[moonchild]

No SSL => MITMer can definitely read your traffic trivially.

Broken SSL => MITMer can possibly negotiate insecure and read your traffic anyway. MITMer can also possibly cause a denial-of-service, or get arbitrary code execution on that one chip that controls your entire CPU.

If I had to choose, I would take the first option.

(This precludes options like removing the IME entirely, or updating it to a version with non-broken SSL.)

[rckoepke]

I'm coming from a place of good faith here so bear with me. My understanding is that any vulnerability here would also exist in accessing any HTTPS website. I'm assuming you wouldn't choose to browse the web without SSL/TLS, so I'm assuming the difference here is that it's the CPU management chip instead of your browser?

I suppose that if you broke SSL/TLS you could commandeer arbitrary AWS/GCP/Azure instances.

For that matter, do you trust SSL/TLS significantly less than SSH?

I guess I'm still having trouble wrapping my head around the idea of not using SSL/TLS.

[wizzwizz4]

My browser is sandboxed. The worst it can do is ransomware my files – and the Tor Browser can't even do that thanks to the AppArmor rules. (If I set my machine up properly, it wouldn't even be able to ransomware my files.)

The CPU management chip can ransomware my files while the computer is "off", corrupt my backups as I load them, steal my passwords, steal my bank details, dynamically modify the traffic to make it look like my bank balance hasn't gone down…

[progval]

Nitpick: Remote code execution is breaking a TLS implementation (eg. openssl), not breaking TLS itself.

> so I'm assuming the difference here is that it's the CPU management chip instead of your browser?

Yes.

If a vulnerability is found in a SSL/TLS implementation, it can be fixed with a software update.

I don't know how Intel ME works, but I'm guessing updating it is harder to update than a browser.

> For that matter, do you trust SSL/TLS significantly less than SSH?

I'm not the GP, but I'm guessing they do. TLS solves a more complex problem than SSH, as SSH assumes the user validates a server's public key manually (even though they usually don't, but TOFU [1] makes it somewhat harder to exploit), whereas a TLS server's key can change at any time.

[1] https://en.wikipedia.org/wiki/Trust_on_first_use

[moonchild]

There is no specific TLS flaw. The TLS spec is very complicated, so it's difficult to make a library that implements it without bugs. Insofar as TLS implementations have bugs, the TLS implementation in by browser can be updated to fix those bugs. The TLS implementation in my IME cannot.