4 ms·
I don't see how getting rid of the ME helps. If you don't trust Intel then you don't trust Intel. They can backdoor the main CPU as easily as they can backdoor
by mrob 10y ago
I don't see how getting rid of the ME helps. If you don't trust Intel then you don't trust Intel. They can backdoor the main CPU as easily as they can backdoor the ME. The only true solution is an Open Source Hardware CPU, and some means of verifying that the hardware matches the HDL.
There are already Open Source CPU designs, but the verification is more difficult. Even if you use a big enough process node that you can decap it and inspect it optically (eg. using similar techniques as used with http://visual6502.org/ http://visual6502.org/ ), there is still the possibility of dopant-level backdoors which are much harder to detect:
http://sharps.org/wp-content/uploads/BECKER-CHES.pdf http://sharps.org/wp-content/uploads/BECKER-CHES.pdf
- dandelion_lover 10y ago>If you don't trust Intel then you don't trust Intel. Security is not a boolean. It's a price to own you. By removing Intel ME you increase this price.
- dlmetcalf 10y agoRISC-V architectures hold the most promise there AFAIAA. BSD-licensed, cutting-edge architecture and it's got the right backers. FreeBSD is already ported to it. There'll be a range silicon implementations. Cambridge's lowRISC is likely to be the first SoC, with design completing this year.
- ashitlerferad 10y agoThe backers will probably be making proprietary implementations with ME-style backdoors.
- dlmetcalf 10y agoNo doubt some of the bigger ones probably will. When it's BSD (not bound by restrictive commercial licensing agreements) though, it massively lowers the barriers to entry for manufacturers who will offer security. Anyway, here's an entirely open source one you can inspect: http://www.lowrisc.org/ http://www.lowrisc.org/. Put it on an FPGA if you want.
- pascal_cuoq 10y ago> I don't see how getting rid of the ME helps. If you don't trust Intel then you don't trust Intel. As dandelion_lover says, security is not boolean. I may or I may not trust Intel to backdoor the processor I run code of my choice on in more or less visible ways (the more visible, the more risk of public opinion backlash if it is revealed what Intel did), but I absolutely do not trust Intel and its business partners to design an independent subsystem that runs their code and is not vulnerable to Intel (or any rogue employee thereof) + Intel's business partners (or any rogue employee thereof) + US federal entities (or any rogue employee thereof) that can coerce any large or small US company. I am not suspecting Intel of ill intent. I am suspecting Intel and its partners of incompetence. There is a big difference. Intel can not actually “backdoor the main CPU as easily as” they can leave the ME open to attacks. The second one is so easy that you might as well assume it is the case until convincing arguments have shown otherwise.
- edwintorok 10y agoTrusting Intel on producing a CPU, and trusting Intel with ME are completely different levels of trust. If I could run my own software on the Intel ME, I wouldn't mind. In some sense having Intel ME is worse than having Windows as your OS: at least you can apply security patches for Windows, and should you choose to do so you can replace it with your favourite OS. However Intel ME is entirely closed source and locked down: probably running out-of-date software with several unpatched vulnerabilities with no way for the user to inspect it, patch it, turn it off with a jumper or replace it.
- yuhong 10y agoAFAIK they can patch the Intel ME firmware and often do with BIOS updates. I am not sure about AMD PSP but that don't have access to the network so patching it is less important.
- killerstorm 10y agoThe problem with ME is that enables a number of additional attacks, even if it's not backdoored.