| Internet-Draft | BRSKI-CLOUD | August 2023 |
| Friel, et al. | Expires 25 February 2024 | [Page] |
Bootstrapping Remote Secure Key Infrastructures defines how to onboard a device securely into an operator maintained infrastructure. It assumes that there is local network infrastructure for the device to discover and to help the device. This document extends the new device behaviour so that if no local infrastructure is available, such as in a home or remote office, that the device can use a well defined "call-home" mechanism to find the operator maintained infrastructure.¶
To this, this document defines how to contact a well-known Cloud Registrar, and two ways in which the new device may be redirected towards the operator maintained infrastructure.¶
This note is to be removed before publishing as an RFC.¶
Status information for this document may be found at https://datatracker.ietf.org/doc/draft-ietf-anima-brski-cloud/.¶
Discussion of this document takes place on the anima Working Group mailing list (mailto:anima@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/anima/.¶
Source for this draft and an issue tracker can be found at https://github.com/anima-wg/brski-cloud.¶
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 25 February 2024.¶
Copyright (c) 2023 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 Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
Bootstrapping Remote Secure Key Infrastructures [BRSKI] and [RFC8994] specifies automated network onboarding of devices, referred to as pledges, within an Autonomic Control Plane or other managed network infrastructure. BRSKI Section 2.7 describes how a pledge "MAY contact a well-known URI of a Cloud Registrar if a local Registrar cannot be discovered or if the pledge's target use cases do not include a local Registrar".¶
This document further specifies use of a BRSKI Cloud Registrar and clarifies operations that are not sufficiently specified in BRSKI.¶
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 BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
This document uses the terms Pledge, Registrar, MASA, and Voucher from [BRSKI] and [RFC8366].¶
The domain where the pledge is physically located and bootstrapping from. This may be different to the pledge owner's domain.¶
The domain that the pledge needs to discover and bootstrap with.¶
The default Registrar that is deployed at a URI that is well known to the pledge.¶
The Registrar that is operated by the Owner, or the Owner's delegate. There may not be an Owner Registrar in all deployment scenarios.¶
The Registrar discovered on the Local Domain. There may not be a Local Domain Registrar in all deployment scenarios.¶
Value Added Reseller¶
Two high level use cases are documented here. There are more details provided in sections Section 4.1 and Section 4.2. While both use cases aid with incremental deployment of BRSKI infrastructure, for many smaller sites (such as teleworkers) no further infrastructure is expected.¶
The pledge is not expected to know which of these two situations it is in. The pledge determines this based upon signals that it receives from the Cloud Registrar. The Cloud Registrar is expected to make the determination based upon the identity presented by the pledge.¶
A Cloud Registrar will typically handle all the devices of a particular product line from a particular manufacturer. This document places no restrictions on how many different deployments or owner sites the Cloud Registrar can handle, or how many devices per site that the Cloud Registrar can handle. It is also entirely possible that all devices sold by through a particular Value Added Reseller (VAR) might be preloaded with a configuration that changes the Cloud Registrar URL to point to a VAR. Such an effort would require unboxing each device in a controlled environment, but the provisioning could occur using a regular BRSKI or SZTP [RFC8572] process.¶
A pledge is bootstrapping from a remote location with no local domain Registrar (specifically: with no local infrastructure to provide for automated discovery), and needs to discover its owner Registrar. The Cloud Registrar is used by the pledge to discover the owner Registrar. The Cloud Registrar redirects the pledge to the owner Registrar, and the pledge completes bootstrap against the owner Registrar.¶
A typical example is an enduser deploying a pledge in a home or small branch office, where the pledge belongs to the enduser's employer. There is no local domain Registrar, and the pledge needs to discover and bootstrap with the employer's Registrar which is deployed in headquarters. For example, an enduser is deploying an IP phone in a home office and the phone needs to register to an IP PBX deployed in their employer's office.¶
A pledge is bootstrapping where the owner organization does not yet have an owner Registrar deployed. The Cloud Registrar issues a voucher, and the pledge completes trust bootstrap using the Cloud Registrar. The voucher issued by the cloud includes domain information for the owner's Enrollment over Secure Transport (EST) [RFC7030] service the pledge should use for certificate enrollment.¶
In one use case, an organization has an EST service deployed, but does not have yet a BRSKI capable Registrar service deployed. The pledge is deployed in the organization's domain, but does not discover a local domain Registrar or owner Registrar. The pledge uses the Cloud Registrar to bootstrap, and the Cloud Registrar provides a voucher that includes instructions on finding the organization's EST service.¶
The high level architecture is illustrated in Figure 1.¶
The pledge connects to the Cloud Registrar during bootstrap.¶
The Cloud Registrar may redirect the pledge to an owner Registrar in order to complete bootstrap against the owner Registrar.¶
If the Cloud Registrar issues a voucher itself without redirecting the pledge to an owner Registrar, the Cloud Registrar will inform the pledge what domain to use for accessing EST services in the voucher response.¶
Finally, when bootstrapping against an owner Registrar, this Registrar may interact with a backend CA to assist in issuing certificates to the pledge. The mechanisms and protocols by which the Registrar interacts with the CA are transparent to the pledge and are out-of-scope of this document.¶
The architecture shows the Cloud Registrar and MASA as being logically separate entities. The two functions could of course be integrated into a single service.¶
There are two different mechanisms for a Cloud Registrar to handle voucher requests. It can redirect the request to Owner Registrar for handling, or it can return a voucher that pins the actual Owner Registrar. When returning a voucher, additional bootstrapping information embedded in the voucher. Both mechanisms are described in detail later in this document.¶
As depicted in Figure 1, there are a number of parties involve in the process. The Manufacturer, or Original Equipment Maker (OEM) builds the device, but also is expected to run the MASA, or arrange for it to exist.¶
The network operator or enterprise is the intended owner of the new device: the pledge. This could be the enterprise itself, or in many cases there is some outsourced IT department that might be involved. They are the operator of the Registrar or EST Server. They may also operate the CA, or they may contract those services from another entity.¶
Unlike in [BRSKI] there is a potential additional party involved, the network integrator, who may operate the Cloud Registrar. This is typically a value added reseller who works with the OEM to ship products with the right configuration to the owner. For example, SIP telephones or other conferencing systems may be installed by this VAR, often shipped directly from a warehouse to the customer's remote office location. The network integrator and manufacturer are aware of which devices have been shipped to the integrator through sales channel integrations, and so the manufacturer's Cloud Registrar is able to redirect the pledge through a chain of Cloud Registrars, as explained in Section 3.3.1.¶
The assumption is that the pledge already has network connectivity prior to connecting to the Cloud Registrar. The pledge must have an IP address that is able to make DNS queries, and be able to send HTTP requests to the Cloud Registrar. There are many ways to accomplish this, from routeable IPv4 or IPv6 addresses, to use of NAT44, to using HTTP or SOCKS proxies. There are are DHCP options that a network operator can configure to accomplish any of these options. The pledge operator has already connected the pledge to the network, and the mechanism by which this has happened is out of scope of this document. For many telephony applications, this is typically going to be a wired connection. For wireless use cases, some kind of existing WiFi onboarding mechanism such as WPS. Similarly, what address space the IP address belongs to, whether it is an IPv4 or IPv6 address, or if there are firewalls or proxies deployed between the pledge and the cloud registar are all out of scope of this document.¶
BRSKI section 5.9.2 specifies that the pledge MUST send an EST [RFC7030] CSR Attributes request to the Registrar. The Registrar MAY use this mechanism to instruct the pledge about the identities it should include in the CSR request it sends as part of enrollment. The Registrar may use this mechanism to tell the pledge what Subject or Subject Alternative Name identity information to include in its CSR request. This can be useful if the Subject must have a specific value in order to complete enrollment with the CA.¶
EST [RFC7030] is not clear on how the CSR Attributes response should be structured, and in particular is not clear on how a server can instruct a client to include specific attribute values in its CSR. [I-D.ietf-lamps-rfc7030-csrattrs] clarifies how a server can use CSR Attributes response to specify specific values for attributes that the client should include in its CSR.¶
For example, the pledge may only be aware of its IDevID Subject which includes a manufacturer serial number, but must include a specific fully qualified domain name in the CSR in order to complete domain ownership proofs required by the CA.¶
As another example, the Registrar may deem the manufacturer serial number in an IDevID as personally identifiable information, and may want to specify a new random opaque identifier that the pledge should use in its CSR.¶
BRSKI defines how a pledge MAY contact a well-known URI of a Cloud Registrar if a local domain Registrar cannot be discovered. Additionally, certain pledge types might never attempt to discover a local domain Registrar and might automatically bootstrap against a Cloud Registrar.¶
The details of the URI are manufacturer specific, with BRSKI giving the example "brski-registrar.manufacturer.example.com".¶
The Pledge SHOULD be provided with the entire URL of the Cloud Registrar, including the path component, which is typically "/.well-known/brski/requestvoucher", but may be another value.¶
The pledge MUST use an Implicit Trust Anchor database (see EST [RFC7030]) to authenticate the Cloud Registrar service. In order to make use of a Cloud Registrar, the Pledge MUST be manufactured with pre-loaded trust-anchors that are used to validate the TLS connection. The TLS connection can be validated using a public Web PKI trust anchors using [RFC6125] DNS-ID mechanisms, a pinned certification authority, or even a pinned raw public key. This is a local implementation decision.¶
The pledge MUST NOT establish a provisional TLS connection (see BRSKI section 5.1) with the Cloud Registrar.¶
The Cloud Registrar MUST validate the identity of the pledge by sending a TLS CertificateRequest message to the pledge during TLS session establishment. The Cloud Registrar MAY include a certificate_authorities field in the message to specify the set of allowed IDevID issuing CAs that pledges may use when establishing connections with the Cloud Registrar.¶
The Cloud Registrar MAY only allow connections from pledges that have an IDevID that is signed by one of a specific set of CAs, e.g. IDevIDs issued by certain manufacturers.¶
The Cloud Registrar MAY allow pledges to connect using self-signed identity certificates or using Raw Public Key [RFC7250] certificates.¶
After the pledge has established a full TLS connection with the Cloud Registrar and has verified the Cloud Registrar PKI identity, the pledge generates a voucher request message as outlined in BRSKI section 5.2, and sends the voucher request message to the Cloud Registrar.¶
The Cloud Registrar must determine pledge ownership. Prior to ownership determination, the Registrar checks the request for correctness and if it is unwilling or unable to handle the request, it MUST return a suitable 4xx or 5xx error response to the pledge as defined by [BRSKI] and HTTP. In the case of an unknown pledge a 404 is returned, for a malformed request 400 is returned, or in case of server overload 503.¶
If the request is correct and the Registrar is able to handle it, but unable to determine ownership, then it MUST return a 401 Unauthorized response to the pledge. This signals to the Pledge that there is currently no known owner domain for it, but that retrying later might resolve this situation. The Registrar MAY also include a Retry-After header that includes a time to defer. A pledge with some kind of indicator (such as a screen or LED) SHOULD consider this an onboarding failure, and indicate this to the operator.¶
If the Cloud Registrar successfully determines ownership, then it MUST take one of the following actions:¶
The Cloud Registrar needs some suitable mechanism for knowing the correct owner of a connecting pledge based on the presented identity certificate. For example, if the pledge establishes TLS using an IDevID that is signed by a known manufacturing CA, the Registrar could extract the serial number from the IDevID and use this to lookup a database of pledge IDevID serial numbers to owners.¶
Alternatively, if the Cloud Registrar allows pledges to connect using self-signed certificates, the Registrar could use the thumbprint of the self-signed certificate to lookup a database of pledge self-signed certificate thumbprints to owners.¶
The mechanism by which the Cloud Registrar determines pledge ownership is out-of-scope of this document.¶
Once the Cloud Registrar has determined pledge ownership, the Cloud Registrar MAY redirect the pledge to the owner Registrar in order to complete bootstrap. Ownership registration will require the owner to register their local domain. The mechanism by which pledge owners register their domain with the Cloud Registrar is out-of-scope of this document.¶
In case of redirection, the Cloud Registrar replies to the voucher request with a HTTP 307 Temporary Redirect response code, including the owner's local domain in the HTTP Location header.¶
If the Cloud Registrar issues a voucher, it returns the voucher in a HTTP response with a 200 response code.¶
The Cloud Registrar MAY issue a 202 response code if it is willing to issue a voucher, but will take some time to prepare the voucher.¶
The voucher MUST include the "est-domain" field as defined in [RFC8366bis]. This tells the pledge where the domain of the EST service to use for completing certificate enrollment.¶
The voucher MAY include the "additional-configuration" field. This points the pledge to a URI where application specific additional configuration information may be retrieved. Pledge and Registrar behavior for handling and specifying the "additional-configuration" field is out-of-scope of this document.¶
The Cloud Registrar returned a 307 response to the voucher request.¶
The pledge SHOULD restart the process using a new voucher request using the location provided in the HTTP redirect. Note if the pledge is able to validate the new server using a trust anchor found in its Implicit Trust Anchor database, then it MAY accept additional 307 redirects.¶
The pledge MUST never visit a location that it has already been to, in order to avoid any kind of cycle. If it happens that a location is repeated, then the pledge MUST fail the onboarding attempt and go back to the beginning, which includes listening to other sources of onboarding information as specified in [BRSKI] section 4.1 and 5.0. The pledge MUST also have a limit on the number of redirects it will a follow, as the cycle detection requires that it keep track of the places it has been. That limit MUST be in the dozens or more redirects such that no reasonable delegation path would be affected.¶
The pledge MUST establish a provisional TLS connection with specified local domain Registrar. The pledge MUST NOT use its Implicit Trust Anchor database for validating the local domain Registrar identity. The pledge MUST send a voucher request message via the local domain Registrar.¶
After the pledge receives the voucher, it validates the TLS connection to the local domain Registrar and continues with enrollment and bootstrap as per standard BRSKI operation.¶
The pledge MUST process any error messages as defined in [BRSKI], and in case of error MUST restart the process from its provisioned Cloud Registrar.¶
The exception is that a 401 Unauthorized code SHOULD cause the Pledge to retry a number of times over a period of a few hours.¶
The Cloud Registrar returned a voucher to the pledge. The pledge SHOULD perform voucher verification as per standard BRSKI operation. The pledge SHOULD verify the voucher signature using the manufacturer-installed trust anchor(s), SHOULD verify the serial number in the voucher, and SHOULD verify any nonce information in the voucher.¶
The pledge SHOULD extract the "est-domain" field from the voucher, and SHOULD continue with EST enrollment as per standard BRSKI operation.¶
This flow illustrates the Owner Registrar Discovery flow. A pledge is bootstrapping in a remote location with no local domain Registrar. The assumption is that the owner Registrar domain is accessible and the pledge can establish a network connection with the owner Registrar. This may require that the owner network firewall exposes the Registrar on the public internet.¶
The process starts, in step 1, when the Pledge establishes a Mutual TLS channel with the Cloud RA using artifacts created during the manufacturing process of the Pledge.¶
In step 2, the Pledge sends a voucher request to the Cloud RA.¶
The Cloud RA completes pledge ownership lookup as outlined in Section 3.2.1, and determines the owner Registrar domain. In step 3, the Cloud RA redirects the pledge to the owner Registrar domain.¶
Steps 4 and onwards follow the standard BRSKI flow. The pledge establishes a provisional TLS connection with the owner Registrar, and sends a voucher request to the owner Registrar. The Registrar forwards the voucher request to the MASA. Assuming the MASA issues a voucher, then the pledge validates the TLS connection with the Registrar using the pinned-domain-cert from the voucher and completes the BRSKI flow.¶
The Voucher includes the EST domain to use for EST enroll. It is assumed services are accessed at that domain too. As trust is already established via the Voucher, the pledge does a full TLS handshake against the local RA indicated by the voucher response.¶
The returned voucher will contain the attribute "est-domain". The pledge is directed to continue enrollment using the EST server found at that URI. The pledge uses the pinned-domain-cert from the voucher to authenticate the EST server.¶
The process starts, in step 1, when the Pledge establishes a Mutual TLS channel with the Cloud RA/MASA using artifacts created during the manufacturing process of the Pledge. In step 2, the Pledge sends a voucher request to the Cloud RA/MASA, and in response the Pledge receives an [RFC8366] format voucher from the Cloud RA/MASA that includes its assigned EST domain in the est-domain attribute.¶
At this stage, the Pledge should be able to establish a TLS channel with the EST server. The connection may involve crossing the Internet requiring a DNS lookup on the provided name. It may also be a local address that includes an IP address literal including both [RFC1918] and IPv6 Unique Local Address. The EST server is validated using the pinned-domain-cert value provided in the voucher as described in [BRSKI] section 5.6.2. This involves treating the artifact provided in the pinned-domain-cert as a trust anchor, and attempting to validate the EST server from this anchor only.¶
There is a case where the pinned-domain-cert is the identical End-Entity (EE) Certificate as the EST server. It also explicitly includes the case where the EST server has a self-signed EE Certificate, but it may also be an EE certificate that is part of a larger PKI. If the certificate is not a self-signed or EE certificate, then the Pledge SHOULD apply [RFC6125] DNS-ID validation on the certificate against the URL provided in the est-domain attribute. If the est-domain was provided by with an IP address literal, then it is unlikely that it can be validated, and in that case, it is expected that either a self-signed certificate or an EE certificate will be pinned.¶
The Pledge also has the details it needs to be able to create the CSR request to send to the RA based on the details provided in the voucher.¶
In step 4, the Pledge establishes a TLS channel with the Cloud RA/MASA, and optionally the pledge should send a request, steps 3.a and 3.b, to the Cloud RA/MASA to inform it that the Pledge was able to establish a secure TLS channel with the EST server.¶
The Pledge then follows that, in step 5, with an EST Enroll request with the CSR and obtains the requested certificate. The Pledge must validate that the issued certificate has the expected identifier obtained from the Cloud RA/MASA in step 3.¶
[RFC8366bis] contains the two needed voucher extensions: est-domain and additional-configuration which are needed when a client is redirected to a local EST server.¶
This document makes no IANA requests.¶
The Cloud Registrar described in this document inherits all of the issues that are described in [BRSKI]. This includes dependency upon continued operation of the manufacturer provided MASA, as well as potential complications where a manufacturer might interfere with resale of a device.¶
In addition to the dependency upon the MASA, the successful enrollment of a device using a Cloud Registrar depends upon the correct and continued operation of this new service. This internet accessible service may be operated by the manufacturer and/or by one or more value-added-resellers. All of the considerations for operation of the MASA also apply to operation of the Cloud Registrar.¶
If the Redirect to Registrar method is used, as described in Section 4.1, there may be a series of 307 redirects. An example of why this might occur is that the manufacturer only knows that it resold the device to a particular value added reseller (VAR), and there may be a chain of such VARs. It is important the pledge avoid being drawn into a loop of redirects. This could happen if a VAR does not think they are authoritative for a particular device. A "helpful" programmer might instead decide to redirect back to the manufacturer in an attempt to restart at the top: perhaps there is another process that updates the manufacturer's database and this process is underway. Instead, the VAR MUST return a 404 error if it cannot process the device. This will force the device to stop, timeout, and then try all mechanisms again.¶
There is another case where a connection problem may occur: when the pledge is behind a captive portal or an intelligent home gateway that provides access control on all connections. Captive portals that do not follow the requirements of [RFC8952] section 1 may forcibly redirect HTTPS connections. While this is a deprecated practice as it breaks TLS in a way that most users can not deal with, it is still common in many networks.¶
On the first connection, the incorrect connection will be discovered because the Pledge will be unable to validate the connection to its Cloud Registrar via DNS-ID. That is, the certificate returned from the captive portal will not match.¶
At this point a network operator who controls the captive portal, noticing the connection to what seems a legitimate destination (the Cloud Registrar), may then permit that connection. This enables the first connection to go through.¶
The connection is then redirected to the Registrar, either via 307, or via est-domain in a voucher. If it is a 307 redirect, then a provisional TLS connection will be initiated, and it will succeed. The provisional TLS connection does not do [RFC6125] DNS-ID validation at the beginning of the connection, so a forced redirection to a captive portal system will not be detected. The subsequent BRSKI POST of a voucher will most likely be met by a 404 or 500 HTTP code. As the connection is provisional, the pledge will be unable to determine this.¶
It is RECOMMENDED therefore that the pledge look for [RFC8910] attributes in DHCP, and if present, use the [RFC8908] API to learn if it is captive.¶
Unlike many other uses of BRSKI, in the Cloud Registrar case it is assumed that the Pledge has connected to a network on which there is addressing and connectivity, but there is no other local configuration available.¶
There is another advantage to being online: the pledge may be able to contact the manufacturer before onboarding in order to apply the latest firmware updates. This may also include updates to the Implicit list of Trust Anchors. In this way, a Pledge that may have been in a dusty box in a warehouse for a long time can be updated to the latest (exploit-free) firmware before attempting onboarding.¶
The Implicit TA database is used to authenticate the Cloud Registrar. This list is built-in by the manufacturer along with a DNS name to which to connect. (The manufacturer could even build in IP addresses as a last resort)¶
The Cloud Registrar does not have a certificate that can be validated using a public (WebPKI) anchor. The pledge may have any kind of Trust Anchor built in: from full multi-level WebPKI to the single self-signed certificate used by the Cloud Registrar. There are many tradeoffs to having more or less of the PKI present in the Pledge, which is addressed in part in [I-D.irtf-t2trg-taxonomy-manufacturer-anchors] in sections 3 and 5.¶
The second redirect case is handled by returning a special extension in the voucher. The Cloud Registrar actually does all of the voucher processing as specified in [BRSKI]. In this case, the Cloud Registrar may be operated by the same entity as the MASA, and it might even be combined into a single server. Whether or not this is the case, it behaves as if it was separate.¶
It may be the case that one or more 307-Redirects have taken the Pledge from the built-in Cloud Registrar to one operated by a VAR.¶
When the Pledge is directed to the owner [RFC7030] Registrar, the Pledge validates the TLS connection with this server using the "pinned-domain-cert" attribute in the voucher. There is no provisional TLS connection, and therefore there are no risks associated with being behind a captive portal.¶
The authors would like to thank for following for their detailed reviews: (ordered by last name): Esko Dijk, Sheng Jiang.¶