机器学习在恶意加密流量检测中的应用及研究
电子技术应用
田睿1,2,张雅勤1,2,董伟1,2,李致成1,2,冯志1,2
1.中国电子信息产业集团有限公司第六研究所;2.华北计算机系统工程研究所
摘要: 随着加密通信的普及,恶意攻击者利用加密流量隐藏活动,传统基于签名和规则的检测方法面临挑战。机器学习为恶意加密流量检测提供了新解决方案。综述了监督学习、非监督学习、深度学习和集成学习在该领域的应用。监督学习通过标记数据识别已知攻击,非监督学习在未标记数据中发现新型攻击模式,深度学习提升了在大数据环境中的特征提取能力,而集成学习则通过模型融合增强系统鲁棒性。研究表明,机器学习显著提高了恶意行为识别的准确性,特别是在复杂数据特征提取和新攻击模式发现方面。
中圖分類號:TP181/TP393.0 文獻標志碼:A DOI: 10.16157/j.issn.0258-7998.245979
中文引用格式: 田睿,張雅勤,董偉,等. 機器學習在惡意加密流量檢測中的應用及研究[J]. 電子技術應用,2025,51(4):1-11.
英文引用格式: Tian Rui,Zhang Yaqin,Dong Wei,et al. The application and research of machine learning in malicious encrypted traffic detection[J]. Application of Electronic Technique,2025,51(4):1-11.
中文引用格式: 田睿,張雅勤,董偉,等. 機器學習在惡意加密流量檢測中的應用及研究[J]. 電子技術應用,2025,51(4):1-11.
英文引用格式: Tian Rui,Zhang Yaqin,Dong Wei,et al. The application and research of machine learning in malicious encrypted traffic detection[J]. Application of Electronic Technique,2025,51(4):1-11.
The application and research of machine learning in malicious encrypted traffic detection
Tian Rui1,2,Zhang Yaqin1,2,Dong Wei1,2,Li Zhicheng1,2,Feng Zhi1,2
1.The Sixth Research Institute of China Electronics Information Industry Group Corporation Limited; 2.North China Research Institute of Computer System Engineering
Abstract: With the widespread use of encrypted communication, malicious attackers increasingly exploit encrypted traffic to conceal their activities, posing challenges to traditional signature-based and rule-based detection methods. Machine learning provides a novel solution for detecting malicious encrypted traffic. This paper reviews the applications of supervised learning, unsupervised learning, deep learning, and ensemble learning in this domain. Supervised learning identifies known attacks using labeled data, while unsupervised learning uncovers new attack patterns in unlabeled data. Deep learning enhances feature extraction capabilities in large-scale data environments, and ensemble learning strengthens system robustness through model fusion. The findings indicate that machine learning significantly improves the accuracy of malicious behavior detection, particularly in complex feature extraction and the identification of new attack patterns.
Key words : encrypted traffic recognition;machine learning;encrypted traffic;malicious behavior detection;ensemble learning
引言
隨著數字化發(fā)展,網絡安全成為全球重要挑戰(zhàn),尤其是加密流量中的惡意行為識別。機器學習在加密流量檢測中展現出重要應用。本文綜述了監(jiān)督學習、非監(jiān)督學習、深度學習和集成學習在惡意加密流量分析中的應用,探討其對網絡安全的影響。首先,分析監(jiān)督學習(如決策樹、SVM、隨機森林)在標注數據集上的模式識別能力及其優(yōu)劣;接著討論非監(jiān)督學習(如K均值、層次聚類)在無標簽環(huán)境中的異常檢測;然后研究深度學習(如CNN、RNN)在時間序列分析中的優(yōu)勢,通過自動提取特征提升檢測性能;最后評估集成學習(如隨機森林、AdaBoost)結合多模型提升檢測精度和魯棒性。
本文詳細內容請下載:
http://ihrv.cn/resource/share/2000006386
作者信息:
田睿1,2,張雅勤1,2,董偉1,2,李致成1,2,馮志1,2
(1.中國電子信息產業(yè)集團有限公司第六研究所,北京 100083;
2.華北計算機系統(tǒng)工程研究所,北京 100083)

此內容為AET網站原創(chuàng),未經授權禁止轉載。
