摘要
提出了一种基于深度学习的高鲁棒性恶意软件识别算法,该算法利用软件的操作码序列来检测恶意软件。首先采用类信息增益进行特征选择,然后提出了基于启发式规则的图生成算法,并将图转换为矢量空间,最后应用基于堆叠自编码器的深度学习框架对恶意和正常软件进行分类。实验评估结果说明了与现有的算法相比,恶意软件识别算法具有较高的鲁棒性。
This paper proposes a high robust malware recognition algorithm based on deep learning,which uses software's opcode sequence to detect malware.Firstly,we use class information gain to select features,then propose a graph generation algorithm based on heuristic rules and convert the graph into vector space.Finally,we apply stacked auto coder-based deep learning method to classify malicious and normal software.The experimental evaluation results show that the malware recognition algorithm of this paper has higher robustness than the existing algorithms.
作者
彭伟
PENG Wei(Department of Computer and Information Engineering , Anhui vocational & technical college of industry & trade, Huainan 232007, China)
出处
《信阳农林学院学报》
2020年第1期117-121,共5页
Journal of Xinyang Agriculture and Forestry University
基金
安徽省教育厅高等学校省级质量工程项目(2018jyxm1328).
关键词
恶意软件识别
深度学习
图生成
鲁棒性
malware recognition
deep learning
graph generation
robustness
