The process of proving that a key originates from a TPM is known as:
- For TPM 2.0: Credential activation, enforced with
TPM2_ActivateCredential
- For TPM 1.2: Identity activation, enforced with
TPM_ActivateIdentity
This technique accomplishes many things, but one of them is proving that the key generated after issuing a request to the TPM actually originates from a trusted TPM and was not spoofed. For TPM 1.2, since that's what the question is about, identity activation is an 8-step process that goes like this (what follows is an excerpt from TCG's AIK Certificate Enrollment):
The crucial part here is the next-to-last sentence:
This command decrypts the (encrypted) symmetric key K2 from the ACA
using the EK-private-key (which resides only in the TPM) after
ensuring an AIK with a matching pub key resides in the TPM.
It is enforced by the spec that the EK will not decrypt the TPM_EK_BLOB
object unless the private key that the activation is requested for is found within the TPM. And because the object was encrypted by the TSS without using the TPM secrets, and you have already verified the EK public key up the manufacturer's CA cert chain, it is ensured that the key that identity activation is requested for has originated within a TPM manufactured by a trusted entity.