Artificial intelligence algorithms for optimizing SDN functioning: modern approaches and prospects

DOI: 10.31673/2412-9070.2025.041241

Abstract

This article presents a comprehensive review of modern approaches to the application of artificial intelligence (AI) algorithms for optimizing the functioning of Software-Defined Networks (SDN). SDN, as a modern approach that separates control and data planes, provides a flexible architecture for centralized network management. However, the increasing complexity of network traffic, rising security threats, and growing energy demands challenge traditional static control mechanisms. In this context, AI emerges as a key enabler of intelligent, adaptive, and scalable SDN solutions.
The paper explores the application of AI algorithms to address key SDN challenges: Quality of Service and Experience (QoS / QoE), security, load balancing and resource distribution, failure detection and recovery, and energy efficiency. For each aspect, the study systematizes current research, highlights algorithmic approaches (including supervised, unsupervised, deep, and reinforcement learning), and provides examples of real-world use cases. AI-based models such as SVM, LSTM, CNN, and DRL have demonstrated significant effectiveness in traffic classification, DDoS detection, intelligent routing, and proactive network reconfiguration.
Results across multiple publications show that AI integration improves SDN metrics: reducing packet loss and latency, increasing attack detection accuracy (up to 95 %), minimizing controller overloads, and saving energy by dynamically optimizing active components. The study highlights existing challenges, including intensive computational demands, the requirement for high-quality training data, and difficulties in explaining the behavior of opaque AI models. SDN offers the infrastructure for programmable control, while AI supplies the intelligence to make proactive, data-driven decisions. The analysis reveals the potential for further integration, particularly through hybrid approaches combining classical engineering with AI-driven optimization for robust, scalable, and auto-nomous networking.

Keywords: Software-Defined Networking; Artificial Intelligence; Machine Learning; QoS; Load Balancing; Network Security; Energy Efficiency; Intrusion Detection.