移动APP多核架构安全漏洞并行检测  

Parallel Detection of Security Vulnerabilities in Multi-core Architecture of Mobile APP

作　　者：严文昊 王宏岩 董蓓 曲艺 孙德艳 YAN Wen-hao;WANG Hong-yan;DONG Bei;QU Yi;SUN De-yan(State Grid Customer Service Centre,Tianjin 300300,China;Beijing China Power Information Technology Co.,Ltd.,Beijing 100031,China)

机构地区：[1]国家电网有限公司客户服务中心,天津300300 [2]北京中电普华信息技术有限公司,北京100031

出　　处：《计算机技术与发展》2025年第2期79-85,共7页Computer Technology and Development

摘　　要：移动APP多核架构的安全漏洞具有多样化、大规模特征,现有检测方法受限于参数选择适应性和特征提取的准确性,难以准确并行检测移动APP的多种漏洞。为了解决移动APP多核架构下大规模安全漏洞数据的准确检测问题,提出了一种基于SimHash算法和LightGBM的移动APP多核架构安全漏洞并行检测方法。该方法利用SimHash算法对漏洞数据进行特征提取和编码。利用TF-IDF算法融合Jaccard指数改进SimHash算法,优化特征词的权重分配计算,生成唯一特征哈希值。结合LightGBM算法构建二分类器,判断漏洞是否存在可利用的EXP(漏洞利用代码)。利用贝叶斯超参数优化LightGBM算法,通过多核架构的并行处理,实现对大量漏洞数据的准确检测。通过实验表明,该方法针对多种常见漏洞的MAE值、RMSE值、MAPE值、R^(2)值分别为0.032、1.017、0.124%、0.976,有效提升了漏洞检测的适应性、精度、稳定性和拟合能力,为移动APP的安全管理提供了有力支持。The security vulnerabilities of multi-core architecture in mobile APP have diverse and large-scale characteristics.Existing detection methods are limited by the adaptability of parameter selection and the accuracy of feature extraction,making it difficult to accurately detect multiple vulnerabilities in mobile APP in parallel.In order to solve the problem of accurate detection of large-scale security vulnerability data in mobile APP multi-core architecture,a parallel detection method for security vulnerabilities in mobile APP multi-core architecture based on SimHash algorithm and LightGBM is proposed.This method utilizes SimHash algorithm for feature extraction and encoding of vulnerability data.Using TF-IDF algorithm and Jaccard index to improve SimHash algorithm,the weight allocation calculation of feature words is optimized,and the unique feature hash values are generated.A binary classifier is built by the LightGBM algorithm to determine if there are exploitable EXP(exploit code)vulnerabilities.By utilizing Bayesian hyperparameters to optimize the LightGBM algorithm and parallel processing with a multi-core architecture,accurate detection of large amounts of vulnerability data can be achieved.Through experiments,it has been shown that the proposed method for various common vulnerabilities have MAE values,RMSE values,MAPE values,and R^(2) values of 0.032,1.017,0.124%,and 0.976,respectively,which effectively improves the adaptability,accuracy,stability,and fitting ability of vulnerability detection,providing strong support for the security management of mobile APP.

关 键 词：SimHash算法 LightGBM 移动APP 多核架构 安全漏洞 并行检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]