| 1 |
.. SPDX-License-Identifier: GPL-2.0-only |
| 2 |
|
| 3 |
TPM Security |
| 4 |
============ |
| 5 |
|
| 6 |
The object of this document is to describe how we make the kernel's |
| 7 |
use of the TPM reasonably robust in the face of external snooping and |
| 8 |
packet alteration attacks (called passive and active interposer attack |
| 9 |
in the literature). The current security document is for TPM 2.0. |
| 10 |
|
| 11 |
Introduction |
| 12 |
------------ |
| 13 |
|
| 14 |
The TPM is usually a discrete chip attached to a PC via some type of |
| 15 |
low bandwidth bus. There are exceptions to this such as the Intel |
| 16 |
PTT, which is a software TPM running inside a software environment |
| 17 |
close to the CPU, which are subject to different attacks, but right at |
| 18 |
the moment, most hardened security environments require a discrete |
| 19 |
hardware TPM, which is the use case discussed here. |
| 20 |
|
| 21 |
Snooping and Alteration Attacks against the bus |
| 22 |
----------------------------------------------- |
| 23 |
|
| 24 |
The current state of the art for snooping the `TPM Genie`_ hardware |
| 25 |
interposer which is a simple external device that can be installed in |
| 26 |
a couple of seconds on any system or laptop. Recently attacks were |
| 27 |
successfully demonstrated against the `Windows Bitlocker TPM`_ system. |
| 28 |
Most recently the same `attack against TPM based Linux disk |
| 29 |
encryption`_ schemes. The next phase of research seems to be hacking |
| 30 |
existing devices on the bus to act as interposers, so the fact that |
| 31 |
the attacker requires physical access for a few seconds might |
| 32 |
evaporate. However, the goal of this document is to protect TPM |
| 33 |
secrets and integrity as far as we are able in this environment and to |
| 34 |
try to insure that if we can't prevent the attack then at least we can |
| 35 |
detect it. |
| 36 |
|
| 37 |
Unfortunately, most of the TPM functionality, including the hardware |
| 38 |
reset capability can be controlled by an attacker who has access to |
| 39 |
the bus, so we'll discuss some of the disruption possibilities below. |
| 40 |
|
| 41 |
Measurement (PCR) Integrity |
| 42 |
--------------------------- |
| 43 |
|
| 44 |
Since the attacker can send their own commands to the TPM, they can |
| 45 |
send arbitrary PCR extends and thus disrupt the measurement system, |
| 46 |
which would be an annoying denial of service attack. However, there |
| 47 |
are two, more serious, classes of attack aimed at entities sealed to |
| 48 |
trust measurements. |
| 49 |
|
| 50 |
1. The attacker could intercept all PCR extends coming from the system |
| 51 |
and completely substitute their own values, producing a replay of |
| 52 |
an untampered state that would cause PCR measurements to attest to |
| 53 |
a trusted state and release secrets |
| 54 |
|
| 55 |
2. At some point in time the attacker could reset the TPM, clearing |
| 56 |
the PCRs and then send down their own measurements which would |
| 57 |
effectively overwrite the boot time measurements the TPM has |
| 58 |
already done. |
| 59 |
|
| 60 |
The first can be thwarted by always doing HMAC protection of the PCR |
| 61 |
extend and read command meaning measurement values cannot be |
| 62 |
substituted without producing a detectable HMAC failure in the |
| 63 |
response. However, the second can only really be detected by relying |
| 64 |
on some sort of mechanism for protection which would change over TPM |
| 65 |
reset. |
| 66 |
|
| 67 |
Secrets Guarding |
| 68 |
---------------- |
| 69 |
|
| 70 |
Certain information passing in and out of the TPM, such as key sealing |
| 71 |
and private key import and random number generation, is vulnerable to |
| 72 |
interception which HMAC protection alone cannot protect against, so |
| 73 |
for these types of command we must also employ request and response |
| 74 |
encryption to prevent the loss of secret information. |
| 75 |
|
| 76 |
Establishing Initial Trust with the TPM |
| 77 |
--------------------------------------- |
| 78 |
|
| 79 |
In order to provide security from the beginning, an initial shared or |
| 80 |
asymmetric secret must be established which must also be unknown to |
| 81 |
the attacker. The most obvious avenues for this are the endorsement |
| 82 |
and storage seeds, which can be used to derive asymmetric keys. |
| 83 |
However, using these keys is difficult because the only way to pass |
| 84 |
them into the kernel would be on the command line, which requires |
| 85 |
extensive support in the boot system, and there's no guarantee that |
| 86 |
either hierarchy would not have some type of authorization. |
| 87 |
|
| 88 |
The mechanism chosen for the Linux Kernel is to derive the primary |
| 89 |
elliptic curve key from the null seed using the standard storage seed |
| 90 |
parameters. The null seed has two advantages: firstly the hierarchy |
| 91 |
physically cannot have an authorization, so we are always able to use |
| 92 |
it and secondly, the null seed changes across TPM resets, meaning if |
| 93 |
we establish trust on the null seed at start of day, all sessions |
| 94 |
salted with the derived key will fail if the TPM is reset and the seed |
| 95 |
changes. |
| 96 |
|
| 97 |
Obviously using the null seed without any other prior shared secrets, |
| 98 |
we have to create and read the initial public key which could, of |
| 99 |
course, be intercepted and substituted by the bus interposer. |
| 100 |
However, the TPM has a key certification mechanism (using the EK |
| 101 |
endorsement certificate, creating an attestation identity key and |
| 102 |
certifying the null seed primary with that key) which is too complex |
| 103 |
to run within the kernel, so we keep a copy of the null primary key |
| 104 |
name, which is what is exported via sysfs so user-space can run the |
| 105 |
full certification when it boots. The definitive guarantee here is |
| 106 |
that if the null primary key certifies correctly, you know all your |
| 107 |
TPM transactions since start of day were secure and if it doesn't, you |
| 108 |
know there's an interposer on your system (and that any secret used |
| 109 |
during boot may have been leaked). |
| 110 |
|
| 111 |
Stacking Trust |
| 112 |
-------------- |
| 113 |
|
| 114 |
In the current null primary scenario, the TPM must be completely |
| 115 |
cleared before handing it on to the next consumer. However the kernel |
| 116 |
hands to user-space the name of the derived null seed key which can |
| 117 |
then be verified by certification in user-space. Therefore, this chain |
| 118 |
of name handoff can be used between the various boot components as |
| 119 |
well (via an unspecified mechanism). For instance, grub could use the |
| 120 |
null seed scheme for security and hand the name off to the kernel in |
| 121 |
the boot area. The kernel could make its own derivation of the key |
| 122 |
and the name and know definitively that if they differ from the handed |
| 123 |
off version that tampering has occurred. Thus it becomes possible to |
| 124 |
chain arbitrary boot components together (UEFI to grub to kernel) via |
| 125 |
the name handoff provided each successive component knows how to |
| 126 |
collect the name and verifies it against its derived key. |
| 127 |
|
| 128 |
Session Properties |
| 129 |
------------------ |
| 130 |
|
| 131 |
All TPM commands the kernel uses allow sessions. HMAC sessions may be |
| 132 |
used to check the integrity of requests and responses and decrypt and |
| 133 |
encrypt flags may be used to shield parameters and responses. The |
| 134 |
HMAC and encryption keys are usually derived from the shared |
| 135 |
authorization secret, but for a lot of kernel operations that is well |
| 136 |
known (and usually empty). Thus, every HMAC session used by the |
| 137 |
kernel must be created using the null primary key as the salt key |
| 138 |
which thus provides a cryptographic input into the session key |
| 139 |
derivation. Thus, the kernel creates the null primary key once (as a |
| 140 |
volatile TPM handle) and keeps it around in a saved context stored in |
| 141 |
tpm_chip for every in-kernel use of the TPM. Currently, because of a |
| 142 |
lack of de-gapping in the in-kernel resource manager, the session must |
| 143 |
be created and destroyed for each operation, but, in future, a single |
| 144 |
session may also be reused for the in-kernel HMAC, encryption and |
| 145 |
decryption sessions. |
| 146 |
|
| 147 |
Protection Types |
| 148 |
---------------- |
| 149 |
|
| 150 |
For every in-kernel operation we use null primary salted HMAC to |
| 151 |
protect the integrity. Additionally, we use parameter encryption to |
| 152 |
protect key sealing and parameter decryption to protect key unsealing |
| 153 |
and random number generation. |
| 154 |
|
| 155 |
Null Primary Key Certification in Userspace |
| 156 |
=========================================== |
| 157 |
|
| 158 |
Every TPM comes shipped with a couple of X.509 certificates for the |
| 159 |
primary endorsement key. This document assumes that the Elliptic |
| 160 |
Curve version of the certificate exists at 01C00002, but will work |
| 161 |
equally well with the RSA certificate (at 01C00001). |
| 162 |
|
| 163 |
The first step in the certification is primary creation using the |
| 164 |
template from the `TCG EK Credential Profile`_ which allows comparison |
| 165 |
of the generated primary key against the one in the certificate (the |
| 166 |
public key must match). Note that generation of the EK primary |
| 167 |
requires the EK hierarchy password, but a pre-generated version of the |
| 168 |
EC primary should exist at 81010002 and a TPM2_ReadPublic() may be |
| 169 |
performed on this without needing the key authority. Next, the |
| 170 |
certificate itself must be verified to chain back to the manufacturer |
| 171 |
root (which should be published on the manufacturer website). Once |
| 172 |
this is done, an attestation key (AK) is generated within the TPM and |
| 173 |
it's name and the EK public key can be used to encrypt a secret using |
| 174 |
TPM2_MakeCredential. The TPM then runs TPM2_ActivateCredential which |
| 175 |
will only recover the secret if the binding between the TPM, the EK |
| 176 |
and the AK is true. the generated AK may now be used to run a |
| 177 |
certification of the null primary key whose name the kernel has |
| 178 |
exported. Since TPM2_MakeCredential/ActivateCredential are somewhat |
| 179 |
complicated, a more simplified process involving an externally |
| 180 |
generated private key is described below. |
| 181 |
|
| 182 |
This process is a simplified abbreviation of the usual privacy CA |
| 183 |
based attestation process. The assumption here is that the |
| 184 |
attestation is done by the TPM owner who thus has access to only the |
| 185 |
owner hierarchy. The owner creates an external public/private key |
| 186 |
pair (assume elliptic curve in this case) and wraps the private key |
| 187 |
for import using an inner wrapping process and parented to the EC |
| 188 |
derived storage primary. The TPM2_Import() is done using a parameter |
| 189 |
decryption HMAC session salted to the EK primary (which also does not |
| 190 |
require the EK key authority) meaning that the inner wrapping key is |
| 191 |
the encrypted parameter and thus the TPM will not be able to perform |
| 192 |
the import unless is possesses the certified EK so if the command |
| 193 |
succeeds and the HMAC verifies on return we know we have a loadable |
| 194 |
copy of the private key only for the certified TPM. This key is now |
| 195 |
loaded into the TPM and the Storage primary flushed (to free up space |
| 196 |
for the null key generation). |
| 197 |
|
| 198 |
The null EC primary is now generated using the Storage profile |
| 199 |
outlined in the `TCG TPM v2.0 Provisioning Guidance`_; the name of |
| 200 |
this key (the hash of the public area) is computed and compared to the |
| 201 |
null seed name presented by the kernel in |
| 202 |
/sys/class/tpm/tpm0/null_name. If the names do not match, the TPM is |
| 203 |
compromised. If the names match, the user performs a TPM2_Certify() |
| 204 |
using the null primary as the object handle and the loaded private key |
| 205 |
as the sign handle and providing randomized qualifying data. The |
| 206 |
signature of the returned certifyInfo is verified against the public |
| 207 |
part of the loaded private key and the qualifying data checked to |
| 208 |
prevent replay. If all of these tests pass, the user is now assured |
| 209 |
that TPM integrity and privacy was preserved across the entire boot |
| 210 |
sequence of this kernel. |
| 211 |
|
| 212 |
.. _TPM Genie: https://www.nccgroup.trust/globalassets/about-us/us/documents/tpm-genie.pdf |
| 213 |
.. _Windows Bitlocker TPM: https://dolosgroup.io/blog/2021/7/9/from-stolen-laptop-to-inside-the-company-network |
| 214 |
.. _attack against TPM based Linux disk encryption: https://www.secura.com/blog/tpm-sniffing-attacks-against-non-bitlocker-targets |
| 215 |
.. _TCG EK Credential Profile: https://trustedcomputinggroup.org/resource/tcg-ek-credential-profile-for-tpm-family-2-0/ |
| 216 |
.. _TCG TPM v2.0 Provisioning Guidance: https://trustedcomputinggroup.org/resource/tcg-tpm-v2-0-provisioning-guidance/ |