Vol 24, No 5 (2018)

The possibilities of adaptive step-by-step routing of information packets in a virtual centralized computer network of a complex distributed control system for dynamic processes and objects in real time are considered. The application of a logical neural network for such routing ensures maximum bandwidth, reliability, stealth and survivability of the network. Methods are proposed for organizing multi-channel parallel data collection by the head node of a centralized computer network when using the Internet. Recommendations are given on the development of the architecture of the Elbrus family and interfaces with the environment of functioning.

The main recommendations are as follows:

1. It is necessary to continue the established traditions of using hardware control and error correction tools to build protected control computing processes.
2. It is necessary to develop a fleet of high-speed external interface devices with system resources, as well as with telecommunication and global computer networks. You should use the new direction of creating fast web-objects for exchange with the Internet. This direction will lead to the revision and development of standard exchange protocols.
3. Virtualization of the resource implemented in ERC "Elbrus" gave an excellent effect when solving the tasks of parallelization on the readiness of work in the queue "to the processor". She also solved the problem of sliding reservation without additional costs. However, the combination of centralized and decentralized management of computing resources should also be available.
4. Application of network technologies introduces a special vulnerability in the management system of advanced complex control systems. It seems that the typing of data and type control, embodied in the early Elbrus models, can effectively counteract today’s rapidly developing threats and attacks. Apparently, the tag architecture, which leads to significant excesses in hardware, is not the only way to implement data typing.
5. It is necessary to develop the principles of hardware support for high-level languages. However, reliance on the stack mechanism of performing procedures in its time did not meet the support of the developer of control systems and is the subject of dispute to date.
6. The set of functions of the operating system must be reasonably chosen in accordance with the use of computing facilities.
7. The complete set of models of the Elbrus family should take into account the requirements of standardization and unification. The modular principle of configuration allows the construction of a number of models differing in power and specialization. Modules of actuators must also be composed of variable sets. Their execution on the microprocessor base allows the development and specialization of the command system, as well as the implementation of procedures of limited scope.

The article proposes models of mathematical multiplexing and demultiplexing of flows, as well as expressions for determining the numerical characteristics of the distribution of the outflow intervals from the queuing system (QS). Together, they allow us to write the equilibrium equations for the mean and variance of the distribution of time intervals between consecutive requirements in queuing networks with arbitrary laws receipts and services. On the place of nodes of a queuing network are considered QS M-/M-/1 with delay in time, H₂/H₂/1 and H₂/M/1. Under general assumptions about the probability distribution of time intervals between requirements in the input flows and the time of service in the nodes, the proposed approach makes it possible to determine the average values and variances of time intervals between the requirements of all queuing network flows, as well as all the main indicators of the functioning of such networks. Such an approach can also be extended, if necessary, for third-order moments, which from the point of view of probability theory is more accurate than the calculation at the level of two moments of the distributions. Such an approach can also be extended, if necessary, for third-order moments. From the point of view of probability theory, this will be more accurate than calculating at the level of two moments of distribution. Under incomplete information on the laws of flows distributions, the proposed approach to analyzing the performance of non-Markov queuing networks is currently acceptable.

This article presents a new approach to developing an adaptive genetic optimization algorithm (MAGAMO/A) using agent modeling techniques. The peculiarity of this approach is the support of the mechanism of adaptive control of key characteristics of GA, in particular, the values of the probabilities of crossover operators and mutations, their types and other important characteristics that affect the population diversity and the rate of convergence of GA. Support for adaptive control is provided by using the mechanism of agent state charts and the specified rules of transition between the corresponding states that determine the values of the control parameters of the GA at the individual level of each agent-process. The review of the most popular GAs used for multicriteria optimization, including SPEA2, NSGA, MOEA, etc., is reviewed. The main metrics for evaluating the effectiveness of such GAs (Hypervolume, Generational Distance, distance between solutions on the Pareto boundary, etc.) are considered. The efficiency of the developed approach in the solution of optimization problems of large dimension on several test examples and in comparison with other known GA is demonstrated. The main directions of further research in the field of development of agent-oriented genetic algorithms are formulated.

Theoretical aspects of the formation of applications for editing data in relational databases are considered. This important problem has not yet been satisfactorily resolved. That problem has been solved by "hard" restrictions on the views and it requires the highly qualified user when using these components. In this paper, the theoretical aspects of the applications formation for work with a relational database are discussed. The application works with a data view such as table formed by a set of database relations using " join-select-projection" queries. The implementation of data editing functions: inserting, deleting and updating of tuples in the table are the requirement for such applications. The application must convert the source database according to the user’s view editing providing that the conversion satisfy the commutativity condition. Formulas of relational algebra are obtained for these transformations. The correctness of the transformations is proved. The theory in this paper is the basis for creating a new technology for generating of multitable applications. The technology involves creating a template program, where a query and its restrictions are implemented. The operations of relational algebra are transformed into SQL commands. As a result, the user obtains a ready-to-run program that implements editing of multitable views.

In this article is dedicated to targeted attacks threat modeling approach, which is marked by use of zero-day vulnerabilities by the potential intruder. This feature radically changes the task of modeling, because of software tools vulnerabilities threat reservation absence in such case (which means a system without reservation must be modeled). Security breakups in this case are not managed because of their unknown status (only the potential intruder knows about this vulnerabilities). In common case taking in mind vulnerabilities appearing and eliminating during system work, attack threat can be present in different forms, which needs to be taken into account while threat modeling (making targeted attack threat description more correct). At the beginning threat is known only by potential intruder, then after it’s identification the process of threat eliminating begins. After such process system became restorable, which needs to be taken into account while attack threat modeling (attack can turn it’s form). In common case while attack implementation potential intruder can use both unknown and known vulnerabilities which create attack threat meanwhile known vulnerability is not yet patched. Let’s review how to calcul ate these attack threat features inside the mathematical model and which attack threat actuality characteristics (measures) can be defined. We will use Markov model with discrete state and continues time to build attack threat model. The suggested targeted attacks threat modeling approach has such advantages as ability to receive actuality quantitate measures without using of any subjective expert decisions.

In this paper is explored the problem of including OFDM modulation into the set of digital modulation types of radio signals using the multilayer neural network. Recognition is conducted using the cumulant features with known carrier frequency. The test results show that the inclusion OFDM modulation into the set of digital modulation types does not affect on the recognition probabilities of other types, while the recognizing probability of OFDM type is almost 100 %.

The article is devoted to the problem of constructing neural networks of a finite ring (NNFR), which serve as the basis for neural network modular computing structures for high-performance cryptographic applications. The methodological basis of the NNFR of the investigated class is the modified reduction method of position-modular transformation of weighted large numbers. The authors give a mathematical formalization of the method, was obtained estimates for the range of change and the number of elements of the residue sequence formed by the reduction scheme of the recursive type, the nature and speed of its convergence are investigated, a flexible tabular mechanism for reducing the number of iterations of the scheme is proposed. Synthesized general reducing algorithm of position-modular code conversion, was developed the parallel structure of the NNFR which performs a basic transformation in one iteration — during the time of order ( ⎡log2b⎤ + 2)tcл, where b is the number of the input number; tcл is the duration of the two-fold addition operation.