基于多特征融合的软件代码攻击识别方法  

Software Code Attack Identification Method Based on Multi-Feature Fusion

作　　者：彭图 彭代渊[2] PENG Tu;PENG Dai-yuan(School of Cyberspace Science and Technology,Beijing Institute of Technology,Beijing 100081,China;School of Information Science and Technology,Southwest Jiaotong University,Chengdu Sichuan 610031,China)

机构地区：[1]北京理工大学网络空间安全学院,北京100081 [2]西南交通大学信息科学与技术学院,四川成都610031

出　　处：《计算机仿真》2024年第10期419-423,共5页Computer Simulation

摘　　要：研究软件代码识别方法有利于软件的安全、稳定运行,提出多特征融合下软件代码攻击自适应识别方法。划分软件代码字节,将其映射到图像三色通道中,将软件代码转变为RGBA图,实现软件代码的可视化;将通道注意力机制设置在深度神经网络中,建立特征提取框架,提取并融合软件代码的特征,获得代码多尺度特征;结合AdaBoost算法与C4.5算法通过迭代训练建立强分类器,将代码多尺度特征输入强分类器中,实现软件代码攻击自适应识别。仿真结果表明,所提方法能够有效整合不同数据集的代码特征信息,代码可视化精度高、攻击识别精度高。Researching the software code recognition method is beneficial to the safe and stable operation of software.In this paper,an adaptive method of identifying software code attack based on multi-feature fusion was proposed.At first,we partitioned the software code bytes and mapped them to the three-color channel of image.And then,we transformed the software code into RGBA diagram,thus visualizing software code.Moreover,we introduced the channel attention mechanism into the deep neural network,and constructed the feature extraction framework.Furthermore,we extracted and fused the features of the software code,so that we could get the multi-scale features of the code.Finally,we combined AdaBoost algorithm with C4.5 algorithm to construct a strong classifier through iterative training,and then we input the multi-scale features of code into the strong classifier.Thus,we realized the adaptive identification of software code attacks.The simulation results show that the proposed method can effectively integrate the code feature of different data sets,and has high accuracy of visualization and attack recognition.

关 键 词：代码可视化 通道注意力机制 软件代码攻击 攻击识别 

分 类 号：TP309.5[自动化与计算机技术—计算机系统结构]