V. A. Litvinenko, S.A. Khovanskov , V. S. Khovanskovа
2021-11-14
Abstract ▼
We consider the issues of protection of distributed computing organized on the basis of a multiagent system for solving problems of multivariate modeling. When modeling, choosing one of the many options may require going through a huge set of parameters that are not available for a high-speed computer. Distributed computing is used to reduce the time needed to solve such problems. There are many different approaches for organizing distributed computing in a computer network: grid technology, metacomputing (BOINC, PVM, and others). All of them are intended for creating centralized distributed computing systems. Distributed computing is organized on the basis of a multi-agent system on the computing nodes of any computer network. When using a large-scale computer network as a computing environment, there may be security threats to distributed computing. One of these threats is getting a false result from hackers during calculations. A false result may lead to making an inappropriate or incorrect decision during the simulation process. Managing agents of a centralized distributed computing system, in addition to managing a distributed system, are forced to detect false results of the calculation process. A method has been developed for calculating the probability of detecting a false result depending on the total number of agents in a multi-agent system and the number of control agents. Examples of calculating the number of control agents that provide the required probability of detecting false results in a multi-agent system are given.
