network security of the microvm (sandbox for egress domain names/IPs)

[nnWhisperer]

Hello,
Assuming that the microvm can run any code, is there a way to limit the domains it can connect to? Pi-hole dns filtering comes to my mind, but simple dns filtering isn't guaranteed, as an app may have an IP to connect to be embedded inside. Hence, on the use-case that comes to my mind, ideally there must be a dns server that replies correctly only for the whitelisted domains and then control the iptables to allow access to those IP addresses only. There is no such tool that I know, but may be anyone knows.

[xmarcalx]

Hi @nnWhisperer ,

Thank you very much for the inquiry.

I am not sure i fully grasped the details of your question.
Firecracker microVM scan run "any code" so that is a good assumption. With this, I think you mean that a Firecracker microVM is a simple Linux environment which can use a network interface. As such, it behaves like a normal computer connected to a network. According your threat model (we usually consider guest untrusted so i would suggest to apply any filter technology outside the microVM), you can limit the connectivity of this entity in the network using any already existing network methodology (e.g. firewalls, proxies, etc. etc.)

[nnWhisperer]

Thanks. As I'm aware that this isn't in the scope of firecracker, I didn't ask it under issues. Do you know a solution for this, to limit the accessible domains from the microVMs?

[xmarcalx]

Hi @nnWhisperer ,

got it.
Nope i do not know any specific solution, sorry. Maybe a way to implement is to have a proxy on the host (so you avoid round trips for the VMs outside the host) with the allowed dns page cached and then a firewall with whitelist rules which constrain the host and the proxy to fetch again from internet whatever you want to allow.