返回
Android Zero-Day Guard: Zero-Shot Malware Detection Using Deep Learning and Generative Models
DOI:10.1109/TNSM.2026.3671305.png)
摘要
En 中文
本文提出了一种面向Android的零日恶意软件检测方法,命名为Android Zero-Day Guard。通过结合深度神经网络与零样本学习,该方法能够在未接触恶意样本的情况下识别新兴威胁。该方法将APK文件转换为图像并提取深度特征,从而有效捕捉恶意软件的行为模式。实验结果表明,所提出的方法在多个恶意软件家族中达到精度94.93%、召回率93.75%和F1分数94.28%。该方法无需依赖动态分析,表现出强大的检测能力和泛化性能,非常适合新兴威胁的早期识别。尽管模型在基准数据集上表现优异,但在快速演变的威胁环境中部署时,对最新家族的持续验证至关重要。
Keyword:
Malware
Feature extraction
Accuracy
Zero shot learning
Smart phones
Generative adversarial networks
Computational modeling
Data models
Convolutional neural networks
Application programming interfaces
Android zero-day malware
zero-shot learning
Wasserstein generative adversarial network
malware detection
期刊
IF:
5.4
论文数:
590
被引数:
9.2K
机构
引用论文
Malware-SMELL: A zero-shot learning strategy for detecting zero-day vulnerabilities
COMPUTERS & SECURITY
IF5.4
Zero-day Ransomware Attack Detection using Deep Contractive Autoencoder and Voting based Ensemble Classifier
APPLIED INTELLIGENCE
IF3.5
Unveiling vulnerabilities in deep learning-based malware detection: Differential privacy driven adversarial attacks
COMPUTERS & SECURITY
IF5.4

