]> AS207960 Cyfyngedig

13 Pen-y-lan Terrace Caerdydd CF23 9EU United Kingdom q@magicalcodewit.ch q@as207970.net https://magicalcodewit.ch

sec Internet Engineering Task Force The documents defines extensions to the Automated Certificate Management Environment (ACME) to allow for the automatic issuance of certificates to Tor hidden services (".onion" domains). Discussion Source for this draft and an issue tracker can be found at . The project website and a reference implementation can be found at .

Introduction The Tor network has the ability to host "Onion Services" only accessible via the Tor network. These use the special use ".onion" top-level domain to identify these services. These can be used as any other domain name could, but do not form part of the DNS infrastructure. The Automated Certificate Management Environment (ACME) defines challenges for validating control of DNS identifiers, and whilst a ".onion" domain may appear as a DNS name, it requires special consideration to validate control of one such that ACME could be used on ".onion" domains. In order to allow ACME to be utilised to issue certificates to ".onion" domains this document specifies challenges suitable to validate control of these domains. Additionally this document defines an alternative to the DNS Certification Authority Authorization (CAA) Resource Record that can be used with ".onion" domains.

Requirements Language 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 when, and only when, they appear in all capitals, as shown here.

Identifier defines the "dns" identifier type. This identifier type MUST be used when requesting a certificate for a ".onion" domain. The value of identifier MUST be the textual representation as defined in §3. The value MAY include subdomain labels. Version 2 addresses MUST NOT be used as these are now considered insecure. Example identifiers: { "type": "dns", "value": "bbcweb3hytmzhn5d532owbu6oqadra5z3ar726vq5kgwwn6aucdccrad.onion" } { "type": "dns", "value": "www.bbcweb3hytmzhn5d532owbu6oqadra5z3ar726vq5kgwwn6aucdccrad.onion" }

Identifier Validation Challenges The CA/Browser Forum Baseline Requirements §B.2 define methods accepted by the CA industry for validation of ".onion" domains. This document incorporates these methods into ACME challenges.

Existing challenges

Existing "dns-01" Challenge The existing "dns-01" challenge MUST NOT be used to validate ".onion" domains.

Existing "http-01" Challenge The "http-01" challenge is defined as in §8.3 may be used to validate a ".onion" domain, with the modifications defined in this standard. The ACME server SHOULD follow redirects; note that these may be redirects to non ".onion" services, and the server SHOULD honour these.

Existing "tls-alpn-01" Challenge The "tls-alpn-01" challenge is defined as in may be used to validate a ".onion" domain, with the modifications defined in this standard.

New "onion-csr-01" Challenge The two methods already defined in ACME and allowed by the CA/BF do not allow issuance of wildcard certificates. This new validation method incorporates the specially signed CSR (as defined by § B.2.b) into ACME to allow for the issuance of wildcard certificates. To this end a new challenge type called "onion-csr-01" is defined, with the following fields:

type (required, string)

The string "onion-csr-01"

nonce (required, string)

A Base64 encoded nonce, including padding characters. It MUST contain at least 64 bits of entropy. It MUST NOT be valid for more than 30 days.

authKey (optional, object)

The Ed25519 public key encoded as per .

{ "type": "onion-csr-01", "url": "https://example.com/acme/chall/bbc625c5", "status": "pending", "nonce": "bI6/MRqV4gw=", "authKey": { ... } } Clients may prove control over the key associated with the ".onion" service by generating their CSR with the following additional extension attributes and signing it with the private key of the ".onion" domain:

• A caSigningNonce attribute containing the nonce provided in the challenge. This MUST be raw bytes, and not the base64 encoded value provided in the challenge object.
• An applicantSigningNonce containing a nonce generated by the client. This MUST have at least 64 bits of entropy. This MUST be raw bytes.

These additional attributes have the following format id-pkix OBJECT IDENTIFIER ::= { iso(1) identified-organization(3) dod(6) internet(1) security(5) mechanisms(5) pkix(7) } id-pe OBJECT IDENTIFIER ::= { id-pkix 1 } cabf OBJECT IDENTIFIER ::= { joint-iso-itu-t(2) international-organizations(23) ca-browser-forum(140) } cabf-caSigningNonce OBJECT IDENTIFIER ::= { cabf 41 } caSigningNonce ATTRIBUTE ::= { WITH SYNTAX OCTET STRING EQUALITY MATCHING RULE octetStringMatch SINGLE VALUE TRUE ID { cabf-caSigningNonce } } cabf-applicantSigningNonce OBJECT IDENTIFIER ::= { cabf 42 } applicantSigningNonce ATTRIBUTE ::= { WITH SYNTAX OCTET STRING EQUALITY MATCHING RULE octetStringMatch SINGLE VALUE TRUE ID { cabf-applicantSigningNonce } } In a variation to the usual state machine of ACME, a client need not respond to the challenge. The act of POSTing a CSR to the finalization endpoint is in itself a response to the challenge. The challenge and order progress directly to either the "valid" or "invalid" state without passing through "processing" or "ready" (respectively). In the case of a CSR posted to the finalization endpoint that does not include the above extensions the order SHOULD remain in the "pending" state and SHOULD NOT transition to the "invalid" state. Only in the case that an "onion-csr-01" CSR is POSTed to the finalization endpoint that subsequently fails validation MUST the order transition to the "invalid" state.

Client authentication to hidden services Some hidden services do not wish to be accessible to the entire Tor network, and so encrypt their hidden service descriptor with the keys of clients authorized to connect. Without a way for the CA to signal what key it will use to connect these services will not be able to obtain a certificate using http-01 or tls-alpn-01, nor enforce CAA with any validation method. To this end, an additional field in the challenge object is defined to allow the ACME server to advertise the Ed25519 public key it will use (as per INTRO-AUTH) to authenticate itself when retrieving the hidden service descriptor.

authKey (optional, object)

The Ed25519 public key encoded as per .

ACME servers MUST NOT use the same public key with multiple hidden services. ACME servers MAY re-use public keys for re-validation of the same hidden service. There is no method to communicate to the CA that client authentication is required; instead the ACME server MUST attempt to calculate its CLIENT-ID as per FIRST-LAYER-CLIENT-BEHAVIOR. If no "auth-client" line in the first layer hidden service descriptor matches the computed client-id then the server MUST assume that the hidden service does not require client authentication and proceed accordingly. In the case the Ed25519 is novel to the client it will have to resign and republish its hidden service descriptor. It SHOULD wait some (indeterminate) amount of time for the new descriptor to propagate the Tor hidden service directory servers, before. This should take no more than a few minutes. CAs MUST NOT expire challenges before a reasonable time to allow publication of the new descriptor (this document suggests at least 30 minutes).

Certification Authority Authorization (CAA) ".onion" domains are not part of the DNS, and as such a variation on CAA is required to allow restrictions to be placed on certificate issuance. To this end a new field is added to the second layer hidden service descriptor § 2.5.2.2. with the following format: "caa" SP flags SP tag SP value NL [Any number of times] The contents of "flag", "tag", and "value" are as per § 4.1.1. Multiple CAA records may be present, as is the case in the DNS. CAA records in a hidden service descriptor are to be treated the same by CAs as if they had been in the DNS for the ".onion" domain. A hidden service's second layer descriptor using CAA may look something like the following: create2-formats 2 single-onion-service caa 0 issue "example.com" caa 0 iodef "mailto:security@example.com" caa 128 validationmethods "onion-csr-01" introduction-point AwAGsAk5nSMpAhRqhMHbTFCTSlfhP8f5PqUhe6DatgMgk7kSL3KHCZUZ3C6tXDeRfM9SyNY0DlgbF8q+QSaGKCs= ...

Relevant Resource Record Set In the absence of the possibility of delegation from a ".onion" domain as there is in the DNS there is no need, nor indeed any possibility to search up a the DNS tree for a relevant CAA record set. Instead all subdomains under a ".onion" domain share the same CAA record set. That is all of these share a CAA record set with "a.onion":

• b.a.onion
• c.a.onion
• e.d.a.onion

But these do not:

• b.c.onion
• c.d.onion
• e.c.d.onion

When to check CAA If the hidden service has client authentication enabled then it will be impossible for the CAA to decrypt the second layer descriptor to read the CAA records until the CAAs public key has been added to first layer descriptor. To this end a CA SHOULD wait until the client responds to an authorization, and treat this as indication that their public key has been added and that the CA will be able to decrypt the second layer descriptor.

Preventing mis-issuance by unknown CAs As the CAA records are in the second layer descriptor and in the case of a hidden service requiring client authentication it is impossible to read them without the hidden service trusting a CA's public key, a method is required to signal that there are CAA records present (but not reveal their contents, which may disclose unwanted information about the hidden service operator). To this end a new field is added to the first layer hidden service descriptor § 2.5.1.2. with the following format: "caa-critical" NL [At most once] If a CA encounters this flag it MUST NOT proceed with issuance until it can decrypt and parse the CAA records from the second layer descriptor.

IANA Considerations

Validation Methods Per this document, one new entry has been added to the "ACME Validation Methods" registry defined in §9.7.8. This entry is defined below: New entries

Label	Identifier Type	ACME	Reference
onion-csr-01	dns	Y	This document

Security Considerations

Use of "dns" identifier type The re-use of the "dns" identifier type for a domain not actually in the DNS infrastructure raises questions regarding its suitability. The reasons the author wish to pursue this path in the first place are detailed in . It is felt that there is little security concern in reuse of the "dns" identifier type with regards the mis-issuance by CAs that are not aware of ".onion" domains.

"http-01" Challenge The CA would follow the procedure set out in §8.3 which specifies that the CA should "Dereference the URL using an HTTP GET request". Given that ".onion" require special handling to dereference, this de-referencing will fail, disallowing issuance.

"tls-alpn-01" Challenge The CA would follow the procedure set out in §3 which specifies that the CA "resolves the domain name being validated and chooses one of the IP addresses returned for validation". Given that ".onion" are not resolvable to IP addresses, this de-referencing will fail, disallowing issuance.

"dns-01" Challenge The CA would follow the procedure set out in §8.4 which specifies that the CA should "query for TXT records for the validation domain name". Given that ".onion" are not present in the DNS infrastructure, this query will fail, disallowing issuance.

Key Authorization with "onion-csr-01" The "onion-csr-01" challenge does not make use of the key authorization string defined in §8.1. This does not weaken the integrity of authorizations. The key authorization exists to ensure that an attacker observing the validation channel can observe the correct validation response, but cannot compromise the integrity of authorizations as the response only be used with the account key for which it was generated. As the validation channel for this challenge is ACME itself, and ACME already requires that the request be signed by the account, the key authorization is not required.

Use of Tor for non ".onion" domains An ACME server MUST NOT utilise Tor for the validation of non ".onion" domains, due to the risk of possible exit hijacking.

Security of CAA records The second layer descriptor is encrypted and MACed in a way that only a party with access to the secret key of the hidden service could manipulate what is published there. For more information about this process see § 2.5.3.

Access of the Tor network The ACME server MUST make its own connection to the hidden service via the Tor network, and MUST NOT outsource this, such as by using Tor2Web.

References Normative References The Tor Project The Tor Project CA/Browser Forum Informative References The Tor Project

Discussion on the use of the "dns" identifier type The reasons for utilising the "dns" identifier type in ACME and not defining a new identifier type for ".onion" domains may not seem obvious at first glance. After all, ".onion" domains are not part of the DNS infrastructure and as such why should they use the "dns" identifier type? The CA/Browser Forum Baseline Requirements §B.2.a.ii define, and this standard allows, using the "http-01" or "tls-alpn-01" validation methods already present in ACME (with some considerations). Given the situation of a web server placed behind a Tor terminating proxy (as per the setup suggested by the Tor project ), existing ACME tooling can be blind to the fact that a ".onion" domain is being utilised, as they simply receive an incoming TCP connection as they would regardless (albeit from the Tor terminating proxy). An example of this would be Certbot placing the ACME challenge response file in the webroot of an NGINX web server. Neither Certbot nor NGINX would require any modification to be aware of any special handling for ".onion" domains. This does raise some questions regarding security within existing implementations, however the authors believe this is of little concern, as per .

Acknowledgements With thanks to the Open Technology Fund for funding the work that went into this document. The authors also wish to thank the following for their input on this document:

• Iain R. Learmonth
• Jan-Frederik Rieckers