Improved Data Flow Matching in SDN Using Machine Learning

Data flow subsequence matching is crucial in Software-Defined Networking (SDN) due to the growing demand for efficient data handling. Traditional methods face limitations in accurately and rapidly matching data subsequences, especially under the complexities introduced by non-linear patterns. This paper proposes a machine learning-based algorithm designed for high-accuracy data flow subsequence matching in SDN. The algorithm utilizes feature extraction and transformation for each subsequence and applies similarity measurement techniques to improve matching precision. Furthermore, it enhances a neural network structure by converting a simple neural network into a modular network, improving convergence and performance. Experimental analysis demonstrates that the proposed algorithm achieves shorter feature transformation times, high accuracy in similarity measurement, and improved matching rates across diverse datasets, surpassing traditional methods. This solution provides a reliable approach to optimizing data flow management in SDN.