基于HMM-CBOW行为模型的动态可信度量系统设计  

Design of Dynamic Trusted Measurement System Based on HMM-CBOW Behavior Model

作　　者：庞飞[1] 唐六华[1] 谢小赋[1] PANG Fei;TANG Liuhua;XIE Xiaofu(No.30 Institute of CETC,Chengdu Sichuan 610041,China)

机构地区：[1]中国电子科技集团公司第三十研究所,四川成都610041

出　　处：《信息安全与通信保密》2022年第9期65-73,共9页Information Security and Communications Privacy

摘　　要：针对静态可信度量系统难以检测软件运行过程中的异常行为,无法保证软件动态运行安全可信等问题,通过对软件序列行为模型进行分析,提出了一种基于HMM-CBOW行为模型的软件动态可信度量系统设计方案。该方案采用软件行为静态分析与短序列动态划分相结合的方法,构建出细粒度的行为基线库,通过对运行行为的动态监控与可信评估,实现了对软件的动态可信度量功能,能够确保软件运行过程安全可信。最后,通过实验表明:相对于传统N-gram模型的可信度量系统,基于HMM-CBOW行为模型具有较高的可信评估准确率与较低的时延。To address the problems that a static trusted measurement system is difficult to detect abnormal behaviors in the process of software running and cannot guarantee safe and trusted dynamic operation of software,this paper proposes design scheme for software dynamic trusted measurement system based on HMM-CBOW behavioral model by analyzing the software sequence behavioral model.The scheme uses a combination of static analysis of software behavior and dynamic division of short sequences to build a fine-grained behavior baseline database,which achieves a dynamic trusted measurement function for software through dynamic monitoring and reliability evaluation of running behavior,and can ensure the reliability of software running process.Finally,experiments indicate that the system based on the HMM-CBOW behavioral model has a higher reliability evaluation accuracy and lower latency compared with the traditional N-gram model trusted measurement system.

关 键 词：行为模型 行为基线 动态可信度量 动态监控 

分 类 号：TP309.1[自动化与计算机技术—计算机系统结构]