Science Belongs to All Mankind

JOURNALS-DETAIL

结合随机森林与XGBoost的恶意软件检测研究

Author:

张静蕾,王佳怡,丁涵,罗养霞

Vol. 2, Issue 1, Pages: 64-66(2025)

Doi:

https://doi.org/10.62639/sspis21.20250201

ISSN:

3006-0737

EISSN:

3006-4309

Views:

86

Downloads:

3

Abstract References Project Publication Info Metrics
Abstract

针对传统的恶意软件检测提取单一特征,输入单个分类器,检测准确率低等问题。本研究提取多重静态特征,构建各自的训练模型,并采用Stacking算法对各模型输出结果进行聚合。将图像纹理特征和操作码特征两种特征与标签数据融合。再集成学习并对比结果。实验结果表明,与传统的恶意软件分类方案相比,基于集成学习的多属性特征恶意软件检测方法的AUC值达到了99.84%。相较于传统的提取单一特征或使用单一分类器的机器学习分类方案,本方法能够更有效的提高对恶意软件随机样本检测和分类的效果。

Keyword

恶意软件检测;特征融合;集成学习;N-Gram方法

International Scientific Studies Press Limited

International Scientific Studies Press Limited is a company boasting rich international communication resources and formidable editorial, translation, and publishing capabilities. Our primary focus revolves around the publication of academic journals. Our establishment's mission is to provide a premier publishing platfor...
FLAT C,23/F,LUCHY PLAZA,315-321 LOCKHART ROAD,WANCHAI,HONG KONG (00852) 65557188
Copyright © 2025 International Scientific Studies Press Limited