Hi Coins,
I'm not coreboot, but I'm a part of it, so I will try to answer your question. CCing the coreboot mailing list for more input, as I can only assume that that list was the intended recipient for your email.
It is unproven that Intel deliberately builds in backdoors into their CPUs. However, a lot of their software / hardware designs create a rather large attack surface that could be exploited, if someone puts the right amount of resources on the problem.
This attack surface lives
- in the SOC's converged security management engine (CSME / ME), which in some SKUs enables remote access to the system through builtin network interfaces. The CSME cannot be fully disabled, but some security issues can be mitigated in a good hardware software design i.e. by using the non-enterprise (aka 1.5M SKU) of the ME firmware or by not using the SOC associated network interfaces (questionable) or by disabling as many CSME features as possible. CSME is particularly problematic because it can access main memory, so a remote attack could steal your private keys, rendering your cryptographical secrets useless.
- FSP / BLOBS. Closed source firmware pieces generally have the problem that they are impossible to audit. Even if there are fixed version out in the field, you can not tell from a binary what is fixed or not. Bugs are also impossible to fix, even when known. Imaginable attack scenarios could also be deliberate changes to memory training data which open known but fixed memory controller issues.
Generally coreboot tries to enable the user / developer / systembuilder to address as many of these concerns as possible, but it can not 100% fix them at this point. If you are concerned about your hardware architecture, please study the source code of coreboot and the available open documentation on x86 hardware (of which there is a fair amount) and help us audit our code.
Stefan
* Coins coins@cryptolab.net [190331 18:29]:
Dear Coreboot,
As far as I know, Intel puts proprietary backdoors in any recent CPU they develop.
How does this affect the security of a PC/laptop with coreboot installed when it is using such a processor?
Best regards,
Coins
To add to this - in my opinion there is no reason to believe that Coreboot is less secure than a proprietary BIOS built on top of the code distributed by Intel to established BIOS vendors (AMI, Phoenix/Award etc). The FSPs are also built out of that same code and because Coreboot is open-source, it at least can be audited. As Stefan points out CSME/ME is the main source of worry for security-conscious, and from this standpoint there is no difference between coreboot and the proprietary BIOS.
Best regards Alex
________________________________________ From: Stefan Reinauer stefan.reinauer@coreboot.org Sent: Monday, April 1, 2019 5:53 PM To: Coins Cc: coreboot@coreboot.org Subject: [coreboot] Re: Question regarding 7th generation Intel CPUs
Hi Coins,
I'm not coreboot, but I'm a part of it, so I will try to answer your question. CCing the coreboot mailing list for more input, as I can only assume that that list was the intended recipient for your email.
It is unproven that Intel deliberately builds in backdoors into their CPUs. However, a lot of their software / hardware designs create a rather large attack surface that could be exploited, if someone puts the right amount of resources on the problem.
This attack surface lives
- in the SOC's converged security management engine (CSME / ME), which in some SKUs enables remote access to the system through builtin network interfaces. The CSME cannot be fully disabled, but some security issues can be mitigated in a good hardware software design i.e. by using the non-enterprise (aka 1.5M SKU) of the ME firmware or by not using the SOC associated network interfaces (questionable) or by disabling as many CSME features as possible. CSME is particularly problematic because it can access main memory, so a remote attack could steal your private keys, rendering your cryptographical secrets useless.
- FSP / BLOBS. Closed source firmware pieces generally have the problem that they are impossible to audit. Even if there are fixed version out in the field, you can not tell from a binary what is fixed or not. Bugs are also impossible to fix, even when known. Imaginable attack scenarios could also be deliberate changes to memory training data which open known but fixed memory controller issues.
Generally coreboot tries to enable the user / developer / systembuilder to address as many of these concerns as possible, but it can not 100% fix them at this point. If you are concerned about your hardware architecture, please study the source code of coreboot and the available open documentation on x86 hardware (of which there is a fair amount) and help us audit our code.
Stefan
* Coins coins@cryptolab.net [190331 18:29]:
Dear Coreboot,
As far as I know, Intel puts proprietary backdoors in any recent CPU they develop.
How does this affect the security of a PC/laptop with coreboot installed when it is using such a processor?
Best regards,
Coins
_______________________________________________ coreboot mailing list -- coreboot@coreboot.org To unsubscribe send an email to coreboot-leave@coreboot.org
Dear Stefan, Alex,
Thank you for your response and information provided. Indeed open source BIOS would not be less secure than proprietary BIOS, even quite the opposite is what I would expect.
Basically what I am wondering is, are there any security benefits from using older generation (perhaps 3rd gen) Intel CPUs rather than newer hardware?
Currently I am using old hardware but I do miss having the latest performance capabilities at disposal. Any thoughts on this?
Kind regards,
Coins
On 4/2/19 5:03 AM, Alex Feinman wrote:
To add to this - in my opinion there is no reason to believe that Coreboot is less secure than a proprietary BIOS built on top of the code distributed by Intel to established BIOS vendors (AMI, Phoenix/Award etc). The FSPs are also built out of that same code and because Coreboot is open-source, it at least can be audited. As Stefan points out CSME/ME is the main source of worry for security-conscious, and from this standpoint there is no difference between coreboot and the proprietary BIOS.
Best regards Alex
From: Stefan Reinauer stefan.reinauer@coreboot.org Sent: Monday, April 1, 2019 5:53 PM To: Coins Cc: coreboot@coreboot.org Subject: [coreboot] Re: Question regarding 7th generation Intel CPUs
Hi Coins,
I'm not coreboot, but I'm a part of it, so I will try to answer your question. CCing the coreboot mailing list for more input, as I can only assume that that list was the intended recipient for your email.
It is unproven that Intel deliberately builds in backdoors into their CPUs. However, a lot of their software / hardware designs create a rather large attack surface that could be exploited, if someone puts the right amount of resources on the problem.
This attack surface lives
in the SOC's converged security management engine (CSME / ME), which in some SKUs enables remote access to the system through builtin network interfaces. The CSME cannot be fully disabled, but some security issues can be mitigated in a good hardware software design i.e. by using the non-enterprise (aka 1.5M SKU) of the ME firmware or by not using the SOC associated network interfaces (questionable) or by disabling as many CSME features as possible. CSME is particularly problematic because it can access main memory, so a remote attack could steal your private keys, rendering your cryptographical secrets useless.
FSP / BLOBS. Closed source firmware pieces generally have the problem that they are impossible to audit. Even if there are fixed version out in the field, you can not tell from a binary what is fixed or not. Bugs are also impossible to fix, even when known. Imaginable attack scenarios could also be deliberate changes to memory training data which open known but fixed memory controller issues.
Generally coreboot tries to enable the user / developer / systembuilder to address as many of these concerns as possible, but it can not 100% fix them at this point. If you are concerned about your hardware architecture, please study the source code of coreboot and the available open documentation on x86 hardware (of which there is a fair amount) and help us audit our code.
Stefan
- Coins coins@cryptolab.net [190331 18:29]:
Dear Coreboot,
As far as I know, Intel puts proprietary backdoors in any recent CPU they develop.
How does this affect the security of a PC/laptop with coreboot installed when it is using such a processor?
Best regards,
Coins
coreboot mailing list -- coreboot@coreboot.org To unsubscribe send an email to coreboot-leave@coreboot.org
Coins wrote on 4/3/19 3:50 PM:
Dear Stefan, Alex,
Thank you for your response and information provided. Indeed open source BIOS would not be less secure than proprietary BIOS, even quite the opposite is what I would expect.
Basically what I am wondering is, are there any security benefits from using older generation (perhaps 3rd gen) Intel CPUs rather than newer hardware?
Currently I am using old hardware but I do miss having the latest performance capabilities at disposal. Any thoughts on this?
Intel ME can be removed more thoroughly from older generations; see https://github.com/corna/me_cleaner. Security benefits of doing so depend on your threat model.
On Wed, Apr 3, 2019 at 8:55 AM Coins coins@cryptolab.net wrote:
Currently I am using old hardware but I do miss having the latest performance capabilities at disposal. Any thoughts on this?
We see this a lot - People want performance of modern CPUs but don't trust the platform. That's totally understandable but is sadly unavoidable in the x86 world. Even though coreboot easily provides the most transparency you'll get in x86 host firmware these days, it's still just one component that you need to worry about. ME cleaner is great but has obvious limitations to what it can remove.
If you're really serious then you need to divorce x86 and try an alternative architecture such as POWER9. Raptor (https://www.raptorcs.com/) sells reasonably affordable systems with Xeon-class performance, PCIe Gen 4.0, and 100% open source firmware for both host and BMC. Freedom isn't free, though, and if that's too expensive you might find a cheap used POWER8 system (not as open as POWER9, but still good compared to x86 systems in terms of openness): https://www.phoronix.com/scan.php?page=article&item=tyan-power8-server&a... .
-
Alex Feinman -
awokd -
Coins -
David Hendricks -
Stefan Reinauer