Sarcouncil Journal of Engineering and Computer Sciences

Abstract: Software-Defined Networking has revolutionized network management by enabling centralized programmability and dynamic traffic control, yet traditional architectures encounter significant scalability bottlenecks during real-time traffic classification and steering operations. The distributed machine learning-based flow classification system presents a paradigm shift toward intelligent, adaptive network infrastructures that autonomously identify, classify, and route traffic without compromising performance. This system embeds lightweight machine learning models directly within network switches and hosts, enabling automatic recognition of diverse traffic types, including multimedia streams, voice communications, bulk data transfers, and malicious traffic patterns through sophisticated packet characteristic evaluation. The distributed architecture incorporates edge intelligence modules, classification engines utilizing optimized algorithms, and rule programming interfaces that generate forwarding rules within microsecond timeframes. Performance optimization through distributed learning mechanisms enables line-rate traffic processing while maintaining adaptive quality-of-service management and comprehensive security enforcement. The system addresses critical implementation challenges, including hardware heterogeneity through platform abstraction layers, security concerns via cryptographic mechanisms and differential privacy techniques, and network dynamics through automated topology adaptation and fault recovery mechanisms. Future integration opportunities encompass edge computing convergence, next-generation wireless networks with network slicing capabilities, intent-based networking frameworks, and explainable artificial intelligence systems that enhance operator trust and troubleshooting capabilities.