深度信念网络的恶意代码分类策略研究

发布时间：2018-04-24 01:13

本文选题：深度信念网络 + 恶意代码　；参考：《小型微型计算机系统》2017年11期

【摘要】：恶意代码的分类是恶意代码分析领域研究的重要问题之一.为解决这一问题,提出深度信念网络(Deep Belief Netw orks,DBN)的恶意代码分类策略.首先,从样本集中提取恶意代码图像特征、指令语句中的频度特征;其次,为确保准确率的提高,将上述两类特征进行融合,训练深度信念网络模型中的限制玻尔兹曼机(Restricted Boltzmann Machine,RBM)和反向传导算法(Back Propagation,BP).实验结果表明,提出的深度信念网络模型对恶意代码的分类平均准确率可达95.7%,明显高于传统浅层机器学习模型KNN的94.5%.
[Abstract]:The classification of malicious code is one of the most important problems in the field of malicious code analysis. In order to solve this problem, a malicious code classification strategy based on Deep Belief Netw (DBN) is proposed. Firstly, the malicious code image features are extracted from the sample set, and the frequency features in instruction statements are extracted. Secondly, in order to ensure the improvement of accuracy, the above two kinds of features are fused. The restricted Boltzmann Boltzmann (RBM) and the reverse conduction algorithm (back Propagation) in the training depth belief network model. The experimental results show that the average accuracy of the proposed model for classification of malicious code can reach 95.775, which is significantly higher than that of the traditional shallow machine learning model (KNN) 94.5.
【作者单位】：新疆大学软件学院;新疆大学网络中心;
【基金】：新疆自治区科技人才培养项目(QN2016YX0051)资助国家社会科学基金项目(12CFX053)资助 2013年度湖北省教育厅科学研究计划项目(B2013041)资助
【分类号】：TP18;TP393.08

【相似文献】