基于改进多因子优化蝙蝠算法的网络入侵检测方法OA北大核心CSTPCD

In addressing the challenge of diminished intrusion detection accuracy resulting from the abundance of redundant and irrelevant features in high-dimensional network data,an improved multi-factorial optimization bat al-gorithm(IMFBA)was introduced for precise data feature selection,with the ultimate goal of improving network in-trusion detection accuracy.Within the multi-factorial optimization framework,global and local feature selection tasks were formulated.Information exchange between these tasks was facilitated by selection and vertical cultural transmission operators,strategically designed based on the bat algorithm.The global feature selection task was ac-celerated in identifying optimal solution spaces,thereby enhancing the algorithm's convergence speed and stability.By incorporating the reverse learning strategy and differential evolution into the bat algorithm,the initial solution se-lection stage and individual updating process were refined to address the absence of a mutation mechanism,foste-ring solution diversity and aiding the algorithm in escaping local optima.An adaptive parameter adjustment strategy was introduced,determining weightings for guiding individual updates based on potential optimal solution quality.This could mitigate the risk of knowledge negative transfer during multi-task feature selection,achieving a balance between global exploration and local exploitation.The feature subsets selected by IMFBA demonstrate classification accuracy of 95.37%and 85.14%on the KDD CUP 99 and NSL-KDD intrusion detection datasets,respectively.This reflected increased by 3.01 percentage points and 9.78 percentage points compared to the complete dataset.Experiment results confirm the efficacy of EMFBA in selecting higher-quality feature subsets and,consequently,enhancing network intrusion detection accuracy.