基于遗传算法和SMOTE的网络入侵检测模型优化研究

A study on improving the network intrusion detection model using genetic algorithm and SMOTE

针对网络入侵检测存在数据不平衡和特征冗余的问题,提出一种新的检测模型。该模型结合了遗传算法和SMOTE算法,通过对数据进行采样和特征选择,提高了网络入侵检测的准确性。首先,为了解决数据不平衡的状况,采用了SMOTE算法。这个算法通过在入侵类样本中嵌入随机样本,有效地提高了入侵类样本的数量,使得数据达到平衡。其次,为了缓解特征冗余,引入了基于遗传算法和随机森林方法的包装式特征选择技术,选择有用特征,减少冗余信息,从而提升最终的入侵检测性能。最后,采用随机森林算法对经过上述处理的数据集进行分类,实现对网络入侵样本的有效检测。在NSL⁃KDD数据集上的实验表明,基于遗传算法和SMOTE的网络入侵检测模型从整体上提高了入侵检测的识别率。

A new detection model is proposed to address the issues of data imbalance and feature redundancy in network intru⁃sion detection.The model integrates genetic algorithms and the Synthetic Minority Over⁃sampling Technique(SMOTE)to enhance the accuracy of network intrusion detection.Firstly,to address data imbalance,the SMOTE algorithm is employed,inserting ran⁃dom samples between minority class instances to effectively increase their quantity and achieve inter⁃class balance.Moreover,a wrapper feature selection process,based on genetic algorithms and random forests,has been established to reduce feature redun⁃dancy.This process not only picks useful features and lessens unnecessary data,but also enhances the overall efficiency of intru⁃sion detection.Finally,the processed dataset undergoes classification using the random forest algorithm,enabling effective detec⁃tion of network intrusion instances.Experiments on the NSL⁃KDD dataset show that the network intrusion detection model based on genetic algorithm and SMOTE improves the overall recognition rate of intrusion detection.