalities and on devices.

12 cl, 6 dwg

FIELD: information technologies.

SUBSTANCE: checking method of certificate validity, which includes the key connected to network devices, involves the step of receiving the encoded content and validity index connected to that content in the network. Certificate validity is evaluated from the time index included in the certificate where the time index has the value corresponding to the certificate issuing date, and from validity index connected to the above encoded content.

EFFECT: simplifying the checking process of certificate validity, which provides access to data without reducing data access security.

20 cl, 12 dwg

FIELD: information technologies.

SUBSTANCE: method and device for determining authenticity of the system user is based on comparing coordinates of peculiar features of papillary patterns of fingers at double finger touch of the receiving scanner surface. During the first registration there obtained are pictures of at least two fingerprints, and during the second registration there obtained is the picture of at least one fingerprint, at that, the second registration is performed upon "request-answer" protocol command. Authenticity is considered confirmed in case of non-linear dependence of coordinate offsets of peculiar features of the first and the second pictures. Device for implementing the method consists of a scanner, picture processing unit, database, comparing unit, protocol forming unit connected to the scanner, and comparing unit. Protocol forming unit display panel is located on the scanner front surface.

EFFECT: ensuring high accuracy of authenticity and excluding the access of occasional persons to the protected system.

3 cl, 3 dwg

FIELD: information technologies.

SUBSTANCE: first initial value is known both to the keyboard and the component. Keyboard and component exchange time values. Both the keyboard and the component compute the second initial value and the third initial value on the basis of time values and the first initial value. Both the keyboard and the component make one and the same computation so that both the keyboard and the component have one and the same second and third initial values. The keyboard encodes keystrokes meant for the component by using CBC-3DES method on the basis of the key and the second initial component, as well as creates message authentication code for each keystroke by using CBC-3DESMAC on the basis of the key and the third initial value. The component encodes and verifies keystrokes by using the key and the second and the third initial values.

EFFECT: providing safety connection between two components, such as a keyboard or a related device, and software component via an unsafe communication channel.

26 cl, 6 dwg

FIELD: engineering of devices and methods for using server for access to processing server, which performs given processing.

SUBSTANCE: for this in accordance to method reservation is requested, reservation is confirmed, authentication information included in reservation information is stored, service is requested on basis of authentication information, server utilization is authenticated and server is utilized on basis of authentication result, while on stage of reservation confirmation device for controlling reservation transfers reservation setting information, and on stage of authentication server utilization is only confirmed when authentication information matches authentication information transferred from user terminal. Device contains receiving means, information generation device and transmitting means.

EFFECT: creation of method for using server, device for controlling server reservation and means for storing a program, capable of providing multiple users with efficient utilization of functions of processing server with simultaneous decrease of interference from unauthorized users without complicated processing or authentication operations.

6 cl, 51 dwg

Up!