I am being asked to run some proprietary software for a period of time in a Docker container on a university HPC cluster where I cannot guarantee that no-one else has access to the hard drive or memory. The software takes around two hours to process information, and I will be running about 200 instances on the cluster. I plan on: encrypting the codebase inside the container and unencrypting it at run via a key got from a remote http request (which will be disabled when the container is not supposed to be running) compiling the python files in the codebase Can you suggest any further ideas? I am aware that it is not possible to make this fully secure, but I would like to make it hard for someone to be able to access the codebase. Continue reading...
