基于深度学习的集成DGA域名检测方法  

DGA domain ensemble detection method based on deep learning

作　　者：罗赟骞 邬江 王艳伟 杨鹤 Luo Yunqian;Wu Jiang;Wang Yanwei;Yang He(China Electronics Cyberspace Great Wall Limited Company,Beijing 102209,China)

机构地区：[1]中电长城网际系统应用有限公司,北京102209

出　　处：《信息技术与网络安全》2018年第10期10-14,共5页Information Technology and Network Security

摘　　要：针对DGA域名难以有效检测的问题,提出了一种融合深度学习中CNN模型和RNN模型的集成检测模型。集成检测模型由字符嵌入层、特征提取层和分类层三部分组成。字符嵌入层完成对输入字符的自动编码;特征提取层采用CNN模型和RNN模型从空间和时间的角度自动提取输入字符的特征;分类层采用三层全连接神经网络实现对DGA域名的自动预测分类。实验结果表明,集成检测模型与集成CNN模型相比能够有效提高检测效果。To solve the problem that DGA domain name is difficult to be effectively detected,the ensemble detection model integrating convolutional neural network(CNN)model and recurrent neural network(RNN)model in deep learning is proposed.The ensemble detection model consists of character embedding layer,feature extraction layer and classification layer.The character embedding layer realizes the automatic encoding of the input characters.Feature extraction layer adopts CNN model and RNN model to automatically extract the characteristics of input characters from the perspective of space and time.The classification layer uses three-layer fully connected neural network to predict the classification of DGA domain names.The results show that compared with the integrated CNN model the ensemble detection model can effectively improve the detection effect.

关 键 词：网络空间安全 动态域名生成算法 卷积神经网络 长短期记忆网络 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]