Is it possible to use self-signed user-side certificates in mutual TLS to stop rogue CA MITM attacks?

After seeing Apalrd’s guide on mTLS I set up caddy with a self-signed root CA and made a few certificates to test it out. It’s been working great if I’m trying to access something from a browser or if the app supports it, but that’s a different topic.

What I’m wondering is if it would be possible to use the fact that I have a self signed root CA with certificates that I have distributed to my or my friends’ computers to stop something like the eMudhra leak or a rogue CA from being able to MITM my websites.

The webserver/reverse proxy which is caddy in my case, could have both a “legit” TLS cert and a self-signed one from my CA. The client could then automatically require that all connections to the domains that it has certificates for should have TLS and the chain of trust should be the one that I control.

Basically if the common name for the client is [email protected] and Alice tries to connect to anything example.com her computer should demand a certificate that is signed by the same authority that signed her user-side cert.

I have basically no idea if any browser or web server supports anything close to this, but this is why I’m asking here.

SSL stripping can still occur on sites using self-signed certificates if the connection starts as HTTP or is downgraded to HTTP, because the attack prevents the HTTPS upgrade rather than exploiting the certificate itself.