METHODOLOGY FOR CONSTRUCTING AN ARTIFICIAL INTELLIGENCE PLATFORM FOR MONITORING CONFIDENTIAL AND ANOMALOUS TRAFFIC WITH MINIMIZATION OF DEPLOYMENT TIME AND RELIABILITY OF OPERATION

Abstract

The paper presents the results of developing a model for building an artificial intelligence platform under the condition of minimizing deployment time and ensuring reliability of operation during monitoring of corporate and anomalous traffic of the information and communication network. When creating the model, MLOps technology was used, which shows that its use allows you to create an effective way to deploy artificial intelligence platforms. A feature of the presented binary tree is its relatively low reliability, when in case of a single failure, the topology can actually be divided into two parts. To minimize this problem, it is advisable to use the De Bruyne technology to form additional ties.

The presented model for building the platform has the ability to automate and monitor the stages of data collection, transformation, training, implementation and presentation of results to the end user. The platform deployment stage provides quick integration, minimizing deployment time, integration and reliability of the model with its constant monitoring. The paper highlights the results of assessing the effectiveness of the application of the developed artificial intelligence platform model, which shows that when developing the platform topology from 3 to 63 nodes, the deployment time increases by only 434 seconds, which indicates the platform's ability to adapt and develop. Load testing showed that the presented platform model is able to function without failures, provided that the request queue is loaded with 25 threads simultaneously, while the use of additional nodes ensures the minimization of the average system response time.