Intel® Trusted Execution Technology (Intel® TXT)
Intel TXT has been shipping since 2002 in client machines and since 2010 in servers. Intel TXT provides a chain of trust that is rooted in the microprocessor's hardware and is extended in stages to the OS and even to applications, depending on how higher levels of software make use of it.
This section describes Intel TXT at a high level first, including its features that offer advantages over a TPM-only solution, and then delves into the details of how it uses TPM 2.0's capabilities. At a high level, the advantages of Intel TXT over a TPM-only solution are a hardware-based root of trust, a smaller TCB, and specific checks of the hardware and software configuration performed by the ACMs. This section highlights how these advantages are implemented.
Other Intel technologies use TPMs, including Intel Boot Guard. This chapter doesn't describe these technologies or how they use TPM 2.0 devices, because Intel TXT is currently the most prevalent technology and a representative example of how TPM 2.0 devices are used. Also note that there are two flavors of Intel TXT: one for client platforms and one for server platforms. Many of the principles of operation are shared, but we focus on the server version, because it uses a superset of TPM functionality.
Intel TXT for servers can defend against BIOS attacks, reset attacks, rootkits, and software attacks and allows the system integrator and user many options for configuring the level of protection. Although it does prevent or mitigate some attacks, its primary purpose is to notify the user and system software of the presence of a possible attack and prevent a verified launch if an attack is detected. Intel TXT hardware and software and the TPM are tightly integrated in a way that protects both the TPM and the TXT registers from unauthorized access. Critical measurements stored in the TPM cannot be spoofed, and the TPM protects OEM and user policies from unauthorized alteration.
How does it do this? A short description is that a chain of trust is extended from the Intel processor and/or chipset hardware through the BIOS. Then, after the OS has booted, if the user desires to enter secure mode at the OS level, a measured launch sequence is initiated by the OS or a software program running on top of the OS (DRTM). This measured launch ensures that there are no security holes in the system before launching the OS and entering secure mode. Basically, a chain of trust may be extended from the hardware all the way up to the highest levels of software, enabling a system administrator or user to create and use security policies. This chain of trust always measures components before actually executing them.