Om sesjonen
Modern software delivery chains create a class of indirect privilege that most security teams don't model. A reusable workflow shared across teams, where the repository that defines it has weaker access controls than the ones that call it. A pipeline that runs on every pull request, including ones from contributors who have no other access to the environment it deploys to. An infrastructure-as-code provider that was silently updated by a third party between two pipeline runs. None of these look dangerous in isolation. They look like normal engineering decisions. Together, they form a path to arbitrary code execution in production.
This talk presents an approach to modeling your CI/CD infrastructure as an attack graph. We model repositories, pipelines, deployment targets, identity groups, and external dependencies as nodes and edges in a queryable graph. Suddenly, you can ask questions that span system boundaries:
- "Who can modify this workflow, and what can that workflow reach?"
- "Which external dependencies, if compromised, would give an attacker a path to a production environment?"
- "How many steps separate a standard developer identity from production deployment access - and what are those steps?"
We walk through the graph schema, how we map identities across platforms, and what this reveals that traditional access reviews miss.