基于改进残差网络的光通信网络漏洞自动辨识研究  

Research on automatic identification of optical communication network vulnerability based on improved residual network

作　　者：任美丽 孟亮 李婷[1] REN Meili;MENG Liang;LI Ting(Shanxi University of Finance and Economics,Taiyuan 030006,China;College of Information and Computer,Taiyuan University of Technology,Taiyuan 030024,China)

机构地区：[1]山西财经大学,太原030006 [2]太原理工大学信息与计算机学院,太原030024

出　　处：《激光杂志》2022年第12期133-138,共6页Laser Journal

基　　金：山西省自然科学基金项目(No.201801D121010)。

摘　　要：光通信网络存在漏洞辨识时延较长等问题,提出基于改进残差网络的光通信网络漏洞自动辨识方法。根据覆盖度和畸变设计一种灰盒漏洞挖掘模型,实施光通信网络漏洞数据挖掘,模型可以分为符号执行、样本选择、运行时跟踪、畸变策略四部分。通过对挖掘的光通信网络漏洞数据实施预处理,实现漏洞自动辨识之前的信息整合,具体预处理步骤包括反编译、代码切片、分词与向量化表示。基于卷积神经网络改进的残差网络构建残差池化识别模型,在模型中输入预处理后的光通信网络漏洞数据,实现光通信网络漏洞自动辨识。设置残差池化识别模型参数,通过matlab软件测试设计方法的性能。测试结果如下:设计方法特异度最高可达89.36%,查全率最高可达90.3210%,查准率最高可达89.2558%,准确率最高可达89.6325%,高于对比测试方法;设计方法的漏洞辨识时延与模型训练时间小于其他三种测试方法,表明设计方法的自动辨识性能良好。Existing research results have problems such as long model training time and long loophole identification delay.An automatic identification method for optical communication network loopholes based on improved residual network is proposed.According to the coverage and distortion,a gray-box vulnerability mining model is designed to implement optical communication network vulnerability data mining.The model can be divided into four parts:symbolic execution,sample selection,runtime tracking,and distortion strategy.By preprocessing the exploited optical communication network vulnerability data,the information integration before the automatic identification of the vulnerability is realized.The specific preprocessing steps include decompilation,code slicing,word segmentation and vectorized representation.Based on the residual network improved by the convolutional neural network,the residual pooling recognition model is constructed,and the pre-processed optical communication network vulnerability data is input into the model to realize the automatic identification of optical communication network vulnerabilities.Set the residual pooling identification model parameters,and test the performance of the design method through matlab software.The test results are as follows:the specificity of the design method is up to 89.36%,the recall rate is up to 90.3210%,the precision rate is up to 89.2558%,and the accuracy rate is up to 89.6325%,which is higher than the comparison test method;loopholes in the design method The identification time delay and model training time are less than the other three test methods,indicating that the design method has good automatic identification performance.

关 键 词：改进残差网络 光通信网络 灰盒漏洞挖掘模型 漏洞辨识 反汇编 

分 类 号：TN311[电子电信—物理电子学]