Dan Farmer's site on IPMI, the server management interface, documenting its widespread exposure and security flaws.
HN commenters cite it as a frightening reference on IPMI and BMC risks, noting that exposed management networks leave servers fully compromised.
I followed what you meant and understand baseband management controllers are more similar to IME in terms of risk profile. They're not (generally?) on-chip though, so it's at least up to individual OEMs, and purchasers have slightly more options (if slim). Agree it's not widely enough recognised either though. There's some coverage at least:
https://www.schneier.com/blog/archives/2013/01/the_eavesdroppi.html
https://isc.sans.edu/diary/IPMI%3A+Hacking+servers+that+are+turned+%22off%22/13399
http://arstechnica.com/security/2013/08/remote-admin-tool-imperils-servers/
https://www.usenix.org/system/files/conference/woot13/woot13-bonkoski_0.pdf
https://community.rapid7.com/community/metasploit/blog/2013/07/02/a-penetration-testers-guide-to-ipmi
https://www.schneier.com/blog/archives/2013/01/the_eavesdroppi.html
http://www.websecuritywatch.com/authentication-bypass-vulnerability-in-ipmi-2-0-rakp-through-the-use-of-cipher-zero/
http://fish2.com/ipmi/itrain.pdf
If I'm assuming everyone involved had good intentions, the best thing I can guess is someone was speaking to the writer about the potential of the BMC being used for spying and got some details mixed up.
Consider: the BMC has access to the system via PCI-e, as well as kvm and comport. In some systems, the BMC is in the path of the main NIC. There have been some major software flaws in BMC software, including revisions that SuperMicro shipped, where passwords could be bypassed in the network interface.
Stuff like this https://web.archive.org/web/20140625065505/http://blog.cari.net/carisirt-yet-another-bmc-vulnerability-and-some-added-extras/ and other things on this page http://fish2.com/ipmi/ are all pretty nasty if you thought IPMI was secure in the neighborhood of 2014.
Nowadays, it's not enough to wipe the machine. You have to toss (or, at least, re-image) the hardware.
Modern servers have firmware running on internal devices, including Linux runnning on full ARM cores, that can be 1) imaged from the base OS, and 2) have full control over hardware or other important data.
This includes:
- IPMI Modules: http://fish2.com/ipmi/itrain.html
- HDD Controllers (in your HDD/SSD drives): http://www.arm.com/markets/enterprise/hdd-ssd.php
... as well as ethernet controllers, any devices with PCI option ROMs (they can control boot), and quite a few other things.
This essentially means that you have a bunch of fixed, poorly documented, runtime reconfigurable and re-imageable hardware, with full access to data and system resources.
We are increasingly finding out that even owning and having physical control over the hardware isn't enough to insure secure a physical device.
Google 'BIOS based Advanced persistent threat' and 'Hardware Trojan'.
Oh but wait, it gets worse. Even "well thought out" industry standards are horrible security problems waiting to happen. Cypher 0 , who thought this was a good idea?
http://fish2.com/ipmi/
The solution you are looking for may be known as homomorphic encryption, which is still rather far out in the theoretical place.
The horse has been out of the barn so long it has had a good life, grown old and has great grand children playing in the fields.
if you want to be really scared read:
http://fish2.com/ipmi/itrain.pdf
all servers in a datacenter have this management interface (iLO is just one type).
if the management network these sit on is poorly secured (like here) your servers are literally powned.
It's worse than that. It's not BIOS-- it's a freestanding computer.
You'll enjoy this (or be horrified by it): http://fish2.com/ipmi/itrain.html