面向高性能计算环境的多维自适应授权访问策略

Multi-dimensional adaptive access contol for high-performance computing environment

高性能计算能力是国家综合实力和创新能力的重要体现,是支撑我国科技持续发展的关键技术之一。随着高性能计算的发展,越来越多领域的科研人员开始关注并使用高性能计算环境。高性能计算环境目前面临资源有限、用户数目增多等挑战。为保证环境的安全性、提高环境资源的利用率,需设置一定的授权访问策略来约束用户的访问行为。本文针对高性能计算环境服务对象用户和应用社区或业务平台,基于机器学习算法对用户行为进行分析获取相关属性,设计并实现了一种多维自适应授权访问策略(MAAC)。实验表明,MAAC可实现对环境资源有效和灵活访问控制,同时该策略的决策时间可控制在1 ms内,与策略响应时间相比可忽略不计。

High-performance computing(HPC)capability is an important manifestation of a country's comprehensive strength and innovation capability,and is one of the key technologies supporting the sustainable development of science and technology in China.With the development of high-performance computing,more and more researchers in the field have started to pay attention to the HPC environment.Now the HPC environment is facing challenges such as limited resources and increasing number of accounts.In order to ensure the security of the environment and improve the utilization of the environment resources,certain authorized access policies need to be set to constrain the access behavior of users.In this paper,a multi-dimensional adaptive access control(MAAC)policy is designed and implemented.The policy is based on machine learning algorithms to analyze user behavior and obtain relevant attributes for users and application communities or business platforms,which are served by the HPC environment.Experimental results show that MAAC can achieve effective and flexible access control to environmental resources.Meanwhile the determination time of the MAAC can be controlled within 1 ms,which is negligible compared with the response time.