微型电脑应用2025，Vol.41Issue(7)：242-246,251,6.
基于改进深度残差网络的大规模网络恶意域名检测
Large-scale Network Malicious Domain Name Detection Based on Improved Deep Residual Network
张玉澎1
作者信息
- 1. 河南医药健康技师学院,基础教学部,河南,开封 475000
- 折叠
摘要
Abstract
In response to the challenge of malicious domain name proliferation in the current large-scale network environment,this paper proposes a large-scale network malicious domain name detection method based on improved deep residual network to efficiently and accurately identify malicious applications in the network.Using TransH model,accurate embedding representa-tions of various types of instance data in the graph model is obtained.On this basis,a deep residual network with attention mechanism is constructed to continuously replace tail entities based on network malicious application candidate entities.The scoring function of candidate entities is used to effectively identify potential malicious applications in the network.The deep re-sidual network is used to extract and fuse deep level features of domain name data,and the attention mechanism is combined to scale and weight the output features of convolutional layers.Through convolution processing,network malicious application features are mapped onto target labels for training,achieving high-precision detection of large-scale network malicious applica-tions.The experimental results show that the AUC of the proposed method remains around 0.9957,which is superior to detec-tion methods of single graph inference algorithm and single deep residual network,and accuracy is stable at over 90％.This in-dicates that the proposed method has good performance in detecting network malicious applications and improves the efficiency and accuracy of large-scale network malicious application detection.关键词
图推理/深度残差网络/大规模网络/恶意应用/注意力机制Key words
graph inference/deep residual network/large-scale network/malicious application/attention mechanism分类
信息技术与安全科学引用本文复制引用
张玉澎..基于改进深度残差网络的大规模网络恶意域名检测[J].微型电脑应用,2025,41(7):242-246,251,6.
