基于CNN的JSP类型Webshell检测  

CNN-based Webshell Detection for JSP Types

作　　者：巩思越 张英韬 王宝会[1] GONG Si-yue;ZHANG Ying-tao;WANG Bao-hui(Beihang University,Beijing 100000,China;Wuhan University,Wuhan 430000,China)

机构地区：[1]北京航空航天大学软件学院,北京100000 [2]武汉大学,湖北武汉430000

出　　处：《新型工业化》2021年第10期97-100,共4页The Journal of New Industrialization

摘　　要：Webshell是一种隐蔽性较高的Web攻击工具,其作用是获取服务器的操作权限。目前针对JSP类型的Webshell检测较少,效果一般。针对这一现状,本文从文本分类的角度出发,提出一种基于CNN的Webshell检测方法。首先将JSP文件进行预处理,得到全文ASCII码,转换为灰度图像,使用CNN算法进行分类,以查全率与查准率作为评估标准。最终实验结果表明,CNN检测效果优于其他算法模型,证明该算法是可行的。Webshell is a stealthy Web attack tool, its role is to obtain the operating privileges of the server.At present, the detection of Webshell for JSP type is less and the effect is general. To address this situation, this paper proposes a CNN-based Webshell detection method from the perspective of text classification. Firstly, JSP files are preprocessed to get full text ASCII codes, converted to grayscale images, classified using CNN algorithm, and evaluated by the check-all rate and check-accuracy rate. The final experimental results show that CNN detection is better than other algorithmic models, which proves that the algorithm is feasible.

关 键 词：WEBSHELL JSP 卷积神经网络 ASCII 

分 类 号：F407.61[经济管理—产业经济]