TEEP | M. Pei |
Internet-Draft | Symantec |
Intended status: Informational | H. Tschofenig |
Expires: January 5, 2019 | Arm Ltd. |
A. Atyeo | |
Intercede | |
D. Liu | |
Alibaba Group | |
July 4, 2018 |
Trusted Execution Environment Provisioning (TEEP) Architecture
draft-ietf-teep-architecture-00.txt
A Trusted Execution Environment (TEE) was designed to provide a hardware-isolation mechanism to separate a regular operating system from security- sensitive applications.
This architecture document motivates the design and standardization of a protocol for managing the lifecyle of trusted applications running inside a TEE.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on January 5, 2019.
Copyright (c) 2018 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
The Trusted Execution Environment (TEE) concept has been designed to separate a regular operating system, also referred as a Rich Execution Environment (REE), from security- sensitive applications. A TEE provides hardware-enforcement so that any data inside the TEE cannot be read by code outside of the TEE. Compromising a REE and normal applications in the REE do not affect code inside the TEE, which is called a Trusted Application (TA), running inside the TEE.
In an TEE ecosystem, a Trusted Application Manager (TAM) is commonly used to manage keys and TAs that run in a device. Different device vendors may use different TEE implementations. Different application providers or device administrators may choose to use different TAM providers.
To simplify the life of developers an interoperable protocol for managing TAs running in different TEEs of various devices is needed.
The protocol addresses the following problems.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119.
There are multiple technologies that can be used to implement a TEE, and the level of security achieved varies accordingly.
This document uses the following abbreviations:
This specification assumes that an applicable device is equipped with one or more TEEs and each TEE is pre-provisioned with a device-unique public/private key pair, which is securely stored. This key pair is referred to as the 'root of trust' for remote attestation of the associated TEE in a device by an TAM.
A Security Domain (SD) concept is used as the security boundary inside a TEE for trusted applications. Each SD is typically associated with one TA provider as the owner, which is a logical space that contains a SP's TAs. One TA provider may request to have multiple SDs in a TEE. One SD may contain multiple TAs. Each Security Domain requires the management operations of TAs in the form of installation, update and deletion.
A TA binary and configuration data can be from two sources:
The architecture covers the first case where the TA binary and configuration data are delivered from a TAM. The second case calls for an extension when a TAM is absent.
Messages exchange with a TAM require some transport and HTTPS is one commonly used transport.
A payment application in a mobile device requires high security and trust about the hosting device. Payments initiated from a mobile device can use a Trusted Application running inside TEE in the device to provide strong identification and proof of transaction.
For a mobile payment application, some biometric identification information could also be stored in the TEE. The mobile payment application can use such information for authentication.
A secure user interface (UI) may be used in a mobile device to prevent malicious software from stealing sensitive user input data. Such an application implementation often relies on TEE for user input protection.
For better security of authentication, a devices may store its sensitive authentication keys inside a TEE of the device, providing hardware-protected security key strength and trusted execution code.
Internet of Things (IoT) has been posing threats to networks and national infrastructures because of existing weak security in devices. It is very desirable that IoT devices can prevent a malware from stealing or modifying sensitive data such as authentication credentials in the device. A TEE can be the best way to implement such IoT security functions.
TEEs could be used to store variety of sensitive data for IoT devices. For example, a TEE could be used in smart door locks to store a user's biometric information for identification, and for protecting access the locking mechanism. Bike-sharing is another example that shares a similar usage scenario.
A tenant can store sensitive data in a TEE in a cloud computing server such that only the tenant can access the data, preventing the cloud host provider from accessing the data. A tenant can run TAs inside a server TEE for secure operation and enhanced data security. This provides benefits not only to tenants with better data security but also to cloud host provider for reduced liability and increased cloud adoption.
The following are the main components in the system.
This architecture leverages asymmetric cryptography to authenticate a device towards a TAM. Additionally, a TEE in a device authenticates a TAM provider and TA signer. The provisioning of trust anchors to a device may different from one use case to the other. The device administrator may want to have the capability to control what TAs are allowed. A device manufacturer enables verification of the TA signers and TAM providers; it may embed a list of default trust anchors that the signer of an allowed TA's signer certificate should chain to. A device administrator may choose to accept a subset of the allowed TAs via consent or action of downloading.
PKI CA -- CA CA -- | | | | | | | | | Device | | --- Agent / Client App --- | SW | | | | | | | | | | | | | | | | -- TEE TAM------- | | FW
Figure 1: Entities
(App Developer) (App Store) (TAM) (Device with TEE) (CAs) | | | --> (Embedded TEE cert) <-- | | | <------------------------------ Get an app cert ----- | | | <-- Get a TAM cert ------ | | 1. Build two apps: Client App TA | | Client App -- 2a. --> | ----- 3. Install -------> | TA ------- 2b. Supply ------> | 4. Messaging-->| | | | |
Figure 2: Developer Experience
Figure 2 shows an application developer building two applications: 1) a rich Client Application; 2) a TA that provides some security functions to be run inside a TEE. At step 2, the application developer uploads the Client Application (2a) to an Application Store. The Client Application may optionally bundle the TA binary. Meanwhile, the application developer may provide its TA to a TAM provider that will be managing the TA in various devices. 3. A user will go to an Application Store to download the Client Application. The Client Application will trigger TA installation by calling TAM. This is the step 4. The Client Application will get messages from TAM, and interacts with device TEE via an Agent.
The following diagram will show a system diagram about the entity relationships between CAs, TAM, SP and devices.
------- Message Protocol ----- | | | | -------------------- --------------- ---------- | REE | TEE | | TAM | | SP | | --- | --- | | --- | | -- | | | | | | | | | Client | SD (TAs)| | SD / TA | | TA | | Apps | | | Mgmt | | | | | | | | | | | | | | | | | | | | | Trusted | | Trusted | | | | Agent | TAM/SP | | FW/TEE | | | | | CAs | | CAs | | | | | | | | | | | |TEE Key/ | | TAM Key/ | |SP Key/ | | | Cert | | Cert | | Cert | | | FW Key/ | | | | | | | Cert | | | | | -------------------- --------------- ---------- | | | | | | ------------- ---------- --------- | TEE CA | | TAM CA | | SP CA | ------------- ---------- ---------
Figure 3: Keys
In the previous diagram, different CAs can be used for different types of certificates. Messages are always signed, where the signer key is the message originator's private key such as that of a TAM, the private key of a trusted firmware (TFW), or a TEE's private key.
The main components consist of a set of standard messages created by a TAM to deliver device SD and TA management commands to a device, and device attestation and response messages created by a TEE that responds to a TAM's message.
It should be noted that network communication capability is generally not available in TAs in today's TEE-powered devices. The networking functionality must be delegated to a rich Client Application. Client Applications will need to rely on an agent in the REE to interact with a TEE for message exchanges. Consequently, a TAM generally communicates with a Client Application about how it gets messages that originates from TEE inside a device. Similarly, a TA or TEE generally gets messages from a TAM via some Client Application, namely, an agent in this protocol architecture, not directly from the internet.
It is imperative to have an interoperable protocol to communicate with different TEEs in different devices that a Client Application needs to run and access a TA inside a TEE. This is the role of the agent, which is a software component that bridges communication between a TAM and a TEE. The agent does not need to know the actual content of messages except for the TEE routing information.
Each TEE comes with a trust store that contains a whitelist of root CA certificates that are used to validate a TAM's certificate. A TEE will accept a TAM to create new Security Domains and install new TAs on behalf of a SP only if the TAM's certificate is chained to one of the root CA certificates in the TEE's trust store.
A TEE's trust store is typically preloaded at manufacturing time. It is out of the scope in this document to specify how the trust store should be updated when a new root certificate should be added or existing one should be updated or removed. A device manufacturer is expected to provide its TEE trust store live update or out-of-band update to devices.
Before a TAM can begin operation in the marketplace to support TEE-powered devices with a particular TEE, it must obtain a TAM certificate from a CA that is listed in the trust store of the TEE.
The trust anchor store in a TAM consists of a list of CA certificates that sign various device TEE certificates. A TAM decides what devices it will trust the TEE in.
This architecture leverages the following credentials, which allow delivering end-to-end security without relying on any transport security.
Key Entity Name | Location | Issuer | Checked Against | Cardinality |
---|---|---|---|---|
1. TFW key pair and certificate | Device secure storage | FW CA | A white list of FW root CA trusted by TAMs | 1 per device |
2. TEE key pair and certificate | Device TEE | TEE CA under a root CA | A white list of TEE root CA trusted by TAMs | 1 per device |
3. TAM key pair and certificate | TAM provider | TAM CA under a root CA | A white list of TAM root CA embedded in TEE | 1 or multiple can be used by a TAM |
4. SP key pair and certificate | SP | SP signer CA | A SP uses a TAM. TA is signed by a SP signer. TEE delegates trust of TA to TAM. SP signer is associated with a SD as the owner. | 1 or multiple can be used by a TAM |
This architecture uses a PKI. Trust anchors exist on the devices to enable the TEE to authenticate TAMs, and TAMs use trust anchors to authenticate TEEs. Since a PKI is used, many intermediate CAs certificates can chain to a root certificate, each of which can issue many certificates. This makes the protocol highly scalable. New factories that produce TEEs can join the ecosystem. In this case, such a factory can get an intermediate CA certificate from one of the existing roots without requiring that TAMs are updated with information about the new device factory. Likewise, new TAMs can join the ecosystem, providing they are issued a TAM certificate that chains to an existing root whereby existing TEEs will be allowed to be personalized by the TAM without requiring changes to the TEE itself. This enables the ecosystem to scale, and avoids the need for centralized databases of all TEEs produced or all TAMs that exist.
Messages created by a TAM are used to deliver device SD and TA management commands to a device, and device attestation and response messages created by the TEE to respond to TAM messages.
These messages are signed end-to-end and are typically encrypted such that only the targeted device TEE or TAM is able to decrypt and view the actual content.
The primary job of a TAM is to help an SP to manage its trusted applications. A TA is typically installed in an SD. An SD is commonly created for an SP.
When an SP delegates its SD and TA management to a TAM, an SD is created on behalf of a TAM in a TEE and the owner of the SD is assigned to the TAM. An SD may be associated with an SP but the TAM has full privilege to manage the SD for the SP.
Each SD for an SP is associated with only one TAM. When an SP changes TAM, a new SP SD must be created to associate with the new TAM. The TEE will maintain a registry of TAM ID and SP SD ID mapping.
From an SD ownership perspective, the SD tree is flat and there is only one level. An SD is associated with its owner. It is up to TEE implementation how it maintains SD binding information for a TAM and different SPs under the same TAM.
It is an important decision in this protocol specification that a TEE doesn't need to know whether a TAM is authorized to manage the SD for an SP. This authorization is implicitly triggered by an SP Client Application, which instructs what TAM it wants to use. An SD is always associated with a TAM in addition to its SP ID. A rogue TAM isn't able to do anything on an unauthorized SP's SD managed by another TAM.
Since a TAM may support multiple SPs, sharing the same SD name for different SPs creates a dependency in deleting an SD. An SD can be deleted only after all TAs associated with this SD is deleted. An SP cannot delete a Security Domain on its own with a TAM if a TAM decides to introduce such sharing. There are cases where multiple virtual SPs belong to the same organization, and a TAM chooses to use the same SD name for those SPs. This is totally up to the TAM implementation and out of scope of this specification.
There is a need of cryptographically binding proof about the owner of an SD in a device. When an SD is created on behalf of a TAM, a future request from the TAM must present itself as a way that the TEE can verify it is the true owner. The certificate itself cannot reliably used as the owner because TAM may change its certificate.
To this end, each TAM will be associated with a trusted identifier defined as an attribute in the TAM certificate. This field is kept the same when the TAM renew its certificates. A TAM CA is responsible to vet the requested TAM attribute value.
This identifier value must not collide among different TAM providers, and one TAM shouldn't be able to claim the identifier used by another TAM provider.
The certificate extension name to carry the identifier can initially use SubjectAltName:registeredID. A dedicated new extension name may be registered later.
One common choice of the identifier value is the TAM's service URL. A CA can verify the domain ownership of the URL with the TAM in the certificate enrollment process.
A TEE can assign this certificate attribute value as the TAM owner ID for the SDs that are created for the TAM.
An alternative way to represent an SD ownership by a TAM is to have a unique secret key upon SD creation such that only the creator TAM is able to produce a proof-of-possession (PoP) data with the secret.
A sample Security Domain hierarchy for the TEE is shown in Figure 4.
---------- | TEE | ---------- | | ---------- |----------| SP1 SD1 | | ---------- | ---------- |----------| SP1 SD2 | | ---------- | ---------- |----------| SP2 SD1 | ----------
Figure 4: Security Domain Hiearchy
The architecture separates SDs and TAs such that a TAM can only manage or retrieve data for SDs and TAs that it previously created for the SPs it represents.
Step 1: Prepare Images for Devices
Step 2: Inject Key Pairs and Images to Devices
Step 3: Setup attestation key pairs in devices
Step 4: Setup trust anchors in devices
A TEE and TAs do not generally have capability to communicate to the outside of the hosting device. For example, the Global Platform [GPTEE] specifies one such architecture. This calls for a software module in the REE world to handle the network communication. Each Client Application in REE may carry this communication functionality but it must also interact with the TEE for the message exchange. The TEE interaction will vary according to different TEEs. In order for a Client Application to transparently support different TEEs, it is imperative to have a common interface for a Client Application to invoke for exchanging messages with TEEs.
A shared agent comes to meed this need. An agent is an application running in the REE of the device or a SDK that facilitates communication between a TAM and TEE. It also provides interfaces for TAM SDK or Client Applications to query and trigger TA installation that the application needs to use.
This interface for Client Applications may be commonly an Android service call for an Android powered device. A Client Application interacts with a TAM, and turns around to pass messages received from TAM to agent.
In all cases, a Client Application needs to be able to identify an agent that it can use.
An agent abstracts the message exchanges with the TEE in a device. The input data is originated from a TAM that a Client Application connects. A Client Application may also directly call Agent for some TA query functions.
The agent may internally process a request from TAM. At least, it needs to know where to route a message, e.g., TEE instance. It does not need to process or verify message content.
The agent returns TEE / TFW generated response messages to the caller. The agent is not expected to handle any network connection with an application or TAM.
The agent only needs to return an agent error message if the TEE is not reachable for some reason. Other errors are represented as response messages returned from the TEE which will then be passed to the TAM.
A Provider should consider methods of distribution, scope and concurrency on device and runtime options when implementing an agent. Several non-exhaustive options are discussed below. Providers are encouraged to take advantage of the latest communication and platform capabilities to offer the best user experience.
The agent installation is commonly carried out at OEM time. A user can dynamically download and install an agent on-demand.
It is important to ensure a legitimate agent is installed and used. If an agent is compromised it may drop messages and thereby introducing a denial of service.
We anticipate only one shared agent instance in a device. The device's TEE vendor will most probably supply one aent.
With one shared agent, the agent provider is responsible to allow multiple TAMs and TEE providers to achieve interoperability. With a standard agent interface, TAM can implement its own SDK for its SP Client Applications to work with this agent.
Multiple independent agent providers can be used as long as they have standard interface to a Client Application or TAM SDK. Only one agent is expected in a device.
TAM providers are generally expected to provide SDK for SP applications to interact with an agent for the TAM and TEE interaction.
The attestation hierarchy and seed required for TAM protocol operation must be built into the device at manufacture. Additional TEEs can be added post-manufacture using the scheme proposed, but it is outside of the current scope of this document to detail that.
It should be noted that the attestation scheme described is based on signatures. The only encryption that takes place may be the use of a so-called eFuse to release the SBM signing key and later during the protocol lifecycle management interchange with the TAM.
SBM attestation can be optional in TEEP architecture where the starting point of device attestion can be at TEE certfificates. TAM can define its policies on what kind of TEE it trusts if TFW attestation isn't included during the TEE attestation.
During manufacture the following steps are required:
During device boot the following steps are required:
Before a TAM can begin operation in the marketplace to support devices of a given TEE, it must obtain a TAM certificate from a CA that is registered in the trust store of devices with that TEE. In this way, the TEE can check the intermediate and root CA and verify that it trusts this TAM to perform operations on the TEE.
The authors thank Dave Thaler for his very thorough review and many important suggestions. Most content of this document are split from a previously combined OTrP protocol document [I-D.ietf-teep-opentrustprotocol]. We thank the former co-authors Nick Cook and Minho Yoo for the initial document content, and contributors Brian Witten, Tyler Kim, and Alin Mutu.
A TA binary is signed by a TA signer certificate. This TA signing certificate/private key belongs to the SP, and may be self-signed (i.e., it need not participate in a trust hierarchy). It is the responsibility of the TAM to only allow verified TAs from trusted SPs into the system. Delivery of that TA to the TEE is then the responsibility of the TEE, using the security mechanisms provided by the protocol.
We allow a way for an (untrusted) application to check the trustworthiness of a TA. An agent has a function to allow an application to query the information about a TA.
An application in the Rich O/S may perform verification of the TA by verifying the signature of the TA. The GetTAInformation function is available to return the TEE supplied TA signer and TAM signer information to the application. An application can do additional trust checks on the certificate returned for this TA. It might trust the TAM, or require additional SP signer trust chaining.
A TA for multiple SPs must have a different identifier per SP. A TA will be installed in a different SD for each respective SP.
An agent could be malware in the vulnerable Rich OS. A Client Application will connect its TAM provider for required TA installation. It gets command messages from the TAM, and passes the message to the agent.
The architecture enables the TAM to communicate with the device's TEE to manage SDs and TAs. All TAM messages are signed and sensitive data is encrypted such that the agent cannot modify or capture sensitive data.
The TEE implementation provides protection of data on the device. It is the responsibility of the TAM to protect data on its servers.
A root CA for TAM certificates might get compromised. Some TEE trust anchor update mechanism is expected from device OEM. A compromised intermediate CA is covered by OCSP stapling and OCSP validation check in the protocol. A TEE should validate certificate revocation about a TAM certificate chain.
If the root CA of some TEE device certificates is compromised, these devices might be rejected by a TAM, which is a decision of the TAM implementation and policy choice. Any intermediate CA for TEE device certificates SHOULD be validated by TAM with a Certificate Revocation List (CRL) or Online Certificate Status Protocol (OCSP) method.
The TEE SHOULD use validation of the supplied TAM certificates and OCSP stapled data to validate that the TAM is trustworthy.
Since PKI is used, the integrity of the clock within the TEE determines the ability of the TEE to reject an expired TAM certificate, or revoked TAM certificate. Since OCSP stapling includes signature generation time, certificate validity dates are compared to the current time.
TFW and TEE device certificates are expected to be long lived, longer than the lifetime of a device. A TAM certificate usually has a moderate lifetime of 2 to 5 years. A TAM should get renewed or rekeyed certificates. The root CA certificates for a TAM, which are embedded into the trust anchor store in a device, should have long lifetimes that don't require device trust anchor update. On the other hand, it is imperative that OEMs or device providers plan for support of trust anchor update in their shipped devices.
[RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997. |
[RFC4648] | Josefsson, S., "The Base16, Base32, and Base64 Data Encodings", RFC 4648, DOI 10.17487/RFC4648, October 2006. |
[RFC7515] | Jones, M., Bradley, J. and N. Sakimura, "JSON Web Signature (JWS)", RFC 7515, DOI 10.17487/RFC7515, May 2015. |
[RFC7516] | Jones, M. and J. Hildebrand, "JSON Web Encryption (JWE)", RFC 7516, DOI 10.17487/RFC7516, May 2015. |
[RFC7517] | Jones, M., "JSON Web Key (JWK)", RFC 7517, DOI 10.17487/RFC7517, May 2015. |
[RFC7518] | Jones, M., "JSON Web Algorithms (JWA)", RFC 7518, DOI 10.17487/RFC7518, May 2015. |
[GPTEE] | Global Platform, "Global Platform, GlobalPlatform Device Technology: TEE System Architecture, v1.0", 2013. |
[GPTEECLAPI] | Global Platform, "Global Platform, GlobalPlatform Device Technology: TEE Client API Specification, v1.0", 2013. |
[I-D.ietf-teep-opentrustprotocol] | Pei, M., Atyeo, A., Cook, N., Yoo, M. and H. Tschofenig, "The Open Trust Protocol (OTrP)", Internet-Draft draft-ietf-teep-opentrustprotocol-01, July 2018. |