基于机器学习的恶意PNG图像识别方法  

Malicious PNG Image Recognition Method Based on Machine Learning

作　　者：马秋豪 MA Qiuhao(Guangxi Industrial Vocational and Technical College,Nanning 530001,China)

机构地区：[1]广西工业职业技术学院,广西南宁530001

出　　处：《电视技术》2024年第4期203-206,共4页Video Engineering

摘　　要：为避免网络病毒借助便携式网络图形(Portable Network Graphics,PNG)传播恶意代码,应用机器学习技术,提出一套行之有效的恶意PNG图像识别方法。该识别方法采用动态分析的方式,评估PNG图像和加载器行为,同时结合基于传统调用特征的机器学习算法,分类并识别恶意PNG图像。结果表明,提出的恶意PNG图像识别方法具有较高的可靠性和可行性,可以保证对多种恶意PNG图像的识别效果,还能避免用户隐写处理健康图像信息,完全符合图像水印隐秘传输应用需求。In order to prevent network virus from spreading malicious code by means of Portable Network Graphics(PNG),this paper uses machine learning technology to propose a set of effective malicious PNG image recognition methods.This recognition method uses dynamic analysis to evaluate PNG images and loader behavior,and combines machine learning algorithms based on traditional call features to classify and identify malicious PNG images.The results show that the malicious PNG image recognition method proposed in this paper has high reliability and feasibility,can ensure the recognition effect of a variety of malicious PNG images,and can avoid the user steganization of healthy image information,which fully meets the application requirements of image watermarking covert transmission.

关 键 词：机器学习 恶意图像识别 便携式网络图形(PNG) 动态分析 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]