Hey everyone 👋 I'm trying to configure Fleet MDM for SOC 2 compliance. It's important that the agent that's installed on users' devices has minimal capabilities, and that it can't possible be abused, even in theory. Is there any way to restrict the agent's capabilities to a handful of queries? I've been reading the relevant fleet and osquery docs, but it's not clear to me if the agent executable is actually restricted. If this is the wrong route, do you have any other recommendations for restricting the agent?

OSQuery itself is read only https://blog.trailofbits.com/2018/04/10/what-do-you-wish-osquery-could-do/ I am amused at the thought of "can't be abused, even in theory" because in reality everything can

Thanks for sharing that link. I’m specifically interested in an agent that can only run a handful of pre-defined queries. Even if it’s read-only, I would like to have an agent is only able to run the pre-defined set of queries it’s been compiled with. Hence “can’t be abused” — an admin with malicious intent would only be able to run those queries, but nothing more. Is such a thing possible with Fleet / osquery?

If you truly want that then I would say you should deploy the agent with packs configured to write results to a place readable by your SIEM and don't connect the osquery agent to Fleet at all

Thanks! Those flags could be a good option