COOPEVO-CLOUDSEC METHOD FOR OPTIMIZING COMPUTING RESOURCES OF CLOUD SYSTEMS TO IMPROVE SECURITY

Abstract

The CoopEvo-CloudSec method represents an innovative cooperative evolutionary approach to optimizing the allocation of computing resources in cloud systems in order to improve the level of information security. This method integrates elements of game theory, multi-criteria optimization and dynamic risk analysis, allowing the formation of coalitions between protective components such as virtual machines, containers, firewalls and network policies. The basis is a variable risk profile λ(t), which is updated in real time based on data streams from security event management (SIEM), threat detection (EDR) and telemetry monitoring systems. This provides an adaptive response to cyber threats such as DDoS attacks, data leaks or anomalous activity, minimizing vulnerabilities without excessive resource consumption. The architecture of the method is based on the principles of modularity, interoperability, reliability and security. The central component – CoopEvo-CloudSec Engine (CEEngine) – implements NSGA-II or NSGA-III evolutionary algorithms to generate a set of Pareto-optimal solutions, taking into account the criteria: risk level, performance, latency, cost and energy consumption. The Telemetry Collector module accumulates metrics from Prometheus or OpenTelemetry, the Risk Analyzer evaluates and normalizes threats to form λ(t), the Policy Adapter adapts recommendations to orchestrators such as Kubernetes or OpenStack, and Audit & Explainability provides transparency through the capture of quality metrics (HV, IGD, Spacing). The data flow involves a continuous cycle: from SIEM event capture to policy application, with a runtime decision loop for iterative optimization, including selection, crossover, mutation and coalition correction. The discussion of the results indicates advantages in dynamic environments with high load variability, but notes limitations: sensitivity to telemetry noise (accuracy drop by 10%) and the need for computational resources. The method outperforms static planners, providing flexibility for sectors: finance (compliance), IoT (energy consumption), science (performance). Parameter configurations allow adaptation, with NSGA-III for high-dimensional problems.