有效的安卓应用软件可信性评估方法

Efficient trustworthiness evaluation method for Android application software

为保障安卓应用环境中智能终端用户的信息安全,利用支持向量机(support vector machine,SVM),结合证据理论,提出一种面向安卓应用软件的可信性评估方法。该方法分为训练阶段和评估阶段实现,在前阶段,对应用软件样本的属性进行度量,即形成属性证据;利用属性证据对各关联SVM进行训练以确定分类模型。在后阶段,利用待测应用软件的度量结果和SVM分类模型,确定与各属性相关联的分类概率赋值;在对各属性证据赋予自适应权重的基础上,利用改进证据理论对各属性实现合成,用于评估软件的可信性。测试结果验证了所提方法的有效性。

To ensure the information security of intelligent terminal users in Android open application environment,a credibility evaluation method for Android application software was proposed using support vector machine(SVM)and evidence theory.The method was divided into training stage and evaluation stage.In the first stage,the attributes of application software samples were measured,i.e.,attribute evidence was formed.Attribute evidence was used to train each associated SVM to determine the classification model.In the later stage,the classification probability assignment associated with each attribute was determined using the measurement results of the application software to be tested and the SVM classification model.On the basis of giving adaptive weight to each attribute evidence,the improved evidence theory was used to synthesize each attribute to evaluate the credibility of the software.The test results verify the effectiveness of the proposed method.