Podman is an open source CLI tool for working with containers, pods and container images. It uses a Linux kernel feature called SECCOMP to filter syscalls made by the processes inside the container. This allows Podman to reduce the attack surface of the kernel which is exposed to the container. Podman ships with a basic SECCOMP profile, which blocks 44 syscalls, but the default profile is very loose. In this talk, I want to go through the technical details of how we implemented the feature which can generate a SECCOMP profile for a specific workload to tighten the security of the kernel by tracing/recording the syscalls of a given container and the set of syscalls is then used to create a new profile to reduce the attack surface. I would talk about the lessons learned while building the feature, all the problems encountered and their solutions.
A video projector and possibly an internet connection
This seems interesting but I'm not sure of the audience. Is it really just "general container interest", or would this be useful for anyone interested in, say, updating their packages' systemd configuration?
Yes, anybody who is interested in containers, how they work, know about SECCOMP, etc would be benefited.
The talk selection committee feels like this would be better at a developer-focused conference. Fedora contributors and package maintainers are unlikely to get direct benefit here.
Metadata Update from @mattdm: - Issue close_status updated to: Talk Not Scheduled - Issue status updated to: Closed (was: Open)
Thanks for the consideration :D