Threat modeling the Kubernetes Agent: from MVC to continuous improvement

This post details the iterative evolution of GitLab's threat modeling process specifically for the Kubernetes Agent. It outlines three distinct review phases, showcasing the architectural diagrams and data flow diagrams at each stage. Key technical contributions include identifying specific vulnerabilities like the Agent exposing public projects with private repositories and path traversal attacks, and detailing the evolution of the threat modeling template to include application decomposition, dedicated threat analysis, and conversion to a dedicated repository. It also highlights the discovery of new threats related to listener access, unencrypted communications, and the mitigation strategies for unauthorized project access.