物联网是继计算机、互联网和移动通信之后的又一次信息产业革命。目前，物联网已经被正式列为国家重点发展的战略性新兴产业之一，其应用范围几乎覆盖了各行各业。物联网中存在的网络入侵等安全问题日趋突出，在大数据背景下，文章提出一种适用于物联网环境的入侵检测模型。该模型把物联网中的入侵检测分为数据预处理、特征提取和数据分类3部分。数据预处理主要解决数据的归一化和冗余数据等问题；特征提取的主要目标是降维，以减少数据分类的时间；数据分类中引入平衡二叉决策树支持向量机（SVM）多分类算法，选用BDT-SVM算法对网络入侵数据进行训练和检测。实验表明，选用BDT-SVM多分类算法可以提高入侵检测系统的精度；通过特征提取，在保证精度的前提下，减少了检测时间。%The Internet of Things (IoT) is another information industry revolution after the computer, the Internet and the mobile communications. At present, IoT has been ofifcially listed as one of the national strategic emerging industries, and its application range covers almost all areas. Secure problems such as network intrusion in the IoT art prominent increasingly. In the big data context, this paper proposes an intrusion detection model that is suitable for IoT which divides the intrusion detection procedure into three parts, which are data preprocessing, features extraction and data classiifcation. Data normalization and data redundancy reduction are solved in the data preprocessing. The main goal of features extraction is to reduce the dimension and thus to reduce the time of data classiifcation. Support vector machine with balanced binary decision tree algorithm that is named BDT-SVM is introduced in the data classiifcation for training and testing the network intrusion data. Experimental results show that it can improve the accuracy of intrusion detection system by using the BDT-SVM algorithm and reduce the detection time with features extraction in the premise of ensuring accuracy.
展开▼
