函数语义识别的PHP漏洞检测  

PHP Vulnerability Detection Based on Function Semantic Recognition

作　　者：李家梁 彭建山[1] LI Jialiang;PENG Jianshan(Information Engineering University,Zhengzhou 450001,China)

机构地区：[1]信息工程大学,河南郑州450001

出　　处：《信息工程大学学报》2025年第1期90-96,共7页Journal of Information Engineering University

基　　金：河南省重大科技专项(221100240100)。

摘　　要：针对基于深度学习的漏洞检测方法存在源代码分析时间开销大、忽略函数语义导致准确率下降等问题,提出一种针对PHP脚本语言的函数语义识别方法,来增强SQL漏洞检测时对函数语义的理解。通过函数原型扩展的语义相似度分析,自动识别目标函数;根据识别结果,利用静态分析方法对识别后的函数和全局代码进行SQL漏洞检测。实验结果表明,所提方法识别敏感数据库操作函数的准确率达到83.24%。通过引入函数语义识别,实现更加高效的漏洞检测,不仅降低漏洞检测的时间开销,而且能够有效检测出Web应用系统中的真实漏洞。To address the issues of high time overhead in source code analysis and decreased accuracy due to the neglect of function semantics in vulnerability detection methods based on deep learning,a function semantic recognition method for PHP script language is proposed to enhance the understanding of function semantics during SQL vulnerability detection.Target functions are automatically identified through semantic similarity analysis extended from function prototypes.Based on the identification results,SQL vulnerability detection is performed using static analysis methods on the identified functions and the global code.Experimental results demonstrate that the accuracy of the proposed method in identifying sensitive database operation functions reaches 83.24%.By incorporating function semantic recognition,a more efficient vulnerability detection is achieved,reducing both the time overhead and effectively identifying real vulnerabilities in Web application systems.

关 键 词：语义识别 深度学习 漏洞检测 静态分析 

分 类 号：TP315[自动化与计算机技术—计算机软件与理论]