| Network Working Group | H. Birkholz |
| Internet-Draft | Fraunhofer SIT |
| Intended status: Standards Track | M. Wiseman |
| Expires: April 27, 2019 | GE Global Research |
| H. Tschofenig | |
| ARM Ltd. | |
| N. Smith | |
| Intel | |
| October 24, 2018 |
Architecture and Reference Terminology for Remote Attestation Procedures
draft-birkholz-rats-architecture-00
Remote ATtestation ProcedureS (RATS), such as Remote Integrity VERification (RIVER), the creation of Entity Attestation Tokens (EAT), software integrity Measurement And ATtestation (MAAT), or the attestation of device characteristics, in general, are based on specific operations and qualities provided by hardware and software. The RATS architecture maps corresponding functions and operation capabilities to specific RATS roles. The goal is to enable an appropriate conveyance of believable evidence about device health or trusted claims about device capabilities via network protocols. The flows of data between these roles depend on the composition of RATS roles and their location with respect to devices or services. The RATS architecture provides these roles as building blocks to enable suitable composition, while remaining hardware-agnostic. This flexibility is intended to address a significant majority of use cases or usage scenarios in the domain of RATS. Examples include, but are not limited to: financial transactions, voting machines, critical safety systems, network equipment health, or trustworthy end-user device management.
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 April 27, 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.
This document provides normative guidance how to use, create or adopt network protocols that facilitate remote attestation procedures. The foundation of the RATS architecture are specific roles that can be chained and as a result compose remote attestation procedures. The term attestation, unfortunately, is an overloaded term. There are different interpretations, connotations and meanings to the term attestation and therefore also to terms related to attestation. In consequence, this document also provides a detailed definition of Attestation Terminology. The intent is to illustrate and remediate the impedance mismatch of terms related to Remote Attestation Procedures used in different domains today. New terms defined by this document provide a consolidated basis to support future work on RATS in the IETF and beyond.
The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “NOT RECOMMENDED”, “MAY”, and “OPTIONAL” in this document are to be interpreted as described in RFC 2119, BCP 14 [RFC2119].
The goal of the RATS architecture is to provide the building blocks - the roles defined by the RATS architecture - to enable the composition of service-chains and work-flows to create and appraise evidence about the trustworthiness of devices and services.
The RATS architecture does not prescribe specific payload definitions, role composition, or protocol use. However, it imposes requirements on payload definitions, interfaces, and network protocols with respect to proofs of freshness, attestation provenance, and required operational primitives that are available to devices and services taking on RATS roles. For brevity, the term “system” is a synonym for “device or service” in this document.
In the RATS architecture, devices or services can take on RATS roles. In this context, devices are typically composite devices (in the case of atomically integrated devices that would result in a composite device with one component). Services are software components - e.g. a daemon, a virtual network function (vnf) or a network security function (nsf) - that can reside on one or more devices and are not necessarily bound to a specific set of devices.
Devices or Services (Systems) can take on one or more RATS roles either by separate functions or via a collapsed functions that take on multiple RATS roles. Systems that take on RATS roles:
Systems can be distinguished on the management plane via identity documents (which includes specific claim sets about device characteristics, see RFC4949) or via trusted claim sets (e.g. the Entity Attestation Token) and can be addressed by network protocols via IP addresses. RATS can be used in environments without network protocols and RATS roles can be used to design work-flows in these domains, correspondingly. However, the primary focus of the RATS architecture is to facilitate network protocols between RATS roles that convey information via the Internet Protocol.
Relevant decision-factors that influence the composition of RATS roles on systems and resulting work-flows are (amongst others):
[RFC4949] provides definitions that highlight the difference between a “trusted system” and a “trustworthy system”. The following definitions exclude the explicit specialization of concepts that are “environmental disruption” as well as “human user and operator errors”.
A trusted system in the context of RATS “operates as expected, according to design and policy, doing what is required and not doing other things” [RFC4949]. A trustworthy system is a system “that not only is trusted, but also warrants that trust because the system’s behavior can be validated in some convincing way, such as through formal analysis or code review” [RFC4949].
The goal of RATS is to convey information about system component characteristics, such as integrity or authenticity, that can be appraised in a convincing way.
RATS require trust relationships with third parties that qualify assertions about, for example, origin of data, the manufacturer or the capabilities of a system, or the origination of attestation evidence (attestation provenance). Without trusted authorities (e.g. a certificate authority) it is virtually impossible to assess the level of assurance (or resulting level of confidence, correspondingly) of information produced by RATS. Trusting a system does not make it trustworthy. Assessing trustworthiness requires the conveyance of evidence that a system is a trustworthy system, which has to originate from the system itself and has to be convincing. If the convincing information is not originating from the system itself, it comprises trusted claim sets and not evidence. In essence, the attestation provenance of attestation evidence is the system that intends to present its trustworthiness in a believable manner.
The essential basis for trust in the information created via RATS are roots of trust.
Roots of trust are defined by the NIST special publication 800-164 draft as “security primitives composed of hardware, firmware and/or software that provide a set of trusted, security-critical functions. They must always behave in an expected manner because their misbehavior cannot be detected. As such, RoTs need to be secured by their design. Hardware RoTs are preferred over software RoTs due to their immutability, smaller attack surface, and more reliable behavior.”
If the root of trust involved is a root of trust for measurement (RTM), the producer of information takes on the role of a asserter. An asserter can also make use of a root of trust for integrity (RTI) in order to increase the level of assurance in the assertions produced. If the root of trust involved is a root of trust for reporting (RTR), the producer of information takes on the role of an attester.
The RATS asserter role produces measurements about the system’s characteristics in the form of signed (sometimes un-signed) claim sets in order to convey information. A secret signing key is required for this procedure, which is typically stored in a shielded location that can be trusted, for example, via a root of trust for storage (RTS).
The RATS attester role produces signed attestation evidence in order to convey information. The secret key required for this procedure is stored in a shielded location that only allows access to that key, if a specific operational state of the system is met. The trust with respect to this origination is based on a root of trust for reporting.
There are six roles defined in the RATS architecture. iFigure 1 provides a simplified overview of the roles defined in this section.
+------------+ +------------------+
| | | |
| Attester | +->| Verifier |
| | | | |
+------------+ | +------------------+
^ |
| | +------------------+
| +----------------+ | | |
+---->| |<-+ | Authentication |
| Interconnect |<--->| Checker |
+---->| |<-+ | |
| +----------------+ | +------------------+
v |
+------------+ | +------------------+
| | | | |
| Claimant | +->| Relying Party |
| | | |
+------------+ +------------------+
Figure 1: Overall Relationships of Roles in the RATS Architecture
In order to provide an intuitive understanding how the roles used in RATS can be composed into work-flows, this document provides a few example work-flows. Boxes in the following examples that include more than one role are systems that take on more than one role.
If there is a trust relationship between a trusted third party that can assert that signed claims created by a claimant guarantee a trustworthy origination of claim, the work-flow depicted in Figure 2 can facilitate a trust-based implicit remote attestation procedure. The information conveyed are signed claim sets that are trusted via an authoritative third party. In this work-flow claim emission is triggered by the claimant. Variations based on requests emitted by the relying party can be easily facilitated by the same set of roles.
+---------------------------------------+
| |
| +------------------+ +-----------+ |
+------------+ +----------------+ | | | | | |
| | | | | | Authentication | | Relying | |
| Claimant |->| Interconnect |--+->| Checker |->| Party | |
| | | | | | | | | |
+------------+ +----------------+ | +------------------+ +-----------+ |
| |
+---------------------------------------+
Figure 2: Conveyance of Trusted Claim Sets Validated by Signature
If there is trust in the root of trust for reporting based on the assertions of a trusted third party, the work-flow depicted in Figure 3 can facilitate an evidence-based explicit remote attestation procedure. The information conveyed is signed attestation evidence that is created by the trusted verifier. In this work-flow claims do not necessarily have to be signed and the work-flow is triggered by the attestor that aggregates claims from a root of trust of measurement. Variations based on requests emitted by the verifier can be easily facilitated by the same set of roles.
+------------------+ +------------------------+ | | | +------------------+ | | +------------+ | +----------------+ | | | | | | | | | | | | Authentication | | | | Attester |--+->| Interconnect |--+->| Checker | | | | | | | | | | | | | +------------+ | +----------------+ | +------------------+ | | ^ | +-------------------+ | | | | | | | | | | | | +-----------+ v | | +-----+------+ | | | | +------------+ | | | | | | | Relying | | | | | | Claimant | | | | Party |<---------| Verifier | | | | | | | | | | | | | +------------+ | | +-----------+ +------------+ | | | | | +------------------+ +--------------------------------------------+
Figure 3: Conveyance of Attestation Evidence Appraised by a Verifier
During its evolution, the term Remote Attestation has been used in multiple contexts and multiple scopes and in consequence accumulated various connotations with slightly different semantic meaning. Correspondingly, Remote Attestation Procedures (RATS) are employed in various usage scenarios and different environments.
In order to better understand and grasp the intent and meaning of specific RATS in the scope of the security area - including the requirements that are addressed by them - this document provides an overview of existing work, its background, and common terminology. As the contribution, from that state-of-the-art a set of terms that provides a stable basis for future work on RATS in the IETF is derived.
In essence, a prerequisite for providing an adequate set of terms and definitions for the RATS architecute is a general understanding and a common definitions of “what” RATS can accomplish “how” RATS can to be used.
Please note that this section is still missing various references and is considered “under construction”. The majority of definitions is still only originating from IETF work. Future iterations will pull in more complementary definitions from other SDO (e.g. Global Platform, TCG, etc.) and a general structure template to highlight semantic relationships and capable of resolving potential discrepancies will be introduced. A section of context awareness will provide further insight on how Attestation procedures are vital to ongoing work in the IETF (e.g. I2NSF & tokbind). The definitions in the section about RATS are still self-describing in this version. Additional explanatory text will be added to provide more context and coherence.
A very prominent goal of RATS is to address the “lying endpoint problem”. The lying endpoint problem is characterized as a condition of a Computing Context where the information or behavior embedded, created, relayed, stored, or emitted by the Computing Context is not “correct” according to expectations of the authorized system designers, operators and users. There can be multiple reasons why these expectations are incorrect, either from malicious Activity, unanticipated conditions or accidental means. The observed behavior, nevertheless, appears to be a compromised Computing Context.
Attempts to “scrub” the data or “proxy” control elements implies the existence of a more fundamental trusted endpoint that is operating correctly. Therefore, Remote Attestation - the technology designed to detect and mitigate the “lying endpoint problem” – must be trusted to behave correctly independent of other controls.
Consequently, a “lying endpoint” cannot also be a “trusted system”.
Remote Attestation procedures are intended to enable the consumer of information emitted by a Computing Context to assess the validity and integrity of the information transferred. The approach is based, for example, on the assumption that if attestation evidence can be provided in order to prove the integrity of every software instance installed involved in the activity of creating the emitted information in question, the emitted information can be considered valid and integer.
In contrast, such Evidence has to be impossible to create if the software instances used in a Computing Context are compromised. Attestation activities that are intended to create this Evidence therefore also provide guarantees about the validity of the Evidence they can create.
RATS imply the involvement of at least two players (roles) who seek to overcome the lying endpoint problem. The Verifier wishes to consume application data supplied by a Computing Context. But before application data is consumed, the Verifier obtains Attestation Evidence about the Computing Context to assess likelihood of poisoned data due to endpoint compromise or failure. Remote Attestation argues that a systems’s integrity characteristics should not be believed until rationale for believability is presented to the relying party seeking to interact with the system.
An Interconnect defines an untrusted channel between subject and object wherein the rationale for believability is securely exchanged. The type of interconnect technology could vary widely, ranging from GPIO pins, to a PC peripheral IO bus, to the Internet, to a direct physical connection, to a wireless radio-receiver association, or to a world wide mesh of peers. In other words, virtually every kind communication path could be used as the “Interconnect” in RATS. In fact, a single party could take on all roles at the same time (e.g. Self Encrypting Devices).
Attestation evidence can be thought of as the topics of the exchange that is created the operational primitives of a root of trust for reporting. Evidence may be structured in an interoperable format called claims that may include references to the claimants which are asserting the claims. RATS aims to define “interoperable Remote Attestation” such that evidence can be created and consumed by different ecosystem systems and can be securely exchanged by a broad set of network protocols.
This document relies on terminology found in [RFC4949]. This document presumes the reader is familiar with the following terms.
Terminology defined by this document is preceded by a dollar sign ($) to distinguish it from terms defined elsewhere and as a way to disambiguate term definition from explanatory text.
Terms defined by this document that are subsequently used by this document are distinguished by capitalizing the first letter of the term (e.g. Term or First_word Second_word).
This section introduces the term Computing Context in order to specialize the notions of environment and endpoint to terminology that has relevance to trusted computing. Attestation is a discipline of trusted computing.
A Computing Context could refer to a large variety of endpoints. Examples include but are not limited to: the compartmentalization of physical resources, the separation of software instances with different dependencies in dedicated containers, and the nesting of virtual components via hardware-based and software-based solutions. The number of approaches and techniques to construct an endpoint continuously changes with new innovation. Hence, it isn’t a goal of this document to define remote attestation for a fixed set of endpoints. Rather, it attempts to define endpoints conceptually and rely on Claims management as a way to clarify the details and specific attributes of conceptual endpoints.
Computing Contexts may be recursive in nature in that it could be composed of a system that is itself a composite of subsystems. In consequence, a system may be composed of other systems that may be further composed of one or more Computing Contexts capable of taking on the RATS roles. The scope and application of these roles can range from:
Analogously, the increasing number of features and functions that constitute components of a device start to blur the lines that are required to categorize each solution and approach precisely. To address this increasingly challenging categorization, the term Computing Context defines the characteristics of the (sub)systems that can take on the role of an Attester and/or the role of a Verifier. This approach is intended to provide a stable basis of definitions for future solutions that continuous to remain viable long-term.
While the semantic relationships highlighted above constitute the fundamental basis to provide a define Computing Context, the following list of object characteristics is intended to improve the application of the term and provide a better understanding of its meaning:
Computing context characteristics do not necessarily include a network interface with associated network addresses (as required by the definition of an endpoint) – although it is very likely to have (access to) one.
[Issue: This conclusion could be incorrect] In contrast, a container [ref docker, find a more general term here] context is not a distinguishable isolated slice of an information system and therefore is not an independent Computing Context. [more feedback on this statement is required as the capabilities of docker-like functions evolve continuously]
Examples include: a smart phone, a nested virtual machine, a virtualized firewall function running distributed on a cluster of physical and virtual nodes, or a trust-zone.
Computing Contexts may relate to other Computing Contexts that are decomposable in a variety of ways.
The scope of Computing Context encompasses a broad spectrum of systems including, but not limited to:
A Computing Context may be realized in a variety of ways including, but not limited to:
Analogously, a computing sub-context is a decomposition of a Computing Context; a subsystem is a decomposition of a system; a sub-component is a decomposition of a component; and a peer node is a decomposition of a node cluster.
A formal semantic relationship is therefore expressed using an information model that captures interactions, relationships, bindings and interfaces among systems, subsystems, system components, system entities or objects.
[Issue: A tangible relationship to an information model is required here] An information model that richly captures Computing Context semantics is therefore believed to be relevant if not fundamental to Remote Attestation.
The identity of a Computing Context implies there is a binding operation between an identifier and the Computing Context.
Confidence in the identity assurance level [NIST SP-800-63-3] or the assurance levels for identity authentication [RFC4949] is a property of the identifier uniqueness properties and binding operation veracity. Such properties impact the trustworthiness of associated attestation Evidence.
Attestation Evidence created by RATS is a form of telemetry about a computing environment that enables better security risk management through disclosure of security properties of the environment. Attestation may be performed locally (within the same computing environment) or remotely (between different computing environments). The exchange of attestation evidence can be formalized to include well-defined protocol, message syntax and semantics.
A form of telemetry involving the delivery of Claims describing various security properties of a Computing Context by an Attester, such that the Claims can be used as Evidence toward convincing a Verifier regarding trustworthiness of the Computing Context.
Evidence conveyed to a Verifier by an Attester is structured to facilitate syntactic and semantic interoperability. An information model defines the tag namespaces used to create tag-value pairs containing discrete bits of Evidence.
Note: Proofs of Claims assertions may be separated from the Claim itself. For example, a secure transport over which Claims are conveyed where Claimant’s signing key integrity protects the transport payload could be used as proof of Claim assertion. Alternatively, each Claim could be separately signed by a Claimant.
This section introduces terms and definitions that are required to illustrate the scope and the granularity of RATS workflows in the domain of security automation. Terms defined in the following sections will be based on this workflow-related definitions.
In general, RATS are composed of iterative activities that can be conducted in intervals. It is neither a generic set of actions nor simply a task, because the actual actions to be conducted by RATS can vary significantly depending on the protocols employed and types of Computing Contexts involved.
A “lying endpoint” is not trustworthy.
This document provides NNN prominent examples of use cases Attestation procedures are intended to address:
[working title, pulled from various sources, vital]
This document will include requests to IANA:
There are always some.
Maybe.
No changes yet.
| [RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997. |
| [I-D.ietf-sacm-terminology] | Birkholz, H., Lu, J., Strassner, J., Cam-Winget, N. and A. Montville, "Security Automation and Continuous Monitoring (SACM) Terminology", Internet-Draft draft-ietf-sacm-terminology-15, June 2018. |
| [RFC4949] | Shirey, R., "Internet Security Glossary, Version 2", FYI 36, RFC 4949, DOI 10.17487/RFC4949, August 2007. |
| [RFC7519] | Jones, M., Bradley, J. and N. Sakimura, "JSON Web Token (JWT)", RFC 7519, DOI 10.17487/RFC7519, May 2015. |