基于深度学习模型的软件源代码归属研究  

Research on Attribution of Software Source Code Based on Deep Learning Model

作　　者：任群 REN Qun(Department of Electronic and Information Engineering,Bozhou University,Bozhou Anhui 236800,China)

机构地区：[1]亳州学院电子与信息工程系,安徽亳州236800

出　　处：《佳木斯大学学报（自然科学版）》2022年第2期66-69,88,共5页Journal of Jiamusi University：Natural Science Edition

基　　金：中药材大数据应用分析与智能预测综合服务系统(2020年亳州市科技重大专项)阶段性成果;安徽省质量工程项目(2020kfkc323)阶段性成果。

摘　　要：针对软件源代码归属问题,结合深度学习模型和程序依赖关系图,设计了软件源代码识别分类方法,以从不同的软件源代码中识别代码的作者。首先,使用程序依赖关系图来捕捉源代码中的控制和数据的关系。通过预处理技术将程序依赖关系图特征转换为具有频率细节的小实例。然后,采用合成少数类过采样技术解决类不平衡问题,采用深度学习算法为每个程序员提取编码风格特征并进行代码作者的分类。最后,使用真实且包含不同编程语言源代码数据集对本方法进行评估,实验结果表明,该方法在精度、召回率和F度量三方面均优于现有技术。Aiming at the problem of software source code attribution,combining the deep learning model and program dependency graph,a software source code identification classification method is designed to identify the author of the code from different software source codes.First of all,the program dependency graph is used here to capture a small example of converting the program dependency graph features into frequency details through advance technology.Then,the synthetic minority class oversampling technology is used to solve the class imbalance problem,and the deep learning algorithm is used to evaluate the method here using real data sets containing source code of different programming languages.The experimental results show that the method here is precision,recall rate and F-measure are all preliminary existing technologies.

关 键 词：软件源代码归属 深度学习 代码特征 软件取证 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]