In this paper, we propose quantitative evaluation method that enable security comparison between Security Container Runtimes. security container runtime technologies have been developed to address security issues such as Container escape caused by containers sharing the host kernel. However, most literature provides only a analysis of the security of container technologies using rough metrics such as the number of available system calls, making it difficult to compare the secureness of container runtimes quantitatively. While the proposed model uses a new method of combining the degree of exposure of host system calls with various external vulnerability metrics. With the proposed technique, we measure and compare the security of runC (Docker default Runtime) and two representative Security Container Runtimes, gVisor, and Kata container.

Container Security, Container Runtime, Vulnerability, System call, Exploit