中文摘要：

为了有效检测恶意软件,减少恶意软件对安卓平台的安全造成的威胁,在对现有数据集分析研究的基础上,提出概率统计和特征抽取两种策略,分别用这两种策略对提取的特征进行降维处理,减少不确定性数据,再用线性支持向量机（support vector Machine,SVM）分类,模型训练时间缩短为原来的16.7%,并且检测未知恶意软件的准确率明显提高。将该降维策略在其他常用算法上进行试验,结果表明改进后的数据有助于提高这些算法的分类准确率。

英文摘要：

In order to detect malware effectively and reduce the threat of malicious software on Android platform security,two strategies that were probability statistics embedding and feature extraction were proposed based on the analysis of existing data sets. These strategies were used to transform high-dimensional data into low-dimensional data so as to reduce the dimension and the uncertainty of the extracted features. Support vector machine were used to classify these data. With these strategies,the time complexity of training process was reduced to 16. 7 percent of the original time,and the ability of detecting unknown malware families was improved obviously. Moreover,these strategies were used with some popular classification algorithms,and the experimental results revealed that these strategies could achieve a better detection rate.