摘要： DNS滥用已成为网络空间安全治理中面临的最具挑战性的威胁之一。针对现有检测方法多以DNS请求流量为研究对象，忽略了响应流量特征的问题，提出一种基于长短期记忆网络(Long-Short Term Memory，LSTM)的DNS隐蔽信道检测方法。综合分析请求与响应流量特征，提取响应流量中时间戳、TTL、响应分组长度等特征点，并构建LSTM模型进行训练。实验结果表明，该方法在准确率、F1评分等指标方面取得了良好的结果，较现有方法有显著提高。

中图分类号： TP393.08
文献标识码： A
DOI： 10.19358/j.issn.2096-5133.2022.04.009
引用格式： 陈解元. 基于LSTM的DNS隐蔽信道检测方法[J].信息技术与网络安全，2022，41(4)：60-64，89.

DNS covert channel detection method based on LSTM

Abstract： DNS abuse has become one of the most challenging threats in cyberspace security governance.As the existing detection methods mostly focus on DNS request traffic but ignore the characteristics of response traffic,this paper proposed a DNS covert channel detection method based on Long Short Term Memory(LSTM). The characteristics of request and response traffic were comprehensively analyzed and the feature points such as timestamp, TTL and response packet length from response traffic were extracted，then the LSTM model was constructed for training.The experimental results show that the proposed method achieves good results in accuracy, F1 score and other indicators, which are significantly improved compared with existing methods.

Key words : DNS covert channel；machine learning；Long-Short Term Memory(LSTM)