That was my initial understanding, but I figured I'd give the parent commenter the benefit of the doubt in case my understanding was flawed.

Some more searching seems to lead back to where I started: the microcode itself isn't phoning home, the only phoning "home" that occurs is the normal process of "when you pull down an update for something, it requires phoning to the repository holding the updates".

How could it possibly patch bugs without phoning home? Are you claiming that it is self modifying code?

I wouldn't consider "running an update where it pulls new code" to be "phoning home", any more than my car is "phoning home" when I drive it to the dealership for repairs.

The implication the comment I was replying to gave was that the device sent unexpected network traffic back to Intel HQ, with the connotation that it was doing so to leak information about my system.

"Running an update where it pulls new code" is definitely an example of phoning home, and it opens the door to all sorts of vulnerabilities, such as the one featured in the Apple vs the FBI case.

Intel microcode updates on Linux are provided through regular distribution repositories. There is no phoning home feature.

Do they get updated when you run the package manager, or do they update automatically?

Binary packaging systems download microcode.dat (a text blob containing microcode) from Intel during the build process. Microcode.dat gets converted into an initramfs image that supplies the new microcode early in the system boot. Users download the built binary package, which then modifies the boot parameters of their system.

So normally only the build servers contact Intel. Source-based distributions may. But even then the source files are verified with cryptographic hashes (which are in turn signed by the maintainers' private keys).

Edit: I may have misread your question. I thought you were asking by implication if the microcode blob is pulled after the install. If not, then: No. Updates happen when the package manager decides.

https://www.kernel.org/doc/Documentation/x86/early-microcode...

https://gitlab.com/iucode-tool/iucode-tool

https://downloadcenter.intel.com/download/26400/Linux-Proces...

Either via the package manager or via the user manually downloading and loading the microcode update of their choosing. For example: http://askubuntu.com/questions/545925/how-to-update-intel-mi...

A more subtle way to communicate with the mother ship would be to ship an update that makes the CPU leak some desired information. For example change how some specific cryptographic operation works under certain conditions to make it possible for eavesdropper to break the encryption. Or just make some funny things with memory and make certain information available to be retrieved via web browser when user visits malicious page.

ME does have ability to do that, tho.

Do you think you could fit drivers for all common ethernet/wifi cards in there and proper TCP/IP implementation? It's below assembly abstraction level.

Intel ME has independent NIC access and usually cannot be disabled by the user.

It has only been six days since it has become possible to neutralize the ME firmware for Sandy Bridge and Ivy Bridge[1].

[1]: https://news.ycombinator.com/item?id=13056997

The ME is basically an independent universal computer in its own right, it comes with its own clock, RAM, CPU etc...

It is like a Matryoshka doll sitting inside the Intel CPU of your computer. Therefore, yes, it can contain all of that. For further details see my other two posts.

Sorry I was just thinking about microcodes and you are talking about ME. Scary stuff indeed[1]

1. https://en.wikipedia.org/wiki/Intel_Active_Management_Techno...

How else would you implement remote server management that works even when the hardware is shut down (put still connected to power) if not with a seperate CPU that has access to the NIC?

Just like IPMI, or any other standard for allowing hardware-level admin of a system.

By using the word 'standard' you're trying to legitimize something that's fully encrypted, impossible to code-audit and fully under control of Intel, while pretending to be your property and controlled by you.

So yes, it's a standard. I believe there was also a standard on how to tie a noose for hanging a human, but that didn't mean much to the one getting hanged.