]> Trinity College Dublin

Dublin 2 Ireland +353-1-896-2354 stephen.farrell@cs.tcd.ie

VPN Consortium

paul.hoffman@vpnc.org

Phresheez

mike@phresheez.com

IETF http authentication origin-bound key HTTP Origin-Bound Authentication (HOBA) is a digital-signature-based design for an HTTP authentication method. The design can also be used in JavaScript-based authentication embedded in HTML. HOBA is an alternative to HTTP authentication schemes that require passwords and therefore avoids all problems related to passwords, such as leakage of server-side password databases.

HTTP Origin-Bound Authentication (HOBA) is an authentication design that can be used as an HTTP authentication scheme and for JavaScript-based authentication embedded in HTML. The main goal of HOBA is to offer an easy-to-implement authentication scheme that is not based on passwords but that can easily replace HTTP or HTML forms-based password authentication. Deployment of HOBA can reduce or eliminate password entries in databases, with potentially significant security benefits. HOBA is an HTTP authentication mechanism that complies with the framework for such schemes . As a JavaScript design, HOBA demonstrates a way for clients and servers to interact using the same credentials that are used by the HTTP authentication scheme. Current username/password authentication methods such as HTTP Basic, HTTP Digest, and web forms have been in use for many years but are susceptible to theft of server-side password databases. Instead of passwords, HOBA uses digital signatures in a challenge-response scheme as its authentication mechanism. HOBA also adds useful features such as credential management and session logout. In HOBA, the client creates a new public-private key pair for each host ("web origin" ) to which it authenticates. These keys are used in HOBA for HTTP clients to authenticate themselves to servers in the HTTP protocol or in a JavaScript authentication program. HOBA session management is identical to username/password session management, with a server-side session management tool or script inserting a session cookie into the output to the browser. Use of Transport Layer Security (TLS) for the HTTP session is still necessary to prevent session cookie hijacking. HOBA keys are "bare keys", so there is no need for the semantic overhead of X.509 public key certificates, particularly with respect to naming and trust anchors. The Client Public Key (CPK) structures in HOBA do not have any publicly visible identifier for the user who possesses the corresponding private key, nor the web origin with which the client is using the CPK. HOBA also defines some services that are needed for modern HTTP authentication: Servers can bind a CPK with an identifier, such as an account name. Servers using HOBA define their own policies for binding CPKs with accounts during account registration. Users are likely to use more than one device or User Agent (UA) for the same HTTP-based service, so HOBA gives a way to associate more than one CPK to the same account without having to register for each separately. Logout features can be useful for UAs, so HOBA defines a way to close a current HTTP "session". Digital signatures can be expensive to compute, so HOBA defines a way for HTTP servers to indicate how long a given challenge value is valid, and a way for UAs to fetch a fresh challenge at any time. Users are also likely to lose a private key, or the client's memory of which key pair is associated with which origin, such as when a user loses the computer or mobile device in which state is stored. HOBA does not define a mechanism for deleting the association between an existing CPK and an account. Such a mechanism can be implemented at the application layer. The HOBA scheme is far from new; for example, the basic idea is pretty much identical to the first two messages from "Mechanism R" on page 6 of , which predates HOBA by 20 years.

HOBA can be used as a drop-in replacement for password-based user authentication schemes used in common web applications. The simplest way is to (re)direct the UA to a HOBA "Login" URL and for the response to a successful HTTP request containing a HOBA signature to set a session cookie . Further interactions with the web application will then be secured via the session cookie, as is commonly done today. While cookies are bearer tokens, and thus weaker than HOBA signatures, they are currently ubiquitously used. If non-bearer token session continuation schemes are developed in the future in the IETF or elsewhere, then those can interface to HOBA as easily as with any password-based authentication scheme.

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. This specification uses the Augmented Backus-Naur Form (ABNF) notation of . Account: The term "account" is (loosely) used to refer to whatever data structure(s) the server maintains that are associated with an identity. That will contain at least one CPK and a web origin; it will also optionally include an HTTP "realm" as defined in the HTTP authentication specification . It might also involve many other non-standard pieces of data that the server accumulates as part of account creation processes. An account may have many CPKs that are considered equivalent in terms of being usable for authentication, but the meaning of "equivalent" is really up to the server and is not defined here. Client public key (CPK): A CPK is the public key and associated cryptographic parameters needed for a server to validate a signature. HOBA-http: We use this term when describing something that is specific to HOBA as an HTTP authentication mechanism. HOBA-js: We use this term when describing something that is unrelated to HOBA-http but is relevant for HOBA as a design pattern that can be implemented in a browser in JavaScript. User agent (UA): typically, but not always, a web browser. User: a person who is running a UA. In this document, "user" does not mean "user name" or "account name". Web client: the content and JavaScript code that run within the context of a single UA instance (such as a tab in a web browser).

Step-by-step, a typical HOBA-http registration and authentication flow might look like this: The client connects to the server and makes a request, and the server's response includes a WWW-Authenticate header field that contains the "HOBA" auth-scheme, along with associated parameters (see ). If the client was not already registered with the web origin and realm it is trying to access, the "joining" process is invoked (see ). This creates a key pair and makes the CPK known to the server so that the server can carry out the account creation processes required. The client uses the challenge from the HOBA auth-scheme parameters, along with other information it knows about the web origin and realm, to create and sign a HOBA to-be-signed (HOBA-TBS) string (see ). The client creates a HOBA client-result (HOBA-RES), using the signed HOBA-TBS for the "sig" value (see ). The client includes the Authorization header field in its next request, using the "HOBA" auth-scheme and putting the HOBA client-result in an auth-param named "result" (see ). The server authenticates the HOBA client-result (see ). Typically, the server's response includes a session cookie that allows the client to indicate its authentication state in future requests (see ).

A UA that implements HOBA maintains a list of web origins and realms. The UA also maintains one or more client credentials for each web origin/realm combination for which it has created a CPK. On receipt of a challenge (and optional realm) from a server, the client marshals a HOBA-TBS blob that includes a client generated nonce, the web origin, the realm, an identifier for the CPK, and the challenge string, and signs that blob with the private key corresponding to the CPK for that web origin. The formatting chosen for this TBS blob is chosen so as to make server-side signature verification as simple as possible for a wide range of current server tooling. specifies the ABNF for the signature input. The term "unreserved" means that the field does not have a specific format defined and allows the characters specified in Section 2.3 of .

HOBA-TBS = len ":" nonce len ":" alg len ":" origin len ":" [ realm ] len ":" kid len ":" challenge len = 1*DIGIT nonce = 1*base64urlchars alg = 1*2DIGIT origin = scheme "://" authority ":" port ; scheme, etc., are from RFC 3986 realm = unreserved ; realm is to be treated as in Section 2.2 of RFC 7235 kid = 1*base64urlchars challenge = 1*base64urlchars ; Characters for Base64URL encoding from Table 2 of RFC 4648 ; all of which are US-ASCII (see RFC 20) base64urlchars = %x30-39 ; Digits / %x41-5A ; Uppercase letters / %x61-7A ; Lowercase letters / "-" / "_" / "=" ; Special characters

The fields above contain the following: len: Each field is preceded by the number of octets of the following field, expressed as a decimal number in ASCII . Lengths are separated from field values by a colon character. So if a nonce with the value "ABCD" were used, then that would be preceeded by "4:" (see the example in for details). nonce: a random value chosen by the UA and MUST be base64url encoded before being included in the HOBA-TBS value. (base64url encoding is defined in ; guidelines for randomness are given in .) UAs MUST be able to use at least 32 bits of randomness in generating a nonce. UAs SHOULD be able to use 64 or more bits of randomness for nonces. alg: specifies the signature algorithm being used. See for details of algorithm support requirements. The IANA-registered algorithm values (see ) are encoded as one- or two-digit ASCII numbers. For example, RSA-SHA256 (number 0) is encoded as the ASCII character "0" (0x30), while a future algorithm registered as number 17 would be encoded as the ASCII characters "17" (0x3137). origin: the web origin expressed as the concatenation of the scheme, authority, and port from . These are not base64 encoded, as they will be most readily available to the server in plain text. For example, if accessing the URL "https://www.example.com:8080/foo", then the bytes input to the signature process will be "https://www.example.com:8080". There is no default for the port number, and the port number MUST be present. realm: a string with the syntactic restrictions defined in . If no realm is specified for this authentication, then this is absent but is preceeded by a length of zero ("0:"). Recall that both sides know when this needs to be there, independent of the encoding via a zero length. kid: a key identifier. This MUST be a base64url-encoded value that is presented to the server in the HOBA client result (see below). challenge: MUST be a base64url-encoded challenge value that the server chose to send to the client. The challenge MUST be chosen so that it is infeasible to guess and SHOULD be indistinguishable from (the base64url encoding of) a random string that is at least 128 bits long. The HOBA-TBS string is the input to the client's signing process but is not itself sent over the network since some fields are already inherent in the HTTP exchange. The challenge, however, is sent over the network so as to reduce the amount of state that needs to be maintained by servers. (One form of stateless challenge might be a ciphertext that the server decrypts and checks, but that is an implementation detail.) The value that is sent over the network by the UA is the HOBA "client result", which we now define. The HOBA "client result" is a dot-separated string that includes the signature and is sent in the HTTP Authorization header field value using the value syntax defined in . The "sig" value is the base64url-encoded version of the binary output of the signing process. The kid, challenge, and nonce are as defined above and are also base64url encoded.

HOBA-RES = kid "." challenge "." nonce "." sig sig = 1*base64urlchars

If a malformed message of any kind is received by a server, the server MUST fail authentication. If a malformed message of any kind is received by a client, the client MUST abandon that authentication attempt. (The client is, of course, free to start another authentication attempt if it desires.)

An HTTP server that supports HOBA authentication includes the "HOBA" auth-scheme value in a WWW-Authenticate header field when it wants the client to authenticate with HOBA. Note that the HOBA auth-scheme might not be the only one that the server includes in a WWW-Authenticate header. The HOBA scheme has two REQUIRED attributes (challenge and max-age) and one OPTIONAL attribute (realm): The "challenge" attribute MUST be included. The challenge is the string made up of the base64url-encoded octets that the server wants the client to sign in its response. The challenge MUST be unique for every 401 HTTP response in order to prevent replay attacks from passive observers. A "max-age" attribute MUST be included. It specifies the number of seconds from the time the HTTP response is emitted for which responses to this challenge can be accepted; for example, "max-age: 10" would indicate ten seconds. If max-age is set to zero, then that means that only one signature will be accepted for this challenge. A "realm" attribute MAY be included to indicate the scope of protection in the manner described in HTTP/1.1, Authentication . The "realm" attribute MUST NOT appear more than once. When the "client response" is created, the UA encodes the HOBA client-result and returns that in the Authorization header. The client-result is a string matching the HOBA-RES production in as an auth-param with the name "result". The server MUST check the cryptographic correctness of the signature based on a public key it knows for the kid in the signatures, and if the server cannot do that, or if the signature fails cryptographic checks, then validation has failed. The server can use any additional mechanisms to validate the signature. If the validation fails, or if the server chooses to reject the signature for any reason whatsoever, the server fails the request with a 401 Unauthorized HTTP response. The server MUST check that the same web origin is used in all of the server's TLS server certificates, the URL being accessed, and the HOBA signature. If any of those checks fail, the server treats the signature as being cryptographically incorrect. Note that a HOBA signature is good for however long a non-zero max-age parameter allows. This means that replay is possible within the time window specified by the "max-age" value chosen by the server. Servers can attempt to detect any such replay (via caching if they so choose) and MAY react to such replays by responding with a second (or subsequent) 401 HTTP response containing a new challenge. To optimize their use of challenges, UAs MAY prefetch a challenge value, for example, after (max-age)/2 seconds have elapsed, using the ".well-known/hoba/getchal" scheme described later in this document. This also allows for precalculation of HOBA signatures, if that is required in order to produce a responsive user interface.

Web sites using JavaScript can also perform origin-bound authentication without needing to involve the HTTP layer and by inference not needing HOBA-http support in browsers. HOBA-js is not an on-the-wire protocol like HOBA-http is; instead, it is a design pattern that can be realized completely in JavaScript served in normal HTML pages. One thing that is highly desirable for HOBA-js is WebCrypto (see <http://www.w3.org/TR/WebCryptoAPI>), which is (at the time of writing) starting to see deployment. In lieu of WebCrypto, JavaScript crypto libraries can be employed with the known deficiencies of their pseudo-random number generators and the general immaturity of those libraries. Without Webcrypto, one element is required for HOBA-js; localStorage (see <http://www.w3.org/TR/webstorage/>) from HTML5 can be used for persistent key storage. For example, an implementation would store a dictionary account identifier as well as public key and private key tuples in the origin's localStorage for subsequent authentication requests. How this information is actually stored in localStorage is an implementation detail. This type of key storage relies on the security properties of the same-origin policy that localStorage enforces. See the security considerations for discussion about attacks on localStorage. Note that IndexedDB (see <http://www.w3.org/TR/IndexedDB/>) is an alternative to localStorage that can also be used here and that is used by WebCrypto. Because of JavaScript's same-origin policy, scripts from subdomains do not have access to the same localStorage that scripts in their parent domains do. For larger or more complex sites, this could be an issue that requires enrollment into subdomains, which could be difficult for users. One way to get around this is to use session cookies because they can be used across subdomains. That is, with HOBA-js, the user might log in using a single well-known domain, and then session cookies are used whilst the user navigates around the site.

This section describes how clients and servers use HOBA for authentication. The interaction between an HTTP client and HTTP server using HOBA happens in three phases: the CPK preparation phase, the signing phase, and the authentication phase. This section also covers the actions that give HOBA features similar to today's password-based schemes.

In the CPK preparation phase, the client determines if it already has a CPK for the web origin with which it needs to authenticate. If the client has a CPK, the client will use it; if the client does not have a CPK, it generates one in anticipation of the server asking for one.

In the signing phase, the client connects to the server, the server asks for HOBA-based authentication, and the client authenticates by signing a blob of information as described in the previous sections.

The authentication phase is completely dependent on the policies and practices of the server. That is, this phase involves no standardized protocol in HOBA-http; in HOBA-js, there is no suggested interaction template. In the authentication phase, the server uses the key identifier (kid) to determine the CPK from the signing phase and decides if it recognizes the CPK. If the server recognizes the CPK, the server may finish the client authentication process. If this stage of the process involves additional information for authentication, such as asking the user which account she wants to use (in the case where a UA is used for multiple accounts on a site), the server can prompt the user for account identifying information, or the user could choose based on HTML offered by the server before the 401 response is triggered. None of this is standardized: it all follows the server's security policy and session flow. At the end of this, the server probably assigns or updates a session cookie for the client. During the authentication phase, if the server cannot determine the correct CPK, it could use HTML and JavaScript to ask the user if they are really a new user or want to associate this new CPK with another CPK. The server can then use some out-of-band method (such as a confirmation email round trip, SMS, or a UA that is already enrolled) to verify that the "new" user is the same as the already-enrolled one. Thus, logging in on a new UA is identical to logging in with an existing account. If the server does not recognize the CPK, the server might send the client through either a join or login-new-UA (see below) process. This process is completely up to the server and probably entails using HTML and JavaScript to ask the user some questions in order to assess whether or not the server wants to give the client an account. Completion of the joining process might require confirmation by email, SMS, CAPTCHA, and so on. Note that there is no necessity for the server to initiate a joining or login process upon completion of the signing phase. Indeed, the server may desire to challenge the UA even for unprotected resources and set a session cookie for later use in a join or login process as it becomes necessary. For example, a server might only want to offer an account to someone who had been to a few pages on the web site; in such a case, the server could use the CPK from an associated session cookie as a way of building reputation for the user until the server wants the user to join.

The authentication process is more than just the act of authentication. In password-based authentication and HOBA, there are other processes that are needed both before and after an authentication step. This section covers those processes. Where possible, it combines practices of HOBA-http and HOBA-js; where that is not possible, the differences are called out. All HOBA interactions other than those defined in MUST be performed in TLS-protected sessions . If the current HTTP traffic is not running under TLS, a new session is started before any of the actions described here are performed. HOBA-http uses a well-known URI "hoba" as a base URI for performing many tasks: "https://www.example.com/.well-known/hoba". These URIs are based on the name of the host that the HTTP client is accessing. There are many use cases for these URLs to redirect to other URLs: a site that does registration through a federated site, a site that only does registration under HTTPS, and so on. Like any HTTP client, HOBA-http clients have to be able to handle redirection of these requests. However, as that would potentially cause security issues when a re-direct brings the client to a different web origin, servers implementing HOBA-http SHOULD NOT redirect to a different web origin from below ".well-known/hoba" URLs. The above is considered sufficient to allow experimentation with HOBA, but if at some point HOBA is placed on the Standards Track, then a full analysis of off-origin redirections would need to be documented.

Normally, a registration (also called "joining") is expected to happen after a UA receives a 401 response for a web origin and realm (for HOBA-http) or on demand (for HOBA-js) for which it has no associated CPK. The process of registration for a HOBA account on a server is relatively lightweight. The UA generates a new key pair and associates it with the web origin/realm in question. Note that if the UA has a CPK associated with the web origin, but not for the realm concerned, then a new registration is REQUIRED. If the server did not wish for that outcome, then it ought to use the same or no realm. The registration message for HOBA-http is sent as a POST message to the URL ".well-known/hoba/register" with an HTML form (x-www-form-encoded, see <http://www.w3.org/TR/2014/REC-html5-20141028/forms.html#url-encoded-form-data>), described below. The registration message for HOBA-js can be in any format specified by the server, but it could be the same as the one described here for HOBA-http. It is up to the server to decide what kind of user interaction is required before the account is finally set up. When the server's chosen registration flow is completed successfully, the server MUST add a Hobareg HTTP header (see ) to the HTTP response message that completes the registration flow. The registration message sent to the server has one mandatory field (pub) and some optional fields that allow the UA to specify the type and value of key and device identifiers that the UA wishes to use. pub: a mandatory field containing the Privacy Enhanced Mail (PEM) formatted public key of the client. See Appendix C of for an example of how to generate this key format. kidtype: contains the type of key identifier. This is a numeric value intended to contain one of the values from . If this is not present, then the mandatory-to-implement hashed public key option MUST be used. kid: contains the key identifier as a base64url-encoded string that is of the type indicated in the kidtype. If the kid is a hash of a public key, then the correct (base64url-encoded) hash value MUST be provided and the server SHOULD check that and refuse the registration if an incorrect value was supplied. didtype: specifies a kind of device identifier intended to contain one of the values from . If absent, then the "string" form of device identifier defined in MUST be used. did: a UTF-8 string that specifies the device identifier. This can be used to help a user be confident that authentication has worked, e.g., following authentication, some web content might say "You last logged in from device 'did' at time T." Note that replay of registration (and other HOBA) messages is quite possible. That, however, can be counteracted if challenge freshness is ensured. See for details. Note also that with HOBA-http, the HOBA signature does not cover the POST message body. If that is required, then HOBA-JS may be a better fit for registration and other account management actions.

Since registration can often be a multi-step process, e.g., requiring a user to fill in contact details, the initial response to the HTTP POST message defined above may not be the end of the registration process even though the HTTP response has a 200 OK status. This creates an issue for the UA since, during the registration process (e.g., while dealing with interstitial pages), the UA doesn't yet know whether the CPK is good for that web origin or not. For this reason, the server MUST add a header field to the response message when the registration has succeeded in order to indicate the new state. The header to be used is "Hobareg", and the value when registration has succeeded is to be "regok". When registration is in an intermediate state (e.g., on an HTTP response for an interstitial page), the server MAY add this header with a value of "reginwork". See for the relevant IANA registration of this header field. For interstitial pages, the client MAY include a HOBA Authorization header. This is not considered a "MUST", as that might needlessly complicate client implementations, but is noted here in case a server implementer assumes that all registration messages contain a HOBA Authorization header.

Hobareg-val = "regok" / "reginwork"

provides an ABNF definition for the values allowed in the Hobareg header field. Note that these (and the header field name) are case insensitive. Section 8.3.1 of calls for documenting the following details for this new header field: Only one single value is allowed in a Hobareg header field. Should more than one (a list) be encountered, or any other ABNF-invalid value, that SHOULD be interpreted as being the same as "reginwork". The Hobareg header field can only be used in HTTP responses. Since Hobareg is only meant for responses, it ought not appear in requests. The HTTP response code does affect the interpretation of Hobareg. Registration is only considered to have succeeded if the regok value is seen in a 2xx response. 4xx and other errors mean that registration has failed regardless of the value of Hobareg seen. The request method has no influence on the interpretation of Hobareg. Intermediaries never insert, delete, or modify a Hobareg header field. As a response-only header field, it is not appropriate to list a Hobareg in a Vary response header field. Hobareg is allowed in trailers. As a response-only header field, Hobareg will not be preserved across re-directs. Hobareg itself discloses little security- or privacy-sensitive information. If an attacker can somehow detect that a Hobareg header field is being added, then that attacker would know that the UA is in the process of registration, which could be significant. However, it is likely that the set of messages between the UA and server would expose this information in many cases, regardless of whether or not TLS is used. Using TLS is still, however, a good plan.

From the user perspective, the UA having a CPK for a web origin will often appear to be the same as having a way to sign in to an account at that web site. Since users often have more than one UA, and since the CPKs are, in general, UA specific, that raises the question of how the user can sign in to that account from different UAs. And from the server perspective, that turns into the question of how to safely bind different CPKs to one account. In this section, we describe some ways in which this can be done, as well as one way in which this ought not be done. Note that the context here is usually that the user has succeeded in registering with one or more UAs (for the purposes of this section, we call this "the first UA" below) and can use HOBA with those, and the user is now adding another UA. The newest UA might or might not have a CPK for the site in question. Since it is in fact trivial, we assume that the site is able to put in place some appropriate, quicker, easier registration for a CPK for the newest UA. The issue then becomes one of binding the CPK from the newest UA with those of other UAs bound to the account.

It is common for a user to have multiple UAs and to want all those UAs to be able to authenticate to a single account. One method to allow a user who has an existing account to be able to authenticate on a second device is to securely transport the private and public keys and the origin information from the first device to the second. If this approach is taken, then there is no impact on the HOBA-http or HOBA-js, so this is a pure UA implementation issue and not discussed further.

It will be tempting for implementers to use a human-memorable One-Time Password (OTP) in order to "authenticate" binding CPKs to the same account. The workflow here would likely be something along the lines of some server administrative utility generating a human-memorable OTP such as "1234" and sending that to the user out of band for the user to enter at two web pages, each authenticated via the relevant CPK. While this seems obvious enough and could even be secure enough in some limited cases, we consider that this is too risky to use in the Internet, and so servers SHOULD NOT provide such a mechanism. The reason this is so dangerous is that it would be trivial for an automated client to guess such tokens and "steal" the binding intended for some other user. At any scale, there would always be some in-process bindings so that even with only a trickle of guesses (and hence not being detectable via message volume), an attacker would have a high probability of succeeding in registering a binding with the attacker's CPK. This method of binding CPKs together is therefore NOT RECOMMENDED.

One easy binding method is to simply provide a web page where, using the first UA, the user can generate a URL (containing some "unguessable" cryptographically generated value) that the user then later dereferences on the newest UA. The user could email that URL to herself, for example, or the web server accessed at the first UA could automatically do that. Such a URL SHOULD contain at least the equivalent of 128 bits of randomness.

The user can tell the server it wishes to log out. With HOBA-http, this is done by sending a HOBA-authenticated POST message to the URL ".well-known/hoba/logout" on the site in question. The UA SHOULD also delete session cookies associated with the session so that the user's state is no longer "logged in." The server MUST NOT allow TLS session resumption for any logged out session. The server SHOULD also revoke or delete any cookies associated with the session.

The UA can get a "fresh" challenge from the server. In HOBA-http, it sends a POST message to ".well-known/hoba/getchal". If successful, the response MUST contain a fresh (base64url-encoded) HOBA challenge for this origin in the body of the response. Whitespace in the response MUST be ignored.

RSA-SHA256 MUST be supported. HOBA implementations MUST use RSA-SHA256 if it is provided by the underlying cryptographic libraries. RSA-SHA1 MAY be used. RSA modulus lengths of at least 2048 bits SHOULD be used. RSA indicates the RSASSA-PKCS1-v1_5 algorithm defined in Section 8.2 of , and SHA-1 and SHA-256 are defined in . Keys with moduli shorter than 2048 bits SHOULD only be used in cases where generating 2048-bit (or longer) keys is impractical, e.g., on very constrained or old devices.

Binding my CPK with someone else's account would be fun and profitable so SHOULD be appropriately hard. In particular, URLs or other values generated by the server as part of any CPK binding process MUST be hard to guess, for whatever level of difficulty is chosen by the server. The server SHOULD NOT allow a random guess to reveal whether or not an account exists. If key binding was server selected, then a bad actor could bind different accounts belonging to the user from the network with possible bad consequences, especially if one of the private keys was compromised somehow. When the max-age parameter is not zero, then a HOBA signature has a property that is like a bearer token for the relevant number of seconds: it can be replayed for a server-selected duration. Similarly, for HOBA-js, signatures might be replayable depending on the specific implementation. The security considerations of therefore apply in any case where the HOBA signature can be replayed. Server administrators can set the max-age to the minimum acceptable value in such cases, which would often be expected to be just a few seconds. There seems to be no reason to ever set the max-age more than a few minutes; the value ought also decrease over time as device capabilities improve. The administrator will most likely want to set the max-age to something that is not too short for the slowest signing device that is significant for that site.

HOBA does, to some extent, impact privacy and could be considered to represent a super-cookie to the server or to any entity on the path from UA to HTTP server that can see the HOBA signature. This is because we need to send a key identifier as part of the signature and that will not vary for a given key. For this reason, and others, it is strongly RECOMMENDED to only use HOBA over server-authenticated TLS and to migrate web sites using HOBA to only use "https" URLs. UAs SHOULD provide users a way to manage their CPKs. Ideally, there would be a way for a user to maintain their HOBA details for a site while at the same time deleting other site information such as cookies or non-HOBA HTML5 localStorage. However, as this is likely to be complex, and appropriate user interfaces counterintuitive, we expect that UAs that implement HOBA will likely treat HOBA information as just some more site data that would disappear should the user choose to "forget" that site. Device identifiers are intended to specify classes of device in a way that can assist with registration and with presentation to the user of information about previous sessions, e.g., last login time. Device identifier types MUST NOT be privacy sensitive, with values that would allow tracking a user in unexpected ways. In particular, using a device identifier type that is analogous to the International Mobile Equipment Identifier (IMEI) would be a really bad idea and is the reason for the "MUST NOT" above. In that case, "mobile phone" could be an acceptable choice. If possible, implementations ought to encourage the use of device identifier values that are not personally identifying except for the user concerned; for example, "Alice's mobile" is likely to be chosen and is somewhat identifying, but "Alice's phone: UUID 1234-5567-89abc-def0" would be a very bad choice.

The use of localStorage (likely with a non-WebCrypto implementation of HOBA-js) will undoubtedly be a cause for concern. localStorage uses the same-origin model that says that the scheme, domain, and port define a localStorage instance. Beyond that, any code executing will have access to private keying material. Of particular concern are Cross-Site Scripting (XSS) attacks, which could conceivably take the keying material and use it to create UAs under the control of an attacker. XSS attacks are, in reality, devastating across the board since they can and do steal credit card information, passwords, perform illicit acts, etc. It's not evident that we are introducing unique threats from which cleartext passwords don't already suffer. Another source of concern is local access to the keys. That is, if an attacker has access to the UA itself, they could snoop on the key through a JavaScript console or find the file(s) that implement localStorage on the host computer. Again, it's not clear that we are worse in this regard because the same attacker could get at browser password files, etc., too. One possible mitigation is to encrypt the keystore with a password/PIN that the user supplies. This may sound counterintuitive, but the object here is to keep passwords off of servers to mitigate the multiplier effect of a large-scale compromise (e.g., ) because of shared passwords across sites. It's worth noting that HOBA uses asymmetric keys and not passwords when evaluating threats. As various password database leaks have shown, the real threat of a password breach is not just to the site that was breached, it's also to all of the sites on which a user used the same password. That is, the collateral damage is severe because password reuse is common. Storing a password in localStorage would also have a similar multiplier effect for an attacker, though perhaps on a smaller scale than a server-side compromise: one successful crack gains the attacker potential access to hundreds if not thousands of sites the user visits. HOBA does not suffer from that attack multiplier since each asymmetric key pair is unique per site/UA/user.

A shared UA with multiple accounts is possible if the account identifier is stored along with the asymmetric key pair binding them to one another. Multiple entries can be kept, one for each account, and selected by the current user. This, of course, is fraught with the possibility for abuse, since a server is potentially enrolling the device for a long period and the user may not want to have to be responsible for the credential for that long. To alleviate this problem, the user could request that the credential be erased from the browser. Similarly, during the enrollment phase, a user could request that the key pair only be kept for a certain amount of time or that it not be stored beyond the current browser session. However, all such features really ought to be part of the operating system or platform and not part of a HOBA implementation, so those are not discussed further.

The repeated length fields in the HOBA-TBS structure are present in order to ensure that there is no possibility that the catenation of different input values can cause confusion that might lead to an attack, either against HOBA as specified here, or else an attack against some other protocol that reused this to-be-signed structure. Those fields ensure that the mapping from input fields to the HOBA-TBS string is an injective mapping.

IANA has made registrations and created new registries as described below. All new registries have been placed beneath a new "HTTP Origin-Bound Authentication (HOBA) Parameters" category.

A new scheme has been registered in the HTTP Authentication Scheme Registry as follows: Authentication Scheme Name: HOBA Reference: of RFC 7486 Notes (optional): The HOBA scheme can be used with either HTTP servers or proxies. When used in response to a 407 Proxy Authentication Required indication, the appropriate proxy authentication header fields are used instead, as with any other HTTP authentication scheme.

A new .well-known URI has been registered in the Well-Known URIs registry as described below. URI Suffix: hoba Change Controller: IETF Reference: of RFC 7486 Related Information: N/A

A new HOBA signature algorithms registry has been created as follows, with Specification Required as the registration procedure. New HOBA signature algorithms SHOULD be in use with other IETF Standards Track protocols before being added to this registry.

RSA is defined in Section 8.2 of , and SHA-1 and SHA-256 are defined in . For this registry, the number column should contain a small positive integer. Following the ABNF in , the maximum value for this is decimal 99.

A new HOBA Key Identifier Types registry has been created as follows, with Specification Required as the registration procedure.

For the number 0, hashed public keys are as done in DNS-Based Authentication of Named Entities (DANE) . For this registry, the number column should contain a small positive integer.

A new HOBA Device Identifier Types registry has been created as follows, with Specification Required as the registration procedure. The designated expert for this registry is to carefully pay attention to the notes on this field in , in particular, the "MUST NOT" stated therein.

For this registry, the number column should contain a small positive integer.

A new identifier has been registered in the Permanent Message Header Field Names registry as described below. Header Field Name: Hobareg Protocol: http (RFC 7230) Status: experimental Author/Change controller: IETF Reference: of RFC 7486 Related information: N/A

&RFC2119; &RFC3447; &RFC3986; &RFC4648; &RFC5234; &RFC5246; &RFC5785; &RFC6454; &RFC6698; &RFC6750; &RFC7231; &RFC7235; NIST &RFC4086; &RFC6265; &RFC6376; Sophos

By far, the most common mechanism for web authentication is passwords that can be remembered by the user, called "human-memorable passwords". There is plenty of good research on how users typically use human-memorable passwords (e.g., see ), but some of the highlights are that users typically try hard to reuse passwords on as many web sites as possible, and that web sites often use either email addresses or users' names as the identifiers that go with these passwords. If an attacker gets access to the database of memorizable passwords, that attacker can impersonate any of the users. Even if the breach is discovered, the attacker can still impersonate users until every password is changed. Even if all the passwords are changed or at least made unusable, the attacker now possesses a list of likely username/password pairs that might exist on other sites. Using memorizable passwords on unencrypted channels also poses risks to the users. If a web site uses either the HTTP Basic authentication method, or an HTML form that does no cryptographic protection of the password in transit, a passive attacker can see the password and immediately impersonate the user. If a hash-based authentication scheme such as HTTP Digest authentication is used, a passive attacker still has a high chance of being able to determine the password using a dictionary of known passwords. Note that passwords that are not human-memorable are still subject to database attack, though they are of course unlikely to be reused across many systems. Similarly, database attacks of some form or other will work against any password-based authentication scheme, regardless of the cryptographic protocol used. So for example, zero-knowledge or Password-Authenticated Key Exchange (PAKE) schemes, though making use of elegant cryptographic protocols, remain as vulnerable to what is clearly the most common exploit seen when it comes to passwords. HOBA is, however, not vulnerable to database theft.

The following values show an example of HOBA-http authentication to the origin "https://example.com:443". Carriage returns have been added and need to be removed to validate the example. Public Key:

Origin: https://example.com:443 Key Identifier: vesscamS2Kze4FFOg3e2UyCJPhuQ6_3_gzN-k_L6t3w Challenge: pUE77w0LylHypHKhBqAiQHuGC751GiOVv4/7pSlo9jc= Signature algorithm: RSA-SHA256 ("0") Nonce: Pm3yUW-sW5Q Signature:

Authorization Header:

Thanks to the following for good comments received during the preparation of this specification: Richard Barnes, David Black, Alissa Cooper, Donald Eastlake, Amos Jeffries, Benjamin Kaduk, Watson Ladd, Barry Leiba, Matt Lepinski, Ilari Liusvaara, James Manger, Alexey Melnikov, Kathleen Moriarty, Yoav Nir, Mark Nottingham, Julian Reschke, Pete Resnick, Michael Richardson, Yaron Sheffer, and Michael Sweet. All errors and stupidities are of course the editors' fault.