一种基于PCA和随机森林分类的入侵检测算法研究

Research on an Intrusion Detection Algorithm Based on PCA and Random-forest Classification

针对现有的入侵检测方法准确率低的问题,文章提出了一种基于PCA(Principal Component Analysis, 主成分分析法)和随机森林分类的入侵检测算法。该算法思想是对训练数据先清洗后分类,首先通过主成分分析法对数据进行特征降维和降噪,然后对处理后的数据使用随机森林分类器进行分类训练。实验使用基于Python的机器学习库scikit-learn,并采用20%的NSL_KDD数据集。实验结果表明,与常用的基于机器学习的入侵检测技术相比,文章提出的入侵检测算法可更有效地提高检测的准确性。

Due to the low accuracy, of existing intrusion detection methods, this paper proposes an intrusion detection algorithm based on PCA （principle component analysis） and random-forest classification. The idea of the algorithm is to clean the training data before classifying. Firstty, PCA is used to decompose the dataset and reduce noises. Then random-forest classifier is used to classify and tram the processed data. The experiment uses machine learning library based on Python called scikit- learn and 20% NSL-KDD dataset. Experimental results show that compared with the commonly used intrusion detection technologies based on machine learning, the intrusion detection algorithm proposed m this paper can improve the detection accuracy more effectively.