While browsing privacy forums, I often see a lot discussions regarding laptop hardware security, especially in regards to resistance against physical attacks. Unfortunately, most of the advice given on these forums are incorrect, recommending people to buy overpriced products that are significantly less secure than standard laptops.
In this post, I will walk you through a quick overview of how hardware security is generally implemented for laptops, what to look for, and what to avoid. We will not discuss MacBooks or Chromebooks, as they are vastly different from normal x86 Windows/Linux hardware.
To start off, the best laptops I have found are modern the Dell Latitude/Precision laptops with an Intel vPro Enterprise CPU. The second best group of laptops I have found are modern Lenovo Thinkpads with Intel vPro Enterprise or AMD Ryzen Pro CPUs. These are relatively easy to acquire and share these common security properties:
- Have Intel Boot Guard or AMD Platform Secure Boot to protect the firmware
- Have regular firmware updates (Monthly updates for Dell, and Bi-monthly updates for Thinkpads)
- Support firmware updates via the [Linux Vendor Firmware Service](https://www.fwupd.org/)
- Support custom Secure Boot key enrollment
- Support disabling Microsoft third-party certificate authoritity
- Support memory encryption
- Support DRTM technologies
- Meet Secure-cored requirements for Windows
- Meet HSI-4 for Linux
- Still recieve Microcode updates from Intel and AMD
## Boot Security
For boot security, there are 2 different technologies: Static Root of Trust Measurement (SRTM) and Dynamic Root of Trust Measurement (DRTM). I will primarily discuss SRTM here, but I will quickly go over DRTM as well.
### SRTM
In general, SRTM works as follows:
- The public key for the firmware by the OEM is fused into an immutable Root of Trust. With Intel Boot Guard, this will be in the chipset (PCH). With AMD Platform Secure Boot, this will be in the CPU.
- The CPU ensures that a significant portion of firmware from the EEPROM (BIOS chip) is signed by the OEM. This includes the Boot Block.
- The Boot Block performs measurements of the firmware and its settings and submit those measurements to PCR 0.
- The next stages measures things like the firmware configuration, boot loader, partition table, Secure Boot state, and so on. These are submitted to PCR 2,3,5,7, and so on.
- UEFI Secure Boot validates the signature of the bootloader against the Secure Boot database.
- The boot loader submits additional measurements to other PCRs. In the case of GRUB, it will use PCR 8 or 9 for instance.
After the user has set up disk encryption, they can seal a encryption key protector to the TPM against certain PCRs, and optionally protect it with a PIN.
Should an attacker attempt to flash malicious firmware onto the EEPROM, the laptop will refuse to boot, and will throw a Boot Guard / Platform Secure Boot violation. Should they try to downgrade the firmware version, change firmware settings or disable UEFI Secure Boot, PCR 0, 1, and 7 will change respectively. The TPM will not recieve the correct measurements and will not release the encryption key protector.
This is not perfect, and the user still needs to set up additional protection for the boot stages after the boot loader, but it is an okay start and is significantly better than how some vendors try to do it as we will discuss down below.
In general, DRTM works by loading in an ACM binary signed by the CPU vendor along what the system is trying to boot. The ACM binary will perform measurements of the environment and submit the results to certain PCRs. In the case of Intel TXT, PCR 17 and 18 are used. The measurements then can be used to for remote attestation or to release a secret stored in the TPM if it matches the TPM policy.
On Windows, DRTM is implemented with [System Guard](https://learn.microsoft.com/en-us/windows/security/hardware-security/how-hardware-based-root-of-trust-helps-protect-windows) for remote attestation. On Linux, DRTM is not widely used yet, but [TrenchBoot](https://trenchboot.org/) is being developed to address that.
It is important to note that DRTM technologies can be bypassed via the System Management Mode (SMM), so you still need to have some level of trust in your firmware to implement SMM mitigations. In essence, you still need to rely on SRTM to some extent.
A very common misinformation among privacy communities is that the Intel Management Engine (ME), its sucessor - Intel Converged Security and Management Engine (CSME), and AMD Platform Security Processor (PSP) are some sort of evil backdoor. Some may go so far as to tell the user to "disable the ME", either out of [paranoia](https://www.whonix.org/wiki/System_Hardening_Checklist#Disable_Intel_ME_Functionality) or for ["attack surface reduction"](https://discuss.privacyguides.net/t/linux-laptops-system76-other-options/16557/19):
- Intel Locker (A nice vPro Enterprise mechanism to purge the encryption key from memory after early boot - not widely used on Linux yet, but is implemented on ChromeOS)
By disabling Intel CSME, you are **increasing the attack surface** by crippling security features. Additionally, if you buy hardware so old that you can run `me_cleaner` to disable the ME yourself, it means that these hardware do not have Boot Guard and it is impossible to implement any kind of boot security.
This excercise also achieves nothing to protect against a hypothetical scenario where Intel and AMD are malicious. Intel and AMD do not need the co-processor to implement a backdoor - they can simply introduce CPU vulnerabilities like Spectre and Meltdown if they want to. If you do not trust a CPU vendor, the only mitigation is to not use said vendor.
Another misinformation regarding CSME is that it is provides some kind of [shady "remote management" system](https://www.fsf.org/blogs/community/active-management-technology) for your computer. In reality, this is the AMT component which only exists on Intel vPro CPUs. It is meant for IT teams to manage systems with technologies like Serial over LAN, Solarwind, etc.
This is not something hidden, people have accidentally [run into it](https://mastodon.lilysthings.org/@i_lost_my_bagel/112228352384742242) on social media.
For attack surface reduction, you should absolutely disable it. With that said, don't let the scary claims about "remote management" by the Free Software Foundation spook you - if some sort of hypothetical backdoor actually implemented this way, it is not hard to detect. There are better ways to implement a backdoor as discussed above, and if you don't trust the CPU vendor you should avoid them as a whole, not just the vPro model.
Some people recommend buying AMD instead of Intel to avoid the possibility of having Intel AMT. However, they also miss a very simple fact that AMD has an equivalent technology for their Ryzen Pro CPU - AMD DASH.
Another false claim regarding Secure Boot by the Free Software Foundation is that UEFI Secure Boot is somehow Microsoft's evil attempt to lock users out of their computer by [only allowing it to run Microsoft approved software](https://www.fsf.org/campaigns/secure-boot-vs-restricted-boot/whitepaper-web):
In reality, most if not all laptops with UEFI Secure Boot allows you to disable it - you can run whichever operating system you want. While it is true that certain lines of laptops like Razer do not allow custom key enrollment, proper business laptops like Dell Latitude/Precision and Lenovo Thinkpad do. You can enroll your own Secure Boot key and tell your laptop to boot only the system you trust.
Another benefit of laptops with Microsoft's Secure-cored certification is that you can have the **Freedom** to disable the Microsoft Secure Boot Third-Party Certificate Authority and still have the laptop function normally. This is especially handy if you plan to run Windows as your operating system.
Heads is built around the desire for the signing key for the firmware to be in the posession of the user instead of the OEM. They also want the user to be able to change the key in case they lose access to it. In general, this is how Heads works:
The problem with this design is that everything hinges on the boot block doing its initial measurements truthfully. Since the user is supposed be able to change the key, no key can be set up with Boot Guard to protect the boot block. The result is that there is nothing stopping an attacker from flashing a piece of malicious firmware that will just lie about the measurements. This is anagolous to asking a potential liar of they are being truthful. At best, Heads can protect against tampering of the disk, but it cannot protect against firmware tampering.
