TY  - EJOU
AU  - Xu, Hui 
AU  - Yang, Yonglei 
AU  - Qu, Shuang 

TI  - Metaheuristic-Based SDN Intrusion Detection Model Based on Adaptive Chinese Pangolin Optimizer Algorithm
T2  - Computers, Materials \& Continua

PY  - 
VL  - 
IS  - 
SN  - 1546-2226

AB  - With the widespread deployment of Software-Defined Networking (SDN), its centralized control architecture faces increasingly severe security threats despite improving network flexibility and programmability. Due to the high dimensionality, redundancy, and nonlinearity of SDN traffic data, existing intrusion detection methods often suffer from high computational cost, unstable feature selection, and limited generalization ability. To address these challenges, this paper proposes an SDN intrusion detection model based on an Adaptive Chinese Pangolin Optimizer (ACPO), termed ACPO-IDM. Unlike the original Chinese Pangolin Optimizer (CPO), ACPO integrates hierarchical initialization, adaptive inertia weight, elite local search, and precomputation-based caching strategies. These improvements enhance population diversity, balance global exploration and local exploitation, strengthen local refinement, and reduce repeated computational operations. Therefore, ACPO provides a more stable and lightweight feature selection mechanism for high-dimensional SDN traffic data. The selected compact feature subsets are then used to train intrusion detection classifiers. Experimental results on benchmark functions, UCI datasets, and the InSDN dataset show that ACPO achieves competitive or superior performance compared with several representative metaheuristic algorithms. Statistical tests further confirm the robustness of the proposed optimizer. On the full InSDN dataset, ACPO-IDM achieves 98.952% accuracy in binary classification and 98.653% accuracy in multiclass classification while selecting compact feature subsets. Additional comparisons with recent SDN-oriented and deep learning baselines, as well as multiple classifiers, further verify the effectiveness, robustness, and practical potential of the proposed framework.
KW  - Software-defined networking; intrusion detection; feature selection; metaheuristic algorithm; Chinese pangolin optimizer

DO  - 10.32604/cmc.2026.087053