Content Security Policy Level 3

Editor’s Draft,

This version:
https://w3c.github.io/webappsec-csp/
Latest published version:
https://www.w3.org/TR/CSP3/
Previous Versions:
Version History:
https://github.com/w3c/webappsec-csp/commits/master/index.src.html
Feedback:
public-webappsec@w3.org with subject line “[csp3] … message topic …” (archives)
Editor:
(Google Inc.)
Participate:
File an issue (open issues)
Tests:
web-platform-tests content-security-policy/ (ongoing work)

Abstract

This document defines a mechanism by which web developers can control the resources which a particular page can fetch or execute, as well as a number of security-relevant policy decisions.

Status of this document

This is a public copy of the editors’ draft. It is provided for discussion only and may change at any moment. Its publication here does not imply endorsement of its contents by W3C. Don’t cite this document other than as work in progress.

Changes to this document may be tracked at https://github.com/w3c/webappsec.

The (archived) public mailing list public-webappsec@w3.org (see instructions) is preferred for discussion of this specification. When sending e-mail, please put the text “csp3” in the subject, preferably like this: “[csp3] …summary of comment…

This document was produced by the Web Application Security Working Group.

This document was produced by a group operating under the 5 February 2004 W3C Patent Policy. W3C maintains a public list of any patent disclosures made in connection with the deliverables of the group; that page also includes instructions for disclosing a patent. An individual who has actual knowledge of a patent which the individual believes contains Essential Claim(s) must disclose the information in accordance with section 6 of the W3C Patent Policy.

This document is governed by the 1 March 2017 W3C Process Document.

The following features are at-risk, and may be dropped during the CR period:

“At-risk” is a W3C Process term-of-art, and does not necessarily imply that the feature is in danger of being dropped or delayed. It means that the WG believes the feature may have difficulty being interoperably implemented in a timely manner, and marking it as such allows the WG to drop the feature if necessary when transitioning to the Proposed Rec stage, without having to publish a new Candidate Rec without the feature first.

1. Introduction

This section is not normative.

This document defines Content Security Policy (CSP), a tool which developers can use to lock down their applications in various ways, mitigating the risk of content injection vulnerabilities such as cross-site scripting, and reducing the privilege with which their applications execute.

CSP is not intended as a first line of defense against content injection vulnerabilities. Instead, CSP is best used as defense-in-depth. It reduces the harm that a malicious injection can cause, but it is not a replacement for careful input validation and output encoding.

This document is an iteration on Content Security Policy Level 2, with the goal of more clearly explaining the interactions between CSP, HTML, and Fetch on the one hand, and providing clear hooks for modular extensibility on the other. Ideally, this will form a stable core upon which we can build new functionality.

1.1. Examples

1.1.1. Control Execution

MegaCorp Inc’s developers want to protect themselves against cross-site scripting attacks. They can mitigate the risk of script injection by ensuring that their trusted CDN is the only origin from which script can load and execute. Moreover, they wish to ensure that no plugins can execute in their pages' contexts. The following policy has that effect:
Content-Security-Policy: script-src https://cdn.example.com/scripts/; object-src 'none'

1.2. Goals

Content Security Policy aims to do to a few related things:

  1. Mitigate the risk of content-injection attacks by giving developers fairly granular control over

    • The resources which can be requested (and subsequently embedded or executed) on behalf of a specific Document or Worker

    • The execution of inline script

    • Dynamic code execution (via eval() and similar constructs)

    • The application of inline style

  2. Mitigate the risk of attacks which require a resource to be embedded in a malicious context (the "Pixel Perfect" attack described in [TIMING], for example) by giving developers granular control over the origins which can embed a given resource.

  3. Provide a policy framework which allows developers to reduce the privilege of their applications.

  4. Provide a reporting mechanism which allows developers to detect flaws being exploited in the wild.

1.3. Changes from Level 2

This document describes an evolution of the Content Security Policy Level 2 specification [CSP2]. The following is a high-level overview of the changes:

  1. The specification has been rewritten from the ground up in terms of the [FETCH] specification, which should make it simpler to integrate CSP’s requirements and restrictions with other specifications (and with Service Workers in particular).

  2. The child-src model has been substantially altered:

    1. The frame-src directive, which was deprecated in CSP Level 2, has been undeprecated, but continues to defer to child-src if not present (which defers to default-src in turn).

    2. A worker-src directive has been added, deferring to script-src if not present (which likewise defers to default-src in turn).

    3. child-src is now deprecated.

    4. Dedicated workers now always inherit their creator’s policy.

  3. The URL matching algorithm now treats insecure schemes and ports as matching their secure variants. That is, the source expression http://example.com:80 will match both http://example.com:80 and https://example.com:443.

    Likewise, 'self' now matches https: and wss: variants of the page’s origin, even on pages whose scheme is http.

  4. Violation reports generated from inline script or style will now report "inline" as the blocked resource. Likewise, blocked eval() execution will report "eval" as the blocked resource.

  5. The manifest-src directive has been added.

  6. The report-uri directive is deprecated in favor of the new report-to directive, which relies on [REPORTING] as infrastructure.

  7. The 'strict-dynamic' source expression will now allow script which executes on a page to load more script via non-"parser-inserted" script elements. Details are in §8.2 Usage of "'strict-dynamic'".

  8. The 'unsafe-hashed-attributes' source expression will now allow event handlers and style attributes to match hash source expressions. Details in §8.3 Usage of "'unsafe-hashed-attributes'".

    unsafe-hashed-attributes is a work in progress. <https://github.com/w3c/webappsec-csp/issues/13>

  9. The source expression matching has been changed to require explicit presence of any non-network scheme, rather than local scheme, unless that non-network scheme is the same as the scheme of protected resource, as described in §6.6.1.6 Does url match expression in origin with redirect count?.

  10. Hash-based source expressions may now match external scripts if the script element that triggers the request specifies a set of integrity metadata which is listed in the current policy. Details in §8.4 Allowing external JavaScript via hashes.

  11. The disown-opener directive ensures that a resource can’t be opened in such a way as to give another browsing context control over its contents.

  12. The navigation-to directive gives a resource control over the endpoints to which it can initiate navigation.

  13. Reports generated for inline violations will contain a sample attribute if the relevant directive contains the 'report-sample' expression.

2. Framework

2.1. Infrastructure

This document uses ABNF grammar to specify syntax, as defined in [RFC5234]. It also relies on the #rule ABNF extension defined in Section 7 of [RFC7230].

This document depends on the Infra Standard for a number of foundational concepts used in its algorithms and prose [INFRA].

2.2. Policies

A policy defines allowed and restricted behaviors, and may be applied to a Window or WorkerGlobalScope as described in §4.2.2 Initialize a global object’s CSP list.

Each policy has an associated directive set, which is an ordered set of directives that define the policy’s implications when applied.

Each policy has an associated disposition, which is either "enforce" or "report".

Each policy has an associated source, which is either "header" or "meta".

Multiple policies can be applied to a single resource, and are collected into a list of policies known as a CSP list.

A CSP list contains a header-delivered Content Security Policy if it contains a policy whose source is "header".

A serialized CSP is an ASCII string consisting of a semicolon-delimited series of serialized directives, adhering to the following ABNF grammar [RFC5234]:

serialized-policy = serialized-directive *( OWS ";" [ OWS serialized-directive ] )
                    ; OWS is defined in section 3.2.3 of RFC 7230

A serialized CSP list is an ASCII string consisting of a comma-delimited series of serialized CSPs, adhering to the following ABNF grammar [RFC5234]:

serialized-policy-list = 1#serialized-policy
                    ; The '#' rule is defined in section 7 of RFC 7230

2.2.1. Parse a serialized CSP

To parse a serialized CSP, given a serialized CSP (serialized), a source (source), and a disposition (disposition), execute the following steps.

This algorithm returns a Content Security Policy object. If serialized could not be parsed, the object’s directive set will be empty.

  1. Let policy be a new policy with an empty directive set, a source of source, and a disposition of disposition.

  2. For each token returned by strictly splitting serialized on the U+003B SEMICOLON character (;):

    1. Strip leading and trailing ASCII whitespace from token.

    2. If token is an empty string, continue.

    3. Let directive name be the result of collecting a sequence of code points from token which are not ASCII whitespace.

    4. If policy’s directive set contains a directive whose name is directive name, continue.

      In this case, the user agent SHOULD notify developers that a duplicate directive was ignored. A console warning might be appropriate, for example.

    5. Let directive value be the result of splitting token on ASCII whitespace.

    6. Let directive be a new directive whose name is directive name, and value is directive value.

    7. Append directive to policy’s directive set.

  3. Return policy.

2.2.2. Parse a serialized CSP list

To parse a serialized CSP list, given a serialized CSP list (list), a source (source), and a disposition (disposition), execute the following steps.

This algorithm returns a list of Content Security Policy objects. If list cannot be parsed, the returned list will be empty.

  1. Let policies be an empty list.

  2. For each token returned by splitting list on commas:

    1. Let policy be the result of parsing token, with a source of source, and disposition of disposition.

    2. If policy’s directive set is empty, continue.

    3. Append policy to policies.

  3. Return policies.

2.3. Directives

Each policy contains an ordered set of directives (its directive set), each of which controls a specific behavior. The directives defined in this document are described in detail in §6 Content Security Policy Directives.

Each directive is a name / value pair. The name is a non-empty string, and the value is a set of non-empty strings. The value MAY be empty.

A serialized directive is an ASCII string, consisting of one or more whitespace-delimited tokens, and adhering to the following ABNF [RFC5234]:

serialized-directive = directive-name [ RWS directive-value ]
directive-name       = 1*( ALPHA / DIGIT / "-" )
directive-value      = *( %x09 / %x20-%x2B / %x2D-%x3A / %x3C-%7E )
                       ; Directive values may contain whitespace and VCHAR characters,
                       ; excluding ";" and ","

; RWS is defined in section 3.2.3 of RFC7230. ALPHA, DIGIT, and
; VCHAR are defined in Appendix B.1 of RFC 5234.

Directives have a number of associated algorithms:

  1. A pre-request check, which takes a request and a policy as an argument, and is executed during §4.1.3 Should request be blocked by Content Security Policy?. This algorithm returns "Allowed" unless otherwise specified.

  2. A post-request check, which takes a request, a response, and a policy as arguments, and is executed during §4.1.4 Should response to request be blocked by Content Security Policy?. This algorithm returns "Allowed" unless otherwise specified.

  3. A response check, which takes a request, a response, and a policy as arguments, and is executed during §4.1.4 Should response to request be blocked by Content Security Policy?. This algorithm returns "Allowed" unless otherwise specified.

  4. An inline check, which takes an Element a type string, and a soure string as arguments, and is executed during §4.2.3 Should element’s inline type behavior be blocked by Content Security Policy?. This algorithm returns "Allowed" unless otherwise specified.

  5. An initialization, which takes a Document or global object, a response, and a policy as arguments. This algorithm is executed during §4.2.1 Initialize a Document's CSP list, and has no effect unless otherwise specified.

  6. A pre-navigation check, which takes a request, type string, and two browsing contexts as arguments, and is executed during §4.2.4 Should navigation request of type from source in target be blocked by Content Security Policy?. It returns "Allowed" unless otherwise specified.

  7. A navigation response check, which takes a request, a response and two browsing contexts as arguments, and is executed during §4.2.5 Should navigation response to navigation request of type from source in target be blocked by Content Security Policy?. It returns "Allowed" unless otherwise specified.

2.3.1. Source Lists

Many directives' values consist of source lists: sets of strings which identify content that can be fetched and potentially embedded or executed. Each string represents one of the following types of source expression:

  1. Keywords such as 'none' and 'self' (which match nothing and the current URL’s origin, respectively)

  2. Serialized URLs such as https://example.com/path/to/file.js (which matches a specific file) or https://example.com/ (which matches everything on that origin)

  3. Schemes such as https: (which matches any resource having the specified scheme)

  4. Hosts such as example.com (which matches any resource on the host, regardless of scheme) or *.example.com (which matches any resource on the host or any of its subdomains (and any of its subdomains' subdomains, and so on))

  5. Nonces such as 'nonce-ch4hvvbHDpv7xCSvXCs3BrNggHdTzxUA' (which can match specific elements on a page)

  6. Digests such as 'sha256-abcd...' (which can match specific elements on a page)

A serialized source list is an ASCII string, consisting of a whitespace-delimited series of source expressions, adhering to the following ABNF grammar [RFC5234]:

serialized-source-list = ( source-expression *( RWS source-expression ) ) / "'none'"
source-expression      = scheme-source / host-source / keyword-source
                         / nonce-source / hash-source

; Schemes: "https:" / "custom-scheme:" / "another.custom-scheme:"
scheme-source = scheme-part ":"

; Hosts: "example.com" / "*.example.com" / "https://*.example.com:12/path/to/file.js"
host-source = [ scheme-part "://" ] host-part [ ":" port-part ] [ path-part ]
scheme-part = scheme
              ; scheme is defined in section 3.1 of RFC 3986.
host-part   = "*" / [ "*." ] 1*host-char *( "." 1*host-char )
host-char   = ALPHA / DIGIT / "-"
port-part   = 1*DIGIT / "*"
path-part   = path-abempty
              ; path-abempty is defined in section 3.3 of RFC 3986.

; Keywords:
keyword-source = "'self'" / "'unsafe-inline'" / "'unsafe-eval'"
                 / "'strict-dynamic'" / "'unsafe-hashed-attributes'" /
                 / "'report-sample'"

; Nonces: 'nonce-[nonce goes here]'
nonce-source  = "'nonce-" base64-value "'"
base64-value  = 1*( ALPHA / DIGIT / "+" / "/" / "-" / "_" )*2( "=" )

; Digests: 'sha256-[digest goes here]'
hash-source    = "'" hash-algorithm "-" base64-value "'"
hash-algorithm = "sha256" / "sha384" / "sha512"

The host-char production intentionally contains only ASCII characters; internationalized domain names cannot be entered directly as part of a serialized CSP, but instead MUST be Punycode-encoded [RFC3492]. For example, the domain üüüüüü.de MUST be represented as xn--tdaaaaaa.de.

Note: Though IP address do match the grammar above, only 127.0.0.1 will actually match a URL when used in a source expression (see §6.6.1.5 Does url match source list in origin with redirect count? for details). The security properties of IP addresses are suspect, and authors ought to prefer hostnames whenever possible.

Note: The base64-value grammar allows both base64 and base64url encoding. These encodings are treated as equivalant when processing hash-source values. Nonces, however, are strict string matches: we use the base64-value grammar to limit the characters available, and reduce the complexity for the server-side operator (encodings, etc), but the user agent doesn’t actually care about any underlying value, nor does it do any decoding of the nonce-source value.

2.4. Violations

A violation represents an action or resource which goes against the set of policy objects associated with a global object.

Each violation has a global object, which is the global object whose policy has been violated.

Each violation has a url which is its global object’s URL.

Each violation has a status which is a non-negative integer representing the HTTP status code of the resource for which the global object was instantiated.

Each violation has a resource, which is either null, "inline", "eval", or a URL. It represents the resource which violated the policy.

Each violation has a referrer, which is either null, or a URL. It represents the referrer of the resource whose policy was violated.

Each violation has a policy, which is the policy that has been violated.

Each violation has a disposition, which is the disposition of the policy that has been violated.

Each violation has an effective directive which is a non-empty string representing the directive whose enforcement caused the violation.

Each violation has a source file, which is either null or a URL.

Each violation has a line number, which is a non-negative integer.

Each violation has a column number, which is a non-negative integer.

Each violation has a element, which is either null or an element.

Each violation has a sample, which is a string. It is the empty string unless otherwise specified.

Note: A violation’s sample will be populated with the first 40 characters of an inline script, event handler, or style that caused an violation. Violations which stem from an external file will not include a sample in the violation report.

2.4.1. Create a violation object for global, policy, and directive

Given a global object (global), a policy (policy), and a string (directive), the following algorithm creates a new violation object, and populates it with an initial set of data:

  1. Let violation be a new violation whose global object is global, policy is policy, effective directive is directive, and resource is null.

  2. If the user agent is currently executing script, and can extract a source file’s URL, line number, and column number from the global, set violation’s source file, line number, and column number accordingly.

    Is this kind of thing specified anywhere? I didn’t see anything that looked useful in [ECMA262].

    Note: User agents need to ensure that the source file is the URL requested by the page, pre-redirects. If that’s not possible, user agents need to strip the URL down to an origin to avoid unintentional leakage.

  3. If global is a Window object, set violation’s referrer to global’s document's referrer.

  4. Set violation’s status to the HTTP status code for the resource associated with violation’s global object.

    How, exactly, do we get the status code? We don’t actually store it anywhere.

  5. Return violation.

2.4.2. Create a violation object for request, policy, and directive

Given a request (request), a policy (policy), and a string (directive), the following algorithm creates a new violation object, and populates it with an initial set of data:

  1. Let violation be the result of executing §2.4.1 Create a violation object for global, policy, and directive on request’s client’s global object, policy, and directive.

  2. Set violation’s resource to request’s url.

    Note: We use request’s url, and not its current url, as the latter might contain information about redirect targets to which the page MUST NOT be given access.

  3. Return violation.

3. Policy Delivery

A server MAY declare a policy for a particular resource representation via an HTTP response header field whose value is a serialized CSP. This mechanism is defined in detail in §3.1 The Content-Security-Policy HTTP Response Header Field and §3.2 The Content-Security-Policy-Report-Only HTTP Response Header Field, and the integration with Fetch and HTML is described in §4.1 Integration with Fetch and §4.2 Integration with HTML.

A policy may also be declared inline in an HTML document via a meta element’s http-equiv attribute, as described in §3.3 The <meta> element.

3.1. The Content-Security-Policy HTTP Response Header Field

The Content-Security-Policy HTTP response header field is the preferred mechanism for delivering a policy from a server to a client. The header’s value is represented by the following ABNF [RFC5234]:

Content-Security-Policy = 1#serialized-policy
Content-Security-Policy: script-src 'self';
                         report-to csp-reporting-endpoint

A server MAY send different Content-Security-Policy header field values with different representations of the same resource.

A server SHOULD NOT send more than one HTTP response header field named "Content-Security-Policy" with a given resource representation.

When the user agent receives a Content-Security-Policy header field, it MUST parse and enforce each serialized CSP it contains as described in §4.1 Integration with Fetch, §4.2 Integration with HTML.

3.2. The Content-Security-Policy-Report-Only HTTP Response Header Field

The Content-Security-Policy-Report-Only HTTP response header field allows web developers to experiment with policies by monitoring (but not enforcing) their effects. The header’s value is represented by the following ABNF [RFC5234]:

Content-Security-Policy-Report-Only = 1#serialized-policy

This header field allows developers to piece together their security policy in an iterative fashion, deploying a report-only policy based on their best estimate of how their site behaves, watching for violation reports, and then moving to an enforced policy once they’ve gained confidence in that behavior.

Content-Security-Policy-Report-Only: script-src 'self';
                                     report-to csp-reporting-endpoint

A server MAY send different Content-Security-Policy-Report-Only header field values with different representations of the same resource.

A server SHOULD NOT send more than one HTTP response header field named "Content-Security-Policy-Report-Only" with a given resource representation.

When the user agent receives a Content-Security-Policy-Report-Only header field, it MUST parse and monitor each serialized CSP it contains as described in §4.1 Integration with Fetch and §4.2 Integration with HTML.

Note: The Content-Security-Policy-Report-Only header is not supported inside a meta element.

3.3. The <meta> element

A Document may deliver a policy via one or more HTML meta elements whose http-equiv attributes are an ASCII case-insensitive match for the string "Content-Security-Policy". For example:

<meta http-equiv="Content-Security-Policy" content="script-src 'self'">

Implementation details can be found in HTML’s Content Security Policy state http-equiv processing instructions [HTML].

Note: The Content-Security-Policy-Report-Only header is not supported inside a meta element. Neither are the report-uri, frame-ancestors, and sandbox directives.

Authors are strongly encouraged to place meta elements as early in the document as possible, because policies in meta elements are not applied to content which precedes them. In particular, note that resources fetched or prefetched using the Link HTTP response header field, and resources fetched or prefetched using link and script elements which precede a meta-delivered policy will not be blocked.

Note: A policy specified via a meta element will be enforced along with any other policies active for the protected resource, regardless of where they’re specified. The general impact of enforcing multiple policies is described in §8.1 The effect of multiple policies.

Note: Modifications to the content attribute of a meta element after the element has been parsed will be ignored.

4. Integrations

This section is non-normative.

This document defines a set of algorithms which are used in other specifications in order to implement the functionality. These integrations are outlined here for clarity, but those external documents are the normative references which ought to be consulted for detailed information.

4.1. Integration with Fetch

A number of directives control resource loading in one way or another. This specification provides algorithms which allow Fetch to make decisions about whether or not a particular request should be blocked or allowed, and about whether a particular response should be replaced with a network error.

  1. §4.1.3 Should request be blocked by Content Security Policy? is called as part of step #5 of its Main Fetch algorithm. This allows directives' pre-request checks to be executed against each request before it hits the network, and against each redirect that a request might go through on its way to reaching a resource.

  2. §4.1.4 Should response to request be blocked by Content Security Policy? is called as part of step #13 of its Main Fetch algorithm. This allows directives' post-request checks and response checks to be executed on the response delivered from the network or from a Service Worker.

A policy is generally enforced upon a global object, but the user agent needs to parse any policy delivered via an HTTP response header field before any global object is created in order to handle directives that require knowledge of a response’s details. To that end:

  1. A response has an associated CSP list which contains any policy objects delivered in the response’s header list.

  2. §4.1.1 Set response’s CSP list is called in the HTTP fetch and HTTP-network fetch algorithms.

    Note: These two calls should ensure that a response’s CSP list is set, regardless of how the response is created. If we hit the network (via HTTP-network fetch, then we parse the policy before we handle the Set-Cookie header. If we get a response from a Service Worker (via HTTP fetch, we’ll process its CSP list before handing the response back to our caller.

4.1.1. Set response’s CSP list

Given a response (response), this algorithm evaluates its header list for serialized CSP values, and populates its CSP list accordingly:

  1. Set response’s CSP list to the empty list.

  2. Let policies be the result of parsing the result of extracting header list values given Content-Security-Policy and response’s header list, with a source of "header", and a disposition of "enforce".

  3. Append to policies the result of parsing the result of extracting header list values given Content-Security-Policy-Report-Only and response’s header list, with a source of "header", and a disposition of "report".

  4. For each policy in policies:

    1. Insert policy into response’s CSP list.

4.1.2. Report Content Security Policy violations for request

Given a request (request), this algorithm reports violations based on client’s "report only" policies.

  1. Let CSP list be request’s client’s global object’s CSP list.

  2. For each policy in CSP list:

    1. If policy’s disposition is "enforce", then skip to the next policy.

    2. Let violates be the result of executing §6.6.1.1 Does request violate policy? on request and policy.

    3. If violates is not "Does Not Violate", then execute §5.3 Report a violation on the result of executing §2.4.2 Create a violation object for request, policy, and directive on request, policy, and violates.

4.1.3. Should request be blocked by Content Security Policy?

Given a request (request), this algorithm returns Blocked or Allowed and reports violations based on request’s client’s Content Security Policy.

  1. Let CSP list be request’s client’s global object’s CSP list.

  2. Let result be "Allowed".

  3. For each policy in CSP list:

    1. If policy’s disposition is "report", then skip to the next policy.

    2. Let violates be the result of executing §6.6.1.1 Does request violate policy? on request and policy.

    3. If violates is not "Does Not Violate", then:

      1. Execute §5.3 Report a violation on the result of executing §2.4.2 Create a violation object for request, policy, and directive on request, policy, and violates.

      2. Set result to "Blocked".

  4. Return result.

4.1.4. Should response to request be blocked by Content Security Policy?

Given a response (response) and a request (request), this algorithm returns Blocked or Allowed, and reports violations based on request’s client’s Content Security Policy.

  1. Let CSP list be request’s client’s global object’s CSP list.

  2. Let result be "Allowed".

  3. For each policy in CSP list:

    1. For each directive in policy:

      1. If the result of executing directive’s post-request check is "Blocked", then:

        1. Execute §5.3 Report a violation on the result of executing §2.4.2 Create a violation object for request, policy, and directive on request, policy, and directive.

        2. If policy’s disposition is "enforce", then set result to "Blocked".

    Note: This portion of the check verifies that the page can load the response. That is, that a Service Worker hasn’t substituted a file which would violate the page’s CSP.

  4. For each policy in response’s CSP list:

    1. For each directive in policy:

      1. If the result of executing directive’s response check on request, response, and policy is "Blocked", then:

        1. Execute §5.3 Report a violation on the result of executing §2.4.2 Create a violation object for request, policy, and directive on request, policy, and directive.

        2. If policy’s disposition is "enforce", then set result to "Blocked".

    Note: This portion of the check allows policies delivered with the response to determine whether the response is allowed to be delivered.

  5. Return result.

4.2. Integration with HTML

  1. The Document and WorkerGlobalScope objects have a CSP list, which holds all the policy objects which are active for a given context. This list is empty unless otherwise specified, and is populated via the §4.2.2 Initialize a global object’s CSP list algorithm.

    This concept is missing from W3C’s Workers. <https://github.com/w3c/html/issues/187>

  2. A policy is enforced or monitored for a global object by inserting it into the global object’s CSP list.

  3. §4.2.2 Initialize a global object’s CSP list is called during the initializing a new Document object and run a worker algorithms in order to bind a set of policy objects associated with a response to a newly created global object.

  4. §4.2.3 Should element’s inline type behavior be blocked by Content Security Policy? is called during the prepare a script and update a style block algorithms in order to determine whether or not an inline script or style block is allowed to execute/render.

  5. §4.2.3 Should element’s inline type behavior be blocked by Content Security Policy? is called during handling of inline event handlers (like onclick) and inline style attributes in order to determine whether or not they ought to be allowed to execute/render.

  6. policy is enforced during processing of the meta element’s http-equiv.

  7. A Document's embedding document is the Document through which the Document's browsing context is nested.

  8. HTML populates each request’s cryptographic nonce metadata and parser metadata with relevant data from the elements responsible for resource loading.

    Stylesheet loading is not yet integrated with Fetch in W3C’s HTML. <https://github.com/whatwg/html/issues/198>

    Stylesheet loading is not yet integrated with Fetch in WHATWG’s HTML. <https://github.com/whatwg/html/issues/968>

  9. §6.2.1.1 Is base allowed for document? is called during base's set the frozen base URL algorithm to ensure that the href attribute’s value is valid.

  10. §6.2.2.2 Should plugin element be blocked a priori by Content Security Policy?: is called during the processing of object, embed, and applet elements to determine whether they may trigger a fetch.

    Note: Fetched plugin resources are handled in §4.1.4 Should response to request be blocked by Content Security Policy?.

    This hook is missing from W3C’s HTML. <https://github.com/w3c/html/issues/547>

  11. §4.2.4 Should navigation request of type from source in target be blocked by Content Security Policy? is called during the process a navigate fetch algorithm, and §4.2.5 Should navigation response to navigation request of type from source in target be blocked by Content Security Policy? is called during the process a navigate response algorithm to apply directive’s navigation checks, as well as inline checks for navigations to javascript:.

    W3C’s HTML is not based on Fetch, and does not have a process a navigate response algorithm into which to hook. <https://github.com/w3c/html/issues/548>

4.2.1. Initialize a Document's CSP list

Given a Document (document), and a response (response), the user agent performs the following steps in order to initialize document’s CSP list:

  1. If response’s url’s scheme is a local scheme:

    1. Let documents be an empty list.

    2. If document has an embedding document (embedding), then add embedding to documents.

    3. If document has an opener browsing context, then add its active document to documents.

    4. For each doc in documents:

      1. For each policy in doc’s CSP list:

        1. Insert an alias to policy in document’s CSP list.

    Note: local scheme includes about:, and this algorithm will therefore alias the embedding document’s policies for an iframe srcdoc Document.

    Note: We do all this to ensure that a page cannot bypass its policy by embedding a frame or popping up a new window containing content it controls (blob: resources, or document.write()).

  2. For each policy in response’s CSP list, insert policy into document’s CSP list.

  3. For each policy in document’s CSP list:

    1. For each directive in policy:

      1. Execute directive’s initialization algorithm on document and response.

4.2.2. Initialize a global object’s CSP list

Given a global object (global), and a response (response), the user agent performs the following steps in order to initialize global’s CSP list:

  1. If response’s url’s scheme is a local scheme, or if global is a DedicatedWorkerGlobalScope:

    1. Let owners be an empty list.

    2. Add each of the items in global’s owner set to owners.

    3. For each owner in owners:

      1. For each policy in owner’s CSP list:

        1. Insert an alias to policy in global’s CSP list.

    Note: local scheme includes about:, and this algorithm will therefore alias the embedding document’s policies for an iframe srcdoc Document.

  2. If global is a SharedWorkerGlobalScope or ServiceWorkerGlobalScope:

    1. For each policy in response’s CSP list, insert policy into global’s CSP list.

4.2.3. Should element’s inline type behavior be blocked by Content Security Policy?

Given an Element (element), a string (type), and a string (source) this algorithm returns "Allowed" if the element is allowed to have inline definition of a particular type of behavior (script execution, style application, event handlers, etc.), and "Blocked" otherwise:

Note: The valid values for type are "script", "script attribute", "style", and "style attribute".

  1. Assert: element is not null.

  2. Let result be "Allowed".

  3. For each policy in element’s Document's global object’s CSP list:

    1. For each directive in policy’s directive set:

      1. If directive’s inline check returns "Allowed" when executed upon element, type, and source, skip to the next directive.

      2. Otherwise, let violation be the result of executing §2.4.1 Create a violation object for global, policy, and directive on the current settings object’s global object, policy, and "style-src" if type is "style" or "style-attribute", or "script-src" otherwise.

      3. Set violation’s resource to "inline".

      4. Set violation’s element to element.

      5. If directive’s value contains the expression "'report-sample'", then set violation’s sample to the substring of source containing its first 40 characters.

      6. Execute §5.3 Report a violation on violation.

      7. If policy’s disposition is "enforce", then set result to "Blocked".

  4. Return result.

4.2.4. Should navigation request of type from source in target be blocked by Content Security Policy?

Given a request (navigation request), a string (type, either "form-submission" or "other"), and two browsing contexts (source and target), this algorithm return "Blocked" if the active policy blocks the navigation, and "Allowed" otherwise:

  1. Let result be "Allowed".

  2. For each policy in source’s active document’s CSP list:

    1. For each directive in policy:

      1. If directive’s pre-navigation check returns "Allowed" when executed upon navigation request, type, source, and target, skip to the next directive.

      2. Otherwise, let violation be the result of executing §2.4.1 Create a violation object for global, policy, and directive on source’s relevant global object, policy, and directive’s name.

      3. Set violation’s resource to navigation request’s URL.

      4. Execute §5.3 Report a violation on violation.

      5. If policy’s disposition is "enforce", then set result to "Blocked".

  3. If result is "Allowed", and if navigation request’s url’s scheme is javascript:

    1. For each policy in source’s active document’s CSP List:

      1. For each directive in policy:

        1. If directive’s inline check returns "Allowed" when executed upon navigation request, type, source, and target, skip to the next directive.

      2. Otherwise, let violation be the result of executing §2.4.1 Create a violation object for global, policy, and directive on source’s relevant global object, policy, and directive’s name.

      3. Set violation’s resource to navigation request’s URL.

      4. Execute §5.3 Report a violation on violation.

      5. If policy’s disposition is "enforce", then set result to "Blocked".

  4. Return result.

4.2.5. Should navigation response to navigation request of type from source in target be blocked by Content Security Policy?

Given a request (navigation request),, a string (type, either "form-submission" or "other"), a response navigation response, and two browsing contexts (source and target), this algorithm returns "Blocked" if the active policy blocks the navigation, and "Allowed" otherwise:

  1. Let result be "Allowed".

  2. For each policy in navigation response’s CSP list:

    1. For each directive in policy:

      1. If directive’s navigation response check returns "Allowed" when executed upon navigation request, type, navigation response, source, and target, skip to the next directive.

      2. Otherwise, let violation be the result of executing §2.4.1 Create a violation object for global, policy, and directive on null, policy, and directive’s name.

        Note: We use null for the global object, as no global exists: we haven’t processed the navigation to create a Document yet.

      3. Set violation’s resource to navigation response’s URL.

      4. Execute §5.3 Report a violation on violation.

      5. If policy’s disposition is "enforce", then set result to "Blocked".

  3. Return result.

4.3. Integration with ECMAScript

ECMAScript defines a HostEnsureCanCompileStrings() abstract operation which allows the host environment to block the compilation of strings into ECMAScript code. This document defines an implementation of that abstract operation thich examines the relevant CSP list to determine whether such compilation ought to be blocked.

4.3.1. EnsureCSPDoesNotBlockStringCompilation(callerRealm, calleeRealm, source)

Given two realms (callerRealm and calleeRealm), and a string (source), this algorithm returns normally if string compilation is allowed, and throws an "EvalError" if not:

  1. Let globals be a list containing callerRealm’s global object and calleeRealm’s global object.

  2. For each global in globals:

    1. Let result be "Allowed".

    2. For each policy in global’s CSP list:

      1. Let source-list be null.

      2. If policy contains a directive whose name is "script-src", then set source-list to that directive's value.

        Otherwise if policy contains a directive whose name is "default-src", then set source-list to that directive’s value.

      3. If source-list is non-null, and does not contain a source expression which is an ASCII case-insensitive match for the string "'unsafe-eval'", then:

        1. Let violation be the result of executing §2.4.1 Create a violation object for global, policy, and directive on global, policy, and "script-src".

        2. Set violation’s resource to "inline".

        3. If source-list contains the expression "'report-sample'", then set violation’s sample to the substring of source containing its first 40 characters.

        4. Execute §5.3 Report a violation on violation.

        5. If policy’s disposition is "enforce", then set result to "Blocked".

    3. If result is "Blocked", throw an EvalError exception.

HostEnsureCanCompileStrings() does not include the string which is going to be compiled as a parameter. We’ll also need to update HTML to pipe that value through to CSP. <https://github.com/tc39/ecma262/issues/938>

5. Reporting

When one or more of a policy’s directives is violated, a violation report may be generated and sent out to a reporting endpoint associated with the policy.

5.1. Violation DOM Events

enum SecurityPolicyViolationEventDisposition {
  "enforce", "report"
};

[Constructor(DOMString type, optional SecurityPolicyViolationEventInit eventInitDict)]
interface SecurityPolicyViolationEvent : Event {
    readonly    attribute DOMString      documentURI;
    readonly    attribute DOMString      referrer;
    readonly    attribute DOMString      blockedURI;
    readonly    attribute DOMString      violatedDirective;
    readonly    attribute DOMString      effectiveDirective;
    readonly    attribute DOMString      originalPolicy;
    readonly    attribute DOMString      sourceFile;
    readonly    attribute DOMString      sample;
    readonly    attribute SecurityPolicyViolationEventDisposition      disposition;
    readonly    attribute unsigned short statusCode;
    readonly    attribute long           lineNumber;
    readonly    attribute long           columnNumber;
};

dictionary SecurityPolicyViolationEventInit : EventInit {
    DOMString      documentURI;
    DOMString      referrer;
    DOMString      blockedURI;
    DOMString      violatedDirective;
    DOMString      effectiveDirective;
    DOMString      originalPolicy;
    DOMString      sourceFile;
    DOMString      sample;
    SecurityPolicyViolationEventDisposition      disposition;
    unsigned short statusCode;
    long           lineNumber;
    long           columnNumber;
};

5.2. Obtain the deprecated serialization of violation

Given a violation (violation), this algorithm returns a JSON text string representation of the violation, suitable for submission to a reporting endpoint associated with the deprecated report-uri directive.

  1. Let object be a new JavaScript object with properties initialized as follows:

    "document-uri"

    The result of executing the URL serializer on violation’s url, with the exclude fragment flag set.

    "referrer"

    The result of executing the URL serializer on violation’s referrer, with the exclude fragment flag set.

    "blocked-uri"

    The result of executing the URL serializer on violation’s resource, with the exclude fragment flag set.

    "effective-directive"

    violation’s effective directive

    "violated-directive"

    violation’s effective directive

    "original-policy"

    The serialization of violation’s policy

    "disposition"

    The disposition of violation’s policy

    "status-code"

    violation’s status

    "script-sample"

    violation’s sample

    Note: The name script-sample was chosen for compatibility with an earlier iteration of this feature which has shipped in Firefox since its initial implementation of CSP. Despite the name, this field will contain samples for non-script violations, like stylesheets. The data contained in a SecurityPolicyViolationEvent object, and in reports generated via the new report-to directive, is named in a more encompassing fashion: sample.

  2. If violation’s source file is not null:

    1. Set object’s "source-file" property to the result of executing the URL serializer on violation’s source file, with the exclude fragment flag set.

    2. Set object’s "line-number" property to violation’s line number.

    3. Set object’s "column-number" property to violation’s column number.

  3. Assert: If object’s "blocked-uri" property is not "inline", then its "sample" property is the empty string.

  4. Return the result of executing JSON.stringify() on object.

5.3. Report a violation

Given a violation (violation), this algorithm reports it to the endpoint specified in violation’s policy, and fires a SecurityPolicyViolationEvent at violation’s element, or at violation’s global object as described below:

  1. Let global be violation’s global object.

  2. Let target be violation’s element.

  3. Queue a task to run the following steps:

    Note: We "queue a task" here to ensure that the event targeting and dispatch happens after JavaScript completes execution of the task responsible for a given violation (which might manipulate the DOM).

    1. If target is not null, and global is a Window, and target’s shadow-including root is not global’s associated Document, set target to null.

      Note: This ensures that we fire events only at elements connected to violation’s policy’s Document. If a violation is caused by an element which isn’t connected to that document, we’ll fire the event at the document rather than the element in order to ensure that the violation is visible to the document’s listeners.

    2. If target is null:

      1. Set target be violation’s global object.

      2. If target is a Window, set target to target’s associated Document.

    3. Fire an event named securitypolicyviolation that uses the SecurityPolicyViolationEvent interface at target with its attributes initialized as follows:

      documentURI

      The result of executing the URL serializer on violation’s url, with the exclude fragment flag set.

      referrer

      The result of executing the URL serializer on violation’s referrer, with the exclude fragment flag set.

      blockedURI

      The result of executing the URL serializer on violation’s resource, with the exclude fragment flag set.

      effectiveDirective

      violation’s effective directive

      violatedDirective

      violation’s effective directive

      originalPolicy

      violation’s policy

      disposition

      violation’s disposition

      sourceFile

      violation’s source file

      statusCode

      violation’s status

      lineNumber

      violation’s line number

      columnNumber

      violation’s column number

      sample

      violation’s sample

      bubbles

      true

      composed

      true

      Note: Both effectiveDirective and violatedDirective are the same value. This is intentional to maintain backwards compatibility.

      Note: We set the composed attribute, which means that this event can be captured on its way into, and will bubble its way out of a shadow tree. target, et al will be automagically scoped correctly for the main tree.

    4. If violation’s policy’s directive set contains a directive named "report-uri" (directive):

      1. If violation’s policy’s directive set contains a directive named "report-to", skip the remaining substeps.

      2. Let endpoint be the result of executing the URL parser with directive’s value as the input, and violation’s url as the base URL.

      3. If endpoint is not a valid URL, skip the remaining substeps.

      4. Let request be a new request, initialized as follows:

        method

        "POST"

        url

        violation’s url

        origin

        violation’s global object’s relevant settings object’s origin

        window

        "no-window"

        client

        violation’s global object’s relevant settings object

        destination

        "report"

        initiator

        ""

        type

        ""

        credentials mode

        "same-origin"

        keepalive flag

        "true"

        header list

        A header list containing a single header whose name is "Content-Type", and value is "application/csp-report"

        body

        The result of executing §5.2 Obtain the deprecated serialization of violation on violation

        redirect mode

        "error"

        Note: request’s mode defaults to "no-cors"; the response is ignored entirely.

      5. Fetch request. The result will be ignored.

      Note: All of this should be considered deprecated. It sends a single request per violation, which simply isn’t scalable. As soon as this behavior can be removed from user agents, it will be.

      Note: report-uri only takes effect if report-to is not present. That is, the latter overrides the former, allowing for backwards compatibility with browsers that don’t support the new mechanism.

    5. If violation’s policy’s directive set contains a directive named "report-to" (directive):

      1. Let group be directive’s value.

      2. Let settings object be violation’s global object’s relevant settings object.

      3. Execute [REPORTING]'s Queue data as type for endpoint group on settings algorithm with the following arguments:

        data

        violation

        type

        "CSP"

        endpoint group

        group

        settings

        settings object

6. Content Security Policy Directives

This specification defines a number of types of directives which allow developers to control certain aspects of their sites' behavior. This document defines directives which govern resource fetching (in §6.1 Fetch Directives), directives which govern the state of a document (in §6.2 Document Directives), directives which govern aspects of navigation (in §6.3 Navigation Directives), and directives which govern reporting (in §6.4 Reporting Directives). These form the core of Content Security Policy; other directives are defined in a modular fashion in ancillary documents (see §6.5 Directives Defined in Other Documents for examples).

To mitigate the risk of cross-site scripting attacks, web developers SHOULD include directives that regulate sources of script and plugins. They can do so by including:

In either case, developers SHOULD NOT include either 'unsafe-inline', or data: as valid sources in their policies. Both enable XSS attacks by allowing code to be included directly in the document itself; they are best avoided completely.

6.1. Fetch Directives

Fetch directives control the locations from which certain resource types may be loaded. For instance, script-src allows developers to allow trusted sources of script to execute on a page, while font-src controls the sources of web fonts.

6.1.1. child-src

The child-src directive is deprecated. Authors who wish to regulate nested browsing contexts and workers SHOULD use the frame-src and worker-src directives, respectively.

The child-src directive governs the creation of nested browsing contexts (e.g. iframe and frame navigations) and Worker execution contexts. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "child-src"
directive-value = serialized-source-list

This directive controls requests which will populate a frame or a worker. More formally, requests falling into one of the following categories:

Given a page with the following Content Security Policy:
Content-Security-Policy: child-src https://example.com/

Fetches for the following code will all return network errors, as the URLs provided do not match child-src's source list:

<iframe src="https://not-example.com"></iframe>
<script>
  var blockedWorker = new Worker("data:application/javascript,...");
</script>
6.1.1.1. child-src Pre-request check

This directive’s pre-request check is as follows:

Given a request (request) and a policy (policy):

  1. Let name be the result of executing §6.6.1.11 Get the effective directive for request on request.

  2. If name is not frame-src, return "Allowed".

  3. If policy contains a directive whose name is name, return "Allowed"

  4. Return the result of executing the pre-request check for the directive whose name is name on request and policy, using this directive’s value for the comparison.

6.1.1.2. child-src Post-request check

This directive’s post-request check is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Let name be the result of executing §6.6.1.11 Get the effective directive for request on request.

  2. If name is not frame-src, return "Allowed".

  3. If policy contains a directive whose name is name, return "Allowed"

  4. Return the result of executing the post-request check for the directive whose name is name on request, response, and policy, using this directive’s value for the comparison.

6.1.2. connect-src

The connect-src directive restricts the URLs which can be loaded using script interfaces. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "connect-src"
directive-value = serialized-source-list

This directive controls requests which transmit or receive data from other origins. This includes APIs like fetch(), [XHR], [EVENTSOURCE], [BEACON], and a's ping. This directive also controls WebSocket [WEBSOCKETS] connections, though those aren’t technically part of Fetch.

JavaScript offers a few mechanisms that directly connect to an external server to send or receive information. EventSource maintains an open HTTP connection to a server in order to receive push notifications, WebSockets open a bidirectional communication channel between your browser and a server, and XMLHttpRequest makes arbitrary HTTP requests on your behalf. These are powerful APIs that enable useful functionality, but also provide tempting avenues for data exfiltration.

The connect-src directive allows you to ensure that these and similar sorts of connections are only opened to origins you trust. Sending a policy that defines a list of source expressions for this directive is straightforward. For example, to limit connections to only https://example.com, send the following header:

Content-Security-Policy: connect-src https://example.com/

Fetches for the following code will all return network errors, as the URLs provided do not match connect-src's source list:

<a ping="https://not-example.com">...
<script>
  var xhr = new XMLHttpRequest();
  xhr.open('GET', 'https://not-example.com/');
  xhr.send();

  var ws = new WebSocket("https://not-example.com/");

  var es = new EventSource("https://not-example.com/");

  navigator.sendBeacon("https://not-example.com/", { ... });
</script>
6.1.2.1. connect-src Pre-request check

This directive’s pre-request check is as follows:

Given a request (request) and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s initiator is "fetch", or its type is "" and destination is "":

    1. If the result of executing §6.6.1.3 Does request match source list? on request and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.2.2. connect-src Post-request check

This directive’s post-request check is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s initiator is "fetch", or its type is "" and destination is "subresource":

    1. If the result of executing §6.6.1.4 Does response to request match source list? on response, request, and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.3. default-src

The default-src directive serves as a fallback for the other fetch directives. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "default-src"
directive-value = serialized-source-list

If a default-src directive is present in a policy, its value will be used as the policy’s default source list. That is, given default-src 'none'; script-src 'self', script requests will use 'self' as the source list to match against. Other requests will use 'none'. This is spelled out in more detail in the §4.1.3 Should request be blocked by Content Security Policy? and §4.1.4 Should response to request be blocked by Content Security Policy? algorithms.

The following header:
Content-Security-Policy: default-src 'self'

will have the same behavior as the following header:

Content-Security-Policy: connect-src 'self';
                         font-src 'self';
                         frame-src 'self';
                         img-src 'self';
                         manifest-src 'self';
                         media-src 'self';
                         object-src 'self';
                         script-src 'self';
                         style-src 'self';
                         worker-src 'self'

That is, when default-src is set, every fetch directive that isn’t explicitly set will fall back to the value default-src specifies.

There is no inheritance. If a script-src directive is explicitly specified, for example, then the value of default-src has no influence on script requests. That is, the following header:
Content-Security-Policy: default-src 'self'; script-src https://example.com

will have the same behavior as the following header:

Content-Security-Policy: connect-src 'self';
                         font-src 'self';
                         frame-src 'self';
                         img-src 'self';
                         manifest-src 'self';
                         media-src 'self';
                         object-src 'self';
                         script-src https://example.com;
                         style-src 'self';
                         worker-src 'self'

Given this behavior, one good way to build a policy for a site would be to begin with a default-src of 'none', and to build up a policy from there which allowed only those resource types which are necessary for the particular page the policy will apply to.

6.1.3.1. default-src Pre-request check

This directive’s pre-request check is as follows:

Given a request (request) and a policy (policy):

  1. Let name be the result of executing §6.6.1.11 Get the effective directive for request on request.

  2. If name is null, return "Allowed".

  3. If policy contains a directive whose name is name, return "Allowed".

  4. If name is "frame-src", and policy contains a directive whose name is "child-src", return "Allowed".

  5. If name is "worker-src", and policy contains a directive whose name is "script-src", return "Allowed".

  6. Otherwise, return the result of executing the pre-request check for the directive whose name is name on request and policy, using this directive’s value for the comparison.

6.1.3.2. default-src Post-request check

This directive’s post-request check is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Let name be the result of executing §6.6.1.11 Get the effective directive for request on request.

  2. If name is null, return "Allowed".

  3. If policy contains a directive whose name is name, return "Allowed".

  4. If name is "frame-src", and policy contains a directive whose name is "child-src", return "Allowed".

  5. If name is "worker-src", and policy contains a directive whose name is "script-src", return "Allowed".

  6. Otherwise, return the result of executing the post-request check for the directive whose name is name on request, response, and policy, using this directive’s value for the comparison.

6.1.4. font-src

The font-src directive restricts the URLs from which font resources may be loaded. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "font-src"
directive-value = serialized-source-list
Given a page with the following Content Security Policy:
Content-Security-Policy: font-src https://example.com/

Fetches for the following code will return a network errors, as the URL provided do not match font-src's source list:

<style>
  @font-face {
    font-family: "Example Font";
    src: url("https://not-example.com/font");
  }
  body {
    font-family: "Example Font";
  }
</style>
6.1.4.1. font-src Pre-request check

This directive’s pre-request check is as follows:

Given a request (request) and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is "font":

    1. If the result of executing §6.6.1.3 Does request match source list? on request and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.4.2. font-src Post-request check

This directive’s post-request check is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is "font":

    1. If the result of executing §6.6.1.4 Does response to request match source list? on response, request, and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.5. frame-src

The frame-src directive restricts the URLs which may be loaded into nested browsing contexts. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "frame-src"
directive-value = serialized-source-list
Given a page with the following Content Security Policy:
Content-Security-Policy: frame-src https://example.com/

Fetches for the following code will return a network errors, as the URL provided do not match frame-src's source list:

<iframe src="https://not-example.com/">
</iframe>
6.1.5.1. frame-src Pre-request check

This directive’s pre-request check is as follows:

Given a request (request) and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is "document" and target browsing context is a nested browsing context:

    1. If the result of executing §6.6.1.3 Does request match source list? on request and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.5.2. frame-src Post-request check

This directive’s post-request check is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is "document" and target browsing context is a nested browsing context:

    1. If the result of executing §6.6.1.4 Does response to request match source list? on response, request, and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.6. img-src

The img-src directive restricts the URLs from which image resources may be loaded. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "img-src"
directive-value = serialized-source-list

This directive controls requests which load images. More formally, this includes requests whose type is "image" [FETCH].

Given a page with the following Content Security Policy:
Content-Security-Policy: img-src https://example.com/

Fetches for the following code will return a network errors, as the URL provided do not match img-src's source list:

<img src="https://not-example.com/img">
6.1.6.1. img-src Pre-request check

This directive’s pre-request check is as follows:

Given a request (request) and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is "image":

    1. If the result of executing §6.6.1.3 Does request match source list? on request and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.6.2. img-src Post-request check

This directive’s post-request check is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is "image":

    1. If the result of executing §6.6.1.4 Does response to request match source list? on response, request, and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.7. manifest-src

The manifest-src directive restricts the URLs from which application manifests may be loaded [APPMANIFEST]. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "manifest-src"
directive-value = serialized-source-list
Given a page with the following Content Security Policy:
Content-Security-Policy: manifest-src https://example.com/

Fetches for the following code will return a network errors, as the URL provided do not match manifest-src's source list:

<link rel="manifest" href="https://not-example.com/manifest">
6.1.7.1. manifest-src Pre-request check

This directive’s pre-request check is as follows:

Given a request (request) and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is "", and its initiator is "manifest":

    1. If the result of executing §6.6.1.3 Does request match source list? on request and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.7.2. manifest-src Post-request check

This directive’s post-request check is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is "", and its initiator is "manifest":

    1. If the result of executing §6.6.1.4 Does response to request match source list? on response, request, and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.8. media-src

The media-src directive restricts the URLs from which video, audio, and associated text track resources may be loaded. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "media-src"
directive-value = serialized-source-list
Given a page with the following Content Security Policy:
Content-Security-Policy: media-src https://example.com/

Fetches for the following code will return a network errors, as the URL provided do not match media-src's source list:

<audio src="https://not-example.com/audio"></audio>
<video src="https://not-example.com/video">
    <track kind="subtitles" src="https://not-example.com/subtitles">
</video>
6.1.8.1. media-src Pre-request check

This directive’s pre-request check is as follows:

Given a request (request) and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is one of "audio", "video", or "track":

    1. If the result of executing §6.6.1.3 Does request match source list? on request and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.8.2. media-src Post-request check

This directive’s post-request check is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is one of "audio", "video", or "track":

    1. If the result of executing §6.6.1.4 Does response to request match source list? on response, request, and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.9. object-src

The object-src directive restricts the URLs from which plugin content may be loaded. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "object-src"
directive-value = serialized-source-list
Given a page with the following Content Security Policy:
Content-Security-Policy: object-src https://example.com/

Fetches for the following code will return a network errors, as the URL provided do not match object-src's source list:

<embed src="https://not-example.com/flash"></embed>
<object data="https://not-example.com/flash"></object>
<applet archive="https://not-example.com/flash"></applet>

If plugin content is loaded without an associated URL (perhaps an object element lacks a data attribute, but loads some default plugin based on the specified type), it MUST be blocked if object-src's value is 'none', but will otherwise be allowed.

Note: The object-src directive acts upon any request made on behalf of an object, embed, or applet element. This includes requests which would populate the nested browsing context generated by the former two (also including navigations). This is true even when the data is semantically equivalent to content which would otherwise be restricted by another directive, such as an object element with a text/html MIME type.

Note: When a plugin resource is navigated to directly (that is, as a plugin document in the top-level browsing context or a nested browsing context, and not as an embedded subresource via embed, object, or applet), any policy delivered along with that resource will be applied to the plugin document. This means, for instance, that developers can prevent the execution of arbitrary resources as plugin content by delivering the policy object-src 'none' along with a response. Given plugins' power (and the sometimes-interesting security model presented by Flash and others), this could mitigate the risk of attack vectors like Rosetta Flash.

6.1.9.1. object-src Pre-request check

This directive’s pre-request check is as follows:

Given a request (request) and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is "", and its destination is "unknown":

    1. If the result of executing §6.6.1.3 Does request match source list? on request and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.9.2. object-src Post-request check

This directive’s post-request check is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is "", and its destination is "unknown":

    1. If the result of executing §6.6.1.4 Does response to request match source list? on response, request, and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.10. script-src

The script-src directive restricts the locations from which scripts may be executed. This includes not only URLs loaded directly into script elements, but also things like inline script blocks and XSLT stylesheets [XSLT] which can trigger script execution. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "script-src"
directive-value = serialized-source-list

The script-src directive governs four things:

  1. Script requests MUST pass through §4.1.3 Should request be blocked by Content Security Policy?.

  2. Script responses MUST pass through §4.1.4 Should response to request be blocked by Content Security Policy?.

  3. Inline script blocks MUST pass through §4.2.3 Should element’s inline type behavior be blocked by Content Security Policy?. Their behavior will be blocked unless every policy allows inline script, either implicitly by not specifying a script-src (or default-src) directive, or explicitly, by specifying "unsafe-inline", a nonce-source or a hash-source that matches the inline block.

  4. The following JavaScript execution sinks are gated on the "unsafe-eval" source expression:

    Note: If a user agent implements non-standard sinks like setImmediate() or execScript(), they SHOULD also be gated on "unsafe-eval".

6.1.10.1. script-src Pre-request check

This directive’s pre-request check is as follows:

Given a request (request) and a policy (policy):

  1. If the result of executing §6.6.1.11 Get the effective directive for request on request is "worker-src", and policy contains a directive whose name is "worker-src", return "Allowed".

    Note: If worker-src is present, we’ll defer to it when handling worker requests.

  2. If request’s type is "script":

    1. If the result of executing §6.6.1.2 Does nonce match source list? on request’s cryptographic nonce metadata and this directive’s value is "Matches", return "Allowed".

    2. Let integrity expressions be the set of source expressions in this directive’s value that match the hash-source grammar.

    3. If integrity expressions is not empty:

      1. Let integrity sources be the result of executing the algorithn defined in Subresource Integrity §parse-metadata on request’s integrity metadata. [SRI]

      2. If integrity sources is "no metadata" or an empty set, skip the remaining substeps.

      3. Let bypass due to integrity match be true.

      4. For each source in integrity sources:

        1. If this directive’s value does not contain a source expression whose hash-algorithm is a case-sensitive match for source’s hash-algo component, and whose base64-value is a case-sensitive match for source’s base64-value, then set bypass due to integrity match to false.

      5. If bypass due to integrity match is true, return "Allowed".

      Note: Here, we verify only that the request contains a set of integrity metadata which is a subset of the hash-source source expressions specified by this directive. We rely on the browser’s enforcement of Subresource Integrity [SRI] to block non-matching resources upon response.

    4. If this directive’s value contains a source expression that is an ASCII case-insensitive match for the "'strict-dynamic'" keyword-source:

      1. If the request’s parser metadata is "parser-inserted", return "Blocked".

        Otherwise, return "Allowed".

        Note: "'strict-dynamic'" is explained in more detail in §8.2 Usage of "'strict-dynamic'".

    5. If the result of executing §6.6.1.3 Does request match source list? on request and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.10.2. script-src Post-request check

This directive’s post-request check is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. If the result of executing §6.6.1.11 Get the effective directive for request on request is "worker-src", and policy contains a directive whose name is "worker-src", return "Allowed".

    Note: If worker-src is present, we’ll defer to it when handling worker requests.

  2. If request’s type is "script":

    1. If the result of executing §6.6.1.2 Does nonce match source list? on request’s cryptographic nonce metadata and this directive’s value is "Matches", return "Allowed".

    2. If this directive’s value contains "'strict-dynamic'", and request’s parser metadata is not "parser-inserted", return "Allowed".

    3. If the result of executing §6.6.1.4 Does response to request match source list? on response, request, and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.10.3. script-src Inline Check

This directive’s inline check algorithm is as follows:

Given an Element (element), a string (type), and a string (source):

  1. If type is "script attribute" or "script":

    1. Assert: element is not null.

    2. If the result of executing §6.6.2.3 Does element match source list for type and source? on element, this directive’s value, type, and source, is "Does Not Match", return "Blocked".

  2. If type is "navigation":

    1. Let unsafe-inline flag be false.

    2. For each expression in this directive’s value:

      1. If expression matches the nonce-source or hash-source grammar, return "Blocked".

      2. If expression matches the keyword-source "'strict-dynamic'", return "Blocked".

      3. If expression matches the keyword-source "'unsafe-inline'", set unsafe-inline flag to true.

    3. If unsafe-inline flag is false, return "Blocked".

    Note: Navigating to a javascript: URL is allowed only in the presence of "'unsafe-inline'" that isn’t overridden by a nonce, hash, or "'strict-dynamic'".

  3. Return "Allowed".

6.1.11. style-src

The style-src directive restricts the locations from which style may be applied to a Document. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "style-src"
directive-value = serialized-source-list

The style-src directive governs several things:

  1. Style requests MUST pass through §4.1.3 Should request be blocked by Content Security Policy?. This includes:

    1. Stylesheet requests originating from a link element.

    2. Stylesheet requests originating from the @import rule.

    3. Stylesheet requests originating from a Link HTTP response header field [RFC5988].

  2. Responses to style requests MUST pass through §4.1.4 Should response to request be blocked by Content Security Policy?.

  3. Inline style blocks MUST pass through §4.2.3 Should element’s inline type behavior be blocked by Content Security Policy?. The styles will be blocked unless every policy allows inline style, either implicitly by not specifying a style-src (or default-src) directive, or explicitly, by specifying "unsafe-inline", a nonce-source or a hash-source that matches the inline block.

  4. The following CSS algorithms are gated on the unsafe-eval source expression:

    1. insert a CSS rule

    2. parse a CSS rule,

    3. parse a CSS declaration block

    4. parse a group of selectors

    This would include, for example, all invocations of CSSOM’s various cssText setters and insertRule methods [CSSOM] [HTML].

    This needs to be better explained.

6.1.11.1. style-src Pre-request Check

This directive’s pre-request check is as follows:

Given a request (request) and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is "style":

    1. If the result of executing §6.6.1.2 Does nonce match source list? on request’s cryptographic nonce metadata and this directive’s value is "Matches", return "Allowed".

    2. If the result of executing §6.6.1.3 Does request match source list? on request and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.11.2. style-src Post-request Check

This directive’s post-request check is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s type is "style":

    1. If the result of executing §6.6.1.2 Does nonce match source list? on request’s cryptographic nonce metadata and this directive’s value is "Matches", return "Allowed".

    2. If the result of executing §6.6.1.4 Does response to request match source list? on response, request, and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.11.3. style-src Inline Check

This directive’s inline check algorithm is as follows:

Given an Element (element), a string (type), and a string (source):

  1. If type is "style" or "style attribute":

    1. If the result of executing §6.6.2.3 Does element match source list for type and source? on element, this directive’s value, type, and source, is "Does Not Match", return "Blocked".

  2. Return "Allowed".

This directive’s initialization algorithm is as follows:

Do something interesting to the execution context in order to lock down interesting CSSOM algorithms. I don’t think CSSOM gives us any hooks here, so let’s work with them to put something reasonable together.

6.1.12. worker-src

The worker-src directive restricts the URLs which may be loaded as a Worker, SharedWorker, or ServiceWorker. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "worker-src"
directive-value = serialized-source-list
Given a page with the following Content Security Policy:
Content-Security-Policy: worker-src https://example.com/

Fetches for the following code will return a network errors, as the URL provided do not match worker-src's source list:

<script>
  var blockedWorker = new Worker("data:application/javascript,...");
  blockedWorker = new SharedWorker("https://not-example.com/");
  navigator.serviceWorker.register('https://not-example.com/sw.js');
</script>
6.1.12.1. worker-src Pre-request Check

This directive’s pre-request check is as follows:

Given a request (request) and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s destination is one of "serviceworker", "sharedworker", or "worker":

    1. If the result of executing §6.6.1.3 Does request match source list? on request and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.1.12.2. worker-src Post-request Check

This directive’s post-request check is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s destination is one of "serviceworker", "sharedworker", or "worker":

    1. If the result of executing §6.6.1.4 Does response to request match source list? on response, request, and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.2. Document Directives

The following directives govern the properties of a document or worker environment to which a policy applies.

6.2.1. base-uri

The base-uri directive restricts the URLs which can be used in a Document's base element. The syntax for the directive’s name and value is described by the following ABNF:

directive-name  = "base-uri"
directive-value = serialized-source-list

The following algorithm is called during HTML’s set the frozen base url algorithm in order to monitor and enforce this directive:

6.2.1.1. Is base allowed for document?

Given a URL (base), and a Document (document), this algorithm returns "Allowed" if base may be used as the value of a base element’s href attribute, and "Blocked" otherwise:

  1. For each policy in document’s global object’s csp list:

    1. Let source list be null.

    2. If a directive whose name is "base-uri" is present in policy’s directive set, set source list to that directive’s value.

    3. If source list is null, skip to the next policy.

    4. If the result of executing §6.6.1.5 Does url match source list in origin with redirect count? on base, source list, document’s fallback base URL’s origin, and 0 is "Does Not Match":

      1. Let violation be the result of executing §2.4.1 Create a violation object for global, policy, and directive on document’s global object, policy, and "base-uri".

      2. Set violation’s resource to "inline".

      3. Execute §5.3 Report a violation on violation.

      4. If policy’s disposition is "enforce", return "Blocked".

    Note: We compare against the fallback base URL in order to deal correctly with things like an iframe srcdoc Document which has been sandboxed into an opaque origin.

  2. Return "Allowed".

6.2.2. plugin-types

The plugin-types directive restricts the set of plugins that can be embedded into a document by limiting the types of resources which can be loaded. The directive’s syntax is described by the following ABNF grammar:

directive-name  = "plugin-types"
directive-value = media-type-list

media-type-list = media-type *( RWS media-type )
media-type = type "/" subtype
; type and subtype are defined in RFC 2045

If a plugin-types directive is present, instantiation of an embed or object element will fail if any of the following conditions hold:

  1. The element does not explicitly declare a valid MIME type via a type attribute.

  2. The declared type does not match one of the items in the directive’s value.

  3. The fetched resource does not match the declared type.

Given a page with the following Content Security Policy:
Content-Security-Policy: plugin-types application/pdf

Fetches for the following code will all return network errors:

<!-- No 'type' declaration -->
<object data="https://example.com/flash"></object>

<!-- Non-matching 'type' declaration -->
<object data="https://example.com/flash" type="application/x-shockwave-flash"></object>

<!-- Non-matching resource -->
<object data="https://example.com/flash" type="application/pdf"></object>

If the page allowed Flash content by sending the following header:

Content-Security-Policy: plugin-types application/x-shockwave-flash

Then the second item above would load successfully:

<!-- Matching 'type' declaration and resource -->
<object data="https://example.com/flash" type="application/x-shockwave-flash"></object>
6.2.2.1. plugin-types Post-Request Check

This directive’s post-request check algorithm is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Assert: policy is unused.

  2. If request’s destination is either "object" or "embed":

    1. Let type be the result of extracting a MIME type from response’s header list.

    2. If type is not an ASCII case-insensitive match for any item in this directive’s value, return "Blocked".

  3. Return "Allowed".

6.2.2.2. Should plugin element be blocked a priori by Content Security Policy?:

Given an Element (plugin element), this algorithm returns "Blocked" or "Allowed" based on the element’s type attribute and the policy applied to its document:

  1. For each policy in plugin element’s node document’s CSP list:

    1. If policy contains a directive (directive) whose name is plugin-types:

      1. Let type be "application/x-java-applet" if plugin element is an applet element, or plugin element’s type attribute’s value if present, or "null" otherwise.

      2. Return "Blocked" if any of the following are true:

        1. type is null.

        2. type is not a valid MIME type.

        3. type is not an ASCII case-insensitive match for any item in directive’s value.

  2. Return "Allowed".

6.2.3. sandbox

The sandbox directive specifies an HTML sandbox policy which the user agent will apply to a resource, just as though it had been included in an iframe with a sandbox property.

The directive’s syntax is described by the following ABNF grammar, with the additional requirement that each token value MUST be one of the keywords defined by HTML specification as allowed values for the iframe sandbox attribute [HTML].

directive-name  = "sandbox"
directive-value = "" / token *( RWS token )

This directive has no reporting requirements; it will be ignored entirely when delivered in a Content-Security-Policy-Report-Only header, or within a meta element.

6.2.3.1. sandbox Response Check

This directive’s response check algorithm is as follows:

Given a request (request), a response (response), and a policy (policy):

  1. Assert: response is unused.

  2. If policy’s disposition is not "enforce", then return "Allowed".

  3. If request’s destination is one of "serviceworker", "sharedworker", or "worker":

    1. If the result of the Parse a sandboxing directive algorithm using this directive’s value as the input contains either the sandboxed scripts browsing context flag or the sandboxed origin browsing context flag flags, return "Blocked".

      Note: This will need to change if we allow Workers to be sandboxed into unique origins, which seems like a pretty reasonable thing to do.

  4. Return "Allowed".

6.2.3.2. sandbox Initialization

This directive’s initialization algorithm is responsible for adjusting a Document's forced sandboxing flag set according to the sandbox values present in its policies, as follows:

Given a Document or global object (context), a response (response), and a policy (policy):

  1. Assert: response is unused.

  2. If policy’s disposition is not "enforce", or context is not a Document, then abort this algorithm.

    Note: This will need to change if we allow Workers to be sandboxed, which seems like a pretty reasonable thing to do.

  3. Parse a sandboxing directive using this directive’s value as the input, and context’s forced sandboxing flag set as the output.

6.2.4. disown-opener

The disown-opener directive ensures that a resource will disown its opener when navigated to. The directive’s syntax is described by the following ABNF grammar:

directive-name  = "disown-opener"
directive-value = ""

This directive has no reporting requirements; it will be ignored entirely when delivered in a Content-Security-Policy-Report-Only header, or within a meta element.

Not sure this is the right model. We need to ensure that we take care of the inverse as well, and there might be a cleverer syntax that could encompass both a document’s opener, and a document’s openees. disown-openee is weird. Maybe disown 'opener' 'openee'? Do we need origin restrictions on either/both?

6.2.4.1. disown-opener Initialization

This directive’s initialization algorithm is as follows:

Given a Document or global object (context), a response (response), and a policy (policy):

  1. Assert: response and policy are unused.

  2. If context’s responsible browsing context has an opener browsing context, disown its opener.

What should this do in an iframe? Anything?

6.3. Navigation Directives

6.3.1. form-action

The form-action directive restricts the URLs which can be used as the target of a form submissions from a given context. The directive’s syntax is described by the following ABNF grammar:

directive-name  = "form-action"
directive-value = serialized-source-list
6.3.1.1. form-action Pre-Navigation Check

Given a request (request), a string (type, "form-submission or "other") and two browsing contexts (source and target), this algorithm returns "Blocked" if one or more of the ancestors of target violate the frame-ancestors directive delivered with the response, and "Allowed" otherwise. This constitutes the form-action' directive’s pre-navigation check:

  1. Assert: source and target are unused in this algorithm, as form-action is concerned only with details of the outgoing request.

  2. If type is "form-submission":

    1. If the result of executing §6.6.1.3 Does request match source list? on request and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.3.2. frame-ancestors

The frame-ancestors directive restricts the URLs which can embed the resource using frame, iframe, object, embed, or applet element. Resources can use this directive to avoid many UI Redressing [UISECURITY] attacks, by avoiding the risk of being embedded into potentially hostile contexts.

The directive’s syntax is described by the following ABNF grammar:

directive-name  = "frame-ancestors"
directive-value = ancestor-source-list

ancestor-source-list = ( ancestor-source *( RWS ancestor-source) ) / "'none'"
ancestor-source      = scheme-source / host-source / "'self'"

The frame-ancestors directive MUST be ignored when contained in a policy declared via a meta element.

Note: The frame-ancestors directive’s syntax is similar to a source list, but frame-ancestors will not fall back to the default-src directive’s value if one is specified. That is, a policy that declares default-src 'none' will still allow the resource to be embedded by anyone.

6.3.2.1. frame-ancestors Navigation Response Check

Given a request (request), a response (navigation response) and two browsing contexts (source and target), this algorithm returns "Blocked" if one or more of the ancestors of target violate the frame-ancestors directive delivered with the response, and "Allowed" otherwise. This constitutes the frame-ancestors' directive’s navigation response check:

  1. Assert: request, navigation response, and source are unused in this algorithm, as frame-ancestors is concerned only with target’s ancestors.

  2. If target is not a nested browsing context, return "Allowed".

  3. Let current be target.

  4. While current has a parent browsing context (parent):

    1. Set current to parent.

    2. Let origin be the result of executing the URL parser on the ASCII serialization of parent’s active document’s relevant settings object’s origin.

    3. If §6.6.1.5 Does url match source list in origin with redirect count? returns Does Not Match when executed upon origin, this directive’s value, navigation response’s url’s origin, and 0, return "Blocked".

  5. Return "Allowed".

6.3.2.2. Relation to X-Frame-Options

This directive is similar to the X-Frame-Options header that several user agents have implemented. The 'none' source expression is roughly equivalent to that header’s DENY, 'self' to SAMEORIGIN, and so on. The major difference is that many user agents implement SAMEORIGIN such that it only matches against the top-level document’s location, while the frame-ancestors directive checks against each ancestor. If _any_ ancestor doesn’t match, the load is cancelled. [RFC7034]

In order to allow backwards-compatible deployment, the frame-ancestors directive _obsoletes_ the X-Frame-Options header. If a resource is delivered with an policy that includes a directive named frame-ancestors and whose disposition is "enforce", then the X-Frame-Options header MUST be ignored.

Spell this out in more detail as part of defining X-Frame-Options integration with the process a navigate response algorithm. <https://github.com/whatwg/html/issues/1230>

6.3.3. navigation-to

The navigation-to directive restricts the URLs to which a document can navigate by any means (a, form, window.location, window.open, etc.). The directive’s syntax is described by the following ABNF grammar:

directive-name  = "navigation-to"
directive-value = serialized-source-list

Should we use ancestor-source-list (basically, origins as opposed to paths?) It doesn’t appear that blocking navigation targets is any worse than blocking any other request type with regard to leakage. Given the redirect behavior, this devolves to an origin check in the presence of a malicious party anyway...

Given a request (request), a string (type, "form-submission or "other") and two browsing contexts (source and target), this algorithm returns "Blocked" if the navigation violates the navigation-to directive’s constraints, and "Allowed" otherwise. This constitutes the navigation-to' directive’s pre-navigation check:

  1. Assert: source and target are unused in this algorithm, as navigation-to is concerned only with details of the outgoing request. Likewise, type is unused, as all navigation requests are treated identically.

  2. If the result of executing §6.6.1.3 Does request match source list? on request and this directive’s value is "Does Not Match", return "Blocked".

  3. Return "Allowed".

6.4. Reporting Directives

Various algorithms in this document hook into the reporting process by constructing a violation object via §2.4.2 Create a violation object for request, policy, and directive or §2.4.1 Create a violation object for global, policy, and directive, and passing that object to §5.3 Report a violation to deliver the report.

6.4.1. report-uri

Note: The report-uri directive is deprecated. Please use the report-to directive instead. If the latter directive is present, this directive will be ignored. To ensure backwards compatibility, we suggest specifying both, like this:
Content-Security-Policy: ...; report-uri https://endpoint.com; report-to groupname

The report-uri directive defines a set of endpoints to which violation reports will be sent when particular behaviors are prevented.

directive-name  = "report-uri"
directive-value = uri-reference *( RWS uri-reference )

; The uri-reference grammar is defined in Section 4.1 of RFC 3986.

The directive has no effect in and of itself, but only gains meaning in combination with other directives.

6.4.2. report-to

The report-to directive defines a reporting group to which violation reports ought to be sent [REPORTING]. The directive’s behavior is defined in §5.3 Report a violation. The directive’s name and value are described by the following ABNF:

directive-name  = "report-to"
directive-value = token

6.5. Directives Defined in Other Documents

This document defines a core set of directives, and sets up a framework for modular extension by other specifications. At the time this document was produced, the following stable documents extend CSP:

Extensions to CSP MUST register themselves via the process outlined in [RFC7762]. In particular, note the criteria discussed in Section 4.2 of that document.

New directives SHOULD use the pre-request check, post-request check, response check, and initialization hooks in order to integrate themselves into Fetch and HTML.

6.6. Matching Algorithms

6.6.1. URL Matching

6.6.1.1. Does request violate policy?

Given a request (request) and a policy (policy), this algorithm returns the violated directive if the request violates the policy, and "Does Not Violate" otherwise.

  1. Let violates be "Does Not Violate".

  2. For each directive in policy:

    1. Let result be the result of executing directive’s pre-request check on request and policy.

    2. If result is "Blocked", then let violates be directive.

  3. Return violates.

6.6.1.2. Does nonce match source list?

Given a request’s cryptographic nonce metadata (nonce) and a source list (source list), this algorithm returns "Matches" if the nonce matches one or more source expressions in the list, and "Does Not Match" otherwise:

  1. Assert: source list is not null.

  2. If nonce is the empty string, return "Does Not Match".

  3. For each expression in source list:

    1. If expression matches the nonce-source grammar, and nonce is a case-sensitive match for expression’s base64-value part, return "Matches".

  4. Return "Does Not Match".

6.6.1.3. Does request match source list?

Given a request (request), and a source list (source list), this algorithm returns the result of executing §6.6.1.5 Does url match source list in origin with redirect count? on request’s current url, source list, request’s origin, and request’s redirect count.

Note: This is generally used in directives' pre-request check algorithms to verify that a given request is reasonable.

6.6.1.4. Does response to request match source list?

Given a request (request), and a source list (source list), this algorithm returns the result of executing §6.6.1.5 Does url match source list in origin with redirect count? on response’s url, source list, request’s origin, and request’s redirect count.

Note: This is generally used in directives' post-request check algorithms to verify that a given response is reasonable.

6.6.1.5. Does url match source list in origin with redirect count?

Given a URL (url), a source list (source list), an origin (origin), and a number (redirect count), this algorithm returns "Matches" if the URL matches one or more source expressions in source list, or "Does Not Match" otherwise:

  1. Assert: source list is not null.

  2. If source list is an empty list, return "Does Not Match".

  3. If source list contains a single item which is an ASCII case-insensitive match for the string "'none'", return "Does Not Match".

    Note: An empty source list (that is, a directive without a value: script-src, as opposed to script-src host1) is equivalent to a source list containing 'none', and will not match any URL.

  4. For each expression in source list:

    1. If §6.6.1.6 Does url match expression in origin with redirect count? returns "Matches" when executed upon url, expression, origin, and redirect count, return "Matches".

  5. Return "Does Not Match".

6.6.1.6. Does url match expression in origin with redirect count?

Given a URL (url), a source expression (expression), an origin (origin), and a number (redirect count), this algorithm returns "Matches" if url matches expression, and "Does Not Match" otherwise.

Note: origin is the origin of the resource relative to which the expression should be resolved. "'self'", for instance, will have distinct meaning depending on that bit of context.

  1. If expression is the string "*", return "Matches" if one or more of the following conditions is met:

    1. url’s scheme is a network scheme.

    2. url’s scheme is the same as origin’s scheme.

    Note: This logic means that in order to allow resource from a non-network scheme, it has to be either explicitly specified (e.g. default-src * data: custom-scheme-1: custom-scheme-2:, or the protected resource must be loaded from the same scheme.

  2. If expression matches the scheme-source or host-source grammar:

    1. If expression has a scheme-part, and it does not scheme-part match url’s scheme, return "Does Not Match".

    2. If expression matches the scheme-source grammar, return "Matches".

  3. If expression matches the host-source grammar:

    1. If url’s host is null, return "Does Not Match".

    2. If expression does not have a scheme-part, and origin’s scheme does not scheme-part match url’s scheme, return "Does Not Match".

      Note: As with scheme-part above, we allow schemeless host-source expressions to be upgraded from insecure schemes to secure schemes.

    3. If expression’s host-part does not host-part match url’s host, return "Does Not Match".

    4. Let port-part be expression’s port-part if present, and null otherwise.

    5. If port-part does not port-part match url’s port and url’s scheme, return "Does Not Match".

    6. If expression contains a non-empty path-part, and redirect count is 0, then:

      1. Let path be the resulting of joining url’s path on the U+002F SOLIDUS character (/).

      2. If expression’s path-part does not path-part match path, return "Does Not Match".

    7. Return "Matches".

  4. If expression is an ASCII case-insensitive match for "'self'", return "Matches" if one or more of the following conditions is met:

    1. origin is the same as url’s origin

    2. origin’s host is the same as url’s host, origin’s port and url’s {{URL/port} are either the same or the default ports for their respective schemes, and one or more of the following conditions is met:

      1. url’s scheme is "https" or "wss"

      2. origin’s scheme is "http"

    Note: Like the scheme-part logic above, the "'self'" matching algorithm allows upgrades to secure schemes when it is safe to do so. We limit these upgrades to endpoints running on the default port for a particular scheme or a port that matches the origin of the protected resource, as this seems sufficient to deal with upgrades that can be reasonably expected to succeed.

  5. Return "Does Not Match".

6.6.1.7. scheme-part matching

An ASCII string scheme-part matches another ASCII string if a CSP source expression that contained the first as a scheme-part could potentially match a URL containing the latter as a scheme. For example, we say that "http" scheme-part matches "https".

Note: The matching relation is asymmetric. For example, the source expressions https: and https://example.com/ do not match the URL http://example.com/. We always allow a secure upgrade from an explicitly insecure expression. script-src http: is treated as equivalent to script-src http: https:, script-src http://example.com to script-src http://example.com https://example.com, and connect-src ws: to connect-src ws: wss:.

More formally, two ASCII strings (A and B) are said to scheme-part match if the following algorithm returns "Matches":

  1. If one of the following is true, return "Matches":

    1. A is an ASCII case-insensitive match for B.

    2. A is an ASCII case-insensitive match for "http", and B is an ASCII case-insensitive match for "https".

    3. A is an ASCII case-insensitive match for "ws", and B is an ASCII case-insensitive match for "wss", "http", or "https".

    4. A is an ASCII case-insensitive match for "wss", and B is an ASCII case-insensitive match for "https".

  2. Return "Does Not Match".

6.6.1.8. host-part matching

An ASCII string host-part matches another ASCII string if a CSP source expression that contained the first as a host-part could potentially match a URL containing the latter as a host. For example, we say that "www.example.com" host-part matches "www.example.com".

More formally, two ASCII strings (A and B) are said to host-part match if the following algorithm returns "Matches":

Note: The matching relation is asymmetric. That is, A matching B does not mean that B will match A. For example, *.example.com host-part matches www.example.com, but www.example.com does not host-part match *.example.com.

  1. If the first character of A is an U+002A ASTERISK character (*):

    1. Let remaining be the result of removing the leading ("*") from A.

    2. If remaining (including the leading U+002E FULL STOP character (.)) is an ASCII case-insensitive match for the rightmost characters of B, then return "Matches". Otherwise, return "Does Not Match".

  2. If A is not an ASCII case-insensitive match for B, return "Does Not Match".

  3. If A matches the IPv4address rule from [RFC3986], and is not "127.0.0.1"; or if A is an IPv6 address, return "Does Not Match".

    Note: A future version of this specification may allow literal IPv6 and IPv4 addresses, depending on usage and demand. Given the weak security properties of IP addresses in relation to named hosts, however, authors are encouraged to prefer the latter whenever possible.

  4. Return "Matches".

6.6.1.9. port-part matching

An ASCII string (port A) port-part matches two other ASCII strings (port B and scheme B) if a CSP source expression that contained the first as a port-part could potentially match a URL containing the latter as port and scheme. For example, "80" port-part matches matches "80"/"http".

  1. If port A is empty:

    1. If port B is the default port for scheme B, return "Matches". Otherwise, return "Does Not Match".

  2. If port A is equal to "*", return "Matches".

  3. If port A is a case-sensitive match for port B, return "Matches".

  4. If port B is empty:

    1. If port A is the default port for scheme B, return "Matches". Otherwise, return "Does not Match".

  5. Return "Does Not Match".

6.6.1.10. path-part matching

An ASCII string (path A) path-part matches another ASCII string (path B) if a CSP source expression that contained the first as a path-part could potentially match a URL containing the latter as a path. For example, we say that "/subdirectory/" path-part matches "/subdirectory/file".

Note: The matching relation is asymmetric. That is, path A matching path B does not mean that path B will match path A.

  1. If path A is empty, return "Matches".

  2. If path A consists of one character that is equal to the U+002F SOLIDUS character (/) and path B is empty, return "Matches".

  3. Let exact match be false if the final character of path A is the U+002F SOLIDUS character (/), and true otherwise.

  4. Let path list A and path list B be the result of strictly splitting path A and path B respestively on the U+002F SOLIDUS character (/).

  5. If path list A has more items than path list B, return "Does Not Match".

  6. If exact match is true, and path list A does not have the same number of items as path list B, return "Does Not Match".

  7. If exact match is false:

    1. Assert: the final item in path list A is the empty string.

    2. Remove the final item from path list A.

  8. For each piece A in path list A:

    1. Let piece B be the next item in path list B.

    2. Percent decode piece A.

    3. Percent decode piece B.

    4. If piece A is not a case-sensitive match for piece B, return "Does Not Match".

  9. Return "Matches".

6.6.1.11. Get the effective directive for request

Each fetch directive controls a specific type of request. Given a request (request), the following algorithm returns either null or the name of the request’s effective directive:

  1. Switch on request’s type, and execute the associated steps:

    ""
    1. If the request’s initiator and destination are both the empty string, return connect-src.

    2. If the request’s initiator is "manifest", return manifest-src.

    3. If the request’s destination is "subresource", return connect-src.

    4. If the request’s destination is "unknown", return object-src.

    5. If the request’s destination is "document" and the request’s target browsing context is a nested browsing context, return frame-src.

    "audio"
    "track"
    "video"
    1. Return media-src.

    "font"
    1. Return font-src.

    "image"
    1. Return img-src.

    "style"
    1. Return style-src.

    "script"
    1. Switch on request’s destination, and execute the associated steps:

      "script"
      "subresource"
      1. Return script-src.

      "serviceworker"
      "sharedworker"
      "worker"
      1. Return worker-src.

  2. Return null.

6.6.2. Element Matching Algorithms

6.6.2.1. Is element nonceable?

Given an Element (element), this algorithm returns "Nonceable" if a nonce-source expression can match the element (as discussed in §7.2 Nonce Stealing), and "Not Nonceable" if such expressions should not be applied.

  1. If element does not have an attribute named "nonce", return "Not Nonceable".

  2. If element is a script element, then for each attribute in element:

    1. If attribute’s name is an ASCII case-insensitive match for the string "<script" or the string "<style", return "Not Nonceable".

    2. If attribute’s value contains an ASCII case-insensitive match the string "<script" or the string "<style", return "Not Nonceable".

  3. Return "Nonceable".

This processing is meant to mitigate the risk of dangling markup attacks that steal the nonce from an existing element in order to load injected script. It is fairly expensive, however, as it requires that we walk through all attributes and their values in order to determine whether the script should execute. Here, we try to minimize the impact by doing this check only for script elements when a nonce is present, but we should probably consider this algorithm as "at risk" until we know its impact. <https://github.com/w3c/webappsec-csp/issues/98>

6.6.2.2. Does a source list allow all inline behavior for type?

A source list allows all inline behavior of a given type if it contains the keyword-source expression 'unsafe-inline', and does not override that expression as described in the following algorithm:

Given a source list (list) and a string (type), the following algorithm returns "Allows" if all inline content of a given type is allowed and "Does Not Allow" otherwise.

  1. Let allow all inline be false.

  2. For each expression in list:

    1. If expression matches the nonce-source or hash-source grammar, return "Does Not Allow".

    2. If type is "script" or "script attribute" and expression matches the keyword-source "'strict-dynamic'", return "Does Not Allow".

      Note: 'strict-dynamic' only applies to scripts, not other resource types. Usage is explained in more detail in §8.2 Usage of "'strict-dynamic'".

    3. If expression is an ASCII case-insensitive match for the keyword-source "'unsafe-inline'", set allow all inline to true.

  3. If allow all inline is true, return "Allows". Otherwise, return "Does Not Allow".

Source lists that allow all inline behavior:
'unsafe-inline' http://a.com http://b.com
'unsafe-inline'

Source lists that do not allow all inline behavior due to the presence of nonces and/or hashes, or absence of 'unsafe-inline':

'sha512-321cba' 'nonce-abc'
http://example.com 'unsafe-inline' 'nonce-abc'

Source lists that do not allow all inline behavior when type is 'script' or 'script attribute' due to the presence of 'strict-dynamic', but allow all inline behavior otherwise:

'unsafe-inline' 'strict-dynamic'
http://example.com 'strict-dynamic' 'unsafe-inline'
6.6.2.3. Does element match source list for type and source?

Given an Element (element), a source list (list), a string (type), and a string (source), this algorithm returns "Matches" or "Does Not Match".

  1. Assert: source contains the value of a script element’s text IDL attribute, the value of a style element’s textContent IDL attribute, or the value of one of a script element’s event handler IDL attribute.

    Note: This means that source will be interpreted with the encoding of the page in which it is embedded. See the integration points in §4.2 Integration with HTML for more detail.

  2. If §6.6.2.2 Does a source list allow all inline behavior for type? returns "Allows" given list and type, return "Matches".

  3. If type is "script" or "style", and §6.6.2.1 Is element nonceable? returns "Nonceable" when executed upon element:

    1. For each expression in list:

      1. If expression matches the nonce-source grammar, and element has a nonce attribute whose value is a case-sensitive match for expression’s base64-value part, return "Matches".

    Note: Nonces only apply to inline script and inline style, not to attributes of either element.

  4. Let hashes match attributes be false.

  5. For each expression in list:

    1. If expression is an ASCII case-insensitive match for the keyword-source "'unsafe-hashed-attributes'", set hashes match attributes to true. Break out of the loop.

  6. If type is "script" or "style", or hashes match attributes is true:

    1. For each expression in list:

      1. If expression matches the hash-source grammar:

        1. Let algorithm be null.

        2. If expression’s hash-algorithm part is an ASCII case-insensitive match for "sha256", set algorithm to SHA-256.

        3. If expression’s hash-algorithm part is an ASCII case-insensitive match for "sha384", set algorithm to SHA-384.

        4. If expression’s hash-algorithm part is an ASCII case-insensitive match for "sha512", set algorithm to SHA-512.

        5. If algorithm is not null:

          1. Let actual be the result of base64 encoding the result of applying algorithm to source.

          2. Let expected be expression’s base64-value part, with all '-' characters replaced with '+', and all '_' characters replaced with '/'.

            Note: This replacement normalizes hashes expressed in base64url encoding into base64 encoding for matching.

          3. If actual is a case-sensitive match for expected, return "Matches".

    Note: Hashes apply to inline script and inline style. If the "'unsafe-hashed-attributes'" source expression is present, they will also apply to event handlers and style attributes.

  7. Return "Does Not Match".

7. Security and Privacy Considerations

7.1. Nonce Reuse

Nonces override the other restrictions present in the directive in which they’re delivered. It is critical, then, that they remain unguessable, as bypassing a resource’s policy is otherwise trivial.

If a server delivers a nonce-source expression as part of a policy, the server MUST generate a unique value each time it transmits a policy. The generated value SHOULD be at least 128 bits long (before encoding), and SHOULD be generated via a cryptographically secure random number generator in order to ensure that the value is difficult for an attacker to predict.

Note: Using a nonce to allow inline script or style is less secure than not using a nonce, as nonces override the restrictions in the directive in which they are present. An attacker who can gain access to the nonce can execute whatever script they like, whenever they like. That said, nonces provide a substantial improvement over 'unsafe-inline' when layering a content security policy on top of old code. When considering 'unsafe-inline', authors are encouraged to consider nonces (or hashes) instead.

7.2. Nonce Stealing

Dangling markup attacks such as those discussed in [FILEDESCRIPTOR-2015] can be used to repurpose a page’s legitimate nonces for injections. For example, given an injection point before a script element:

<p>Hello, [INJECTION POINT]</p>
<script nonce=abc src=/good.js></script>

If an attacker injects the string "<script src='https://evil.com/evil.js' ", then the browser will receive the following:

<p>Hello, <script src='https://evil.com/evil.js' </p>
<script nonce=abc src=/good.js></script>

It will then parse that code, ending up with a script element with a src attribute pointing to a malicious payload, an attribute named </p>, an attribute named "<script", a nonce attribute, and a second src attribute which is helpfully discarded as duplicate by the parser.

The §6.6.2.1 Is element nonceable? algorithm attempts to mitigate this specific attack by walking through script element attributes, looking for the string "<script" or "<style" in their names or values.

7.3. Nonce Retargeting

Nonces bypass host-source expressions, enabling developers to load code from any origin. This, generally, is fine, and desirable from the developer’s perspective. However, if an attacker can inject a base element, then an otherwise safe page can be subverted when relative URLs are resolved. That is, on https://example.com/ the following code will load https://example.com/good.js:

<script nonce=abc src=/good.js></script>

However, the following will load https://evil.com/good.js:

<base href="https://evil.com">
<script nonce=abc src=/good.js></script>

To mitigate this risk, it is advisable to set an explicit base element on every page, or to limit the ability of an attacker to inject their own base element by setting a base-uri directive in your page’s policy. For example, base-uri 'none'.

7.4. CSS Parsing

The style-src directive restricts the locations from which the protected resource can load styles. However, if the user agent uses a lax CSS parsing algorithm, an attacker might be able to trick the user agent into accepting malicious "stylesheets" hosted by an otherwise trustworthy origin.

These attacks are similar to the CSS cross-origin data leakage attack described by Chris Evans in 2009 [CSS-ABUSE]. User agents SHOULD defend against both attacks using the same mechanism: stricter CSS parsing rules for style sheets with improper MIME types.

7.5. Violation Reports

The violation reporting mechanism in this document has been designed to mitigate the risk that a malicious web site could use violation reports to probe the behavior of other servers. For example, consider a malicious web site that allows https://example.com as a source of images. If the malicious site attempts to load https://example.com/login as an image, and the example.com server redirects to an identity provider (e.g. identityprovider.example.net), CSP will block the request. If violation reports contained the full blocked URL, the violation report might contain sensitive information contained in the redirected URL, such as session identifiers or purported identities. For this reason, the user agent includes only the URL of the original request, not the redirect target.

Note also that violation reports should be considered attacker-controlled data. Developers who wish to collect violation reports in a dashboard or similar service should be careful to properly escape their content before rendering it (and should probably themselves use CSP to further mitigate the risk of injection). This is especially true for the "script-sample" property of violation reports, and the sample property of SecurityPolicyViolationEvent, which are both completely attacker-controlled strings.

7.6. Paths and Redirects

To avoid leaking path information cross-origin (as discussed in Egor Homakov’s Using Content-Security-Policy for Evil), the matching algorithm ignores the path component of a source expression if the resource being loaded is the result of a redirect. For example, given a page with an active policy of img-src example.com not-example.com/path:

This restriction reduces the granularity of a document’s policy when redirects are in play, a necessary compromise to avoid brute-forced information leaks of this type.

The relatively long thread "Remove paths from CSP?" from public-webappsec@w3.org has more detailed discussion around alternate proposals.

7.7. Secure Upgrades

To mitigate one variant of history-scanning attacks like Yan Zhu’s Sniffly, CSP will not allow pages to lock themselves into insecure URLs via policies like script-src http://example.com. As described in §6.6.1.7 scheme-part matching, the scheme portion of a source expression will always allow upgrading to a secure variant.

8. Authoring Considerations

8.1. The effect of multiple policies

This section is not normative.

The above sections note that when multiple policies are present, each must be enforced or reported, according to its type. An example will help clarify how that ought to work in practice. The behavior of an XMLHttpRequest might seem unclear given a site that, for whatever reason, delivered the following HTTP headers:

Content-Security-Policy: default-src 'self' http://example.com http://example.net;
                         connect-src 'none';
Content-Security-Policy: connect-src http://example.com/;
                         script-src http://example.com/

Is a connection to example.com allowed or not? The short answer is that the connection is not allowed. Enforcing both policies means that a potential connection would have to pass through both unscathed. Even though the second policy would allow this connection, the first policy contains connect-src 'none', so its enforcement blocks the connection. The impact is that adding additional policies to the list of policies to enforce can only further restrict the capabilities of the protected resource.

To demonstrate that further, consider a script tag on this page. The first policy would lock scripts down to 'self', http://example.com and http://example.net via the default-src directive. The second, however, would only allow script from http://example.com/. Script will only load if it meets both policy’s criteria: in this case, the only origin that can match is http://example.com, as both policies allow it.

8.2. Usage of "'strict-dynamic'"

Host- and path-based policies are tough to get right, especially on sprawling origins like CDNs. The solutions to Cure53’s H5SC Minichallenge 3: "Sh*t, it’s CSP!" [H5SC3] are good examples of the kinds of bypasses which such policies can enable, and though CSP is capable of mitigating these bypasses via exhaustive declaration of specific resources, those lists end up being brittle, awkward, and difficult to implement and maintain.

The "'strict-dynamic'" source expression aims to make Content Security Policy simpler to deploy for existing applications who have a high degree of confidence in the scripts they load directly, but low confidence in their ability to provide a reasonable list of resources to load up front.

If present in a script-src or default-src directive, it has two main effects:

  1. host-source and scheme-source expressions, as well as the "'unsafe-inline'" and "'self' keyword-sources will be ignored when loading script.

    hash-source and nonce-source expressions will be honored.

  2. Script requests which are triggered by non-"parser-inserted" script elements are allowed.

The first change allows you to deploy "'strict-dynamic' in a backwards compatible way, without requiring user-agent sniffing: the policy 'unsafe-inline' https: 'nonce-abcdefg' 'strict-dynamic' will act like 'unsafe-inline' https: in browsers that support CSP1, https: 'nonce-DhcnhD3khTMePgXwdayK9BsMqXjhguVV' in browsers that support CSP2, and 'nonce-DhcnhD3khTMePgXwdayK9BsMqXjhguVV' 'strict-dynamic' in browsers that support CSP3.

The second allows scripts which are given access to the page via nonces or hashes to bring in their dependencies without adding them explicitly to the page’s policy.

Suppose MegaCorp, Inc. deploys the following policy:
Content-Security-Policy: script-src 'nonce-DhcnhD3khTMePgXwdayK9BsMqXjhguVV' 'strict-dynamic'

And serves the following HTML with that policy active:

...
<script src="https://cdn.example.com/script.js" nonce="DhcnhD3khTMePgXwdayK9BsMqXjhguVV" ></script>
...

This will generate a request for https://cdn.example.com/script.js, which will not be blocked because of the matching nonce attribute.

If script.js contains the following code:

var s = document.createElement('script');
s.src = 'https://othercdn.not-example.net/dependency.js';
document.head.appendChild('s');

document.write('<scr' + 'ipt src='/sadness.js'></scr' + 'ipt>');

dependency.js will load, as the script element created by createElement() is not "parser-inserted".

sadness.js will not load, however, as document.write() produces script elements which are "parser-inserted".

8.3. Usage of "'unsafe-hashed-attributes'"

This section is not normative.

Work in progress. <https://github.com/w3c/webappsec-csp/issues/13>

Legacy websites and websites with legacy dependencies might find it difficult to entirely externalize event handlers. These sites could enable such handlers by allowing 'unsafe-inline', but that’s a big hammer with a lot of associated risk (and cannot be used in conjunction with nonces or hashes).

The "'unsafe-hashed-attributes'" source expression aims to make CSP deployment simpler and safer in these situations by allowing developers to enable specific handlers via hashes.

MegaCorp, Inc. can’t quite get rid of the following HTML on anything resembling a reasonable schedule:
<button id="action" onclick="doSubmit()">

Rather than reducing security by specifying "'unsafe-inline'", they decide to use "'unsafe-hashed-attributes'" along with a hash source expression, as follows:

Content-Security-Policy:  script-src 'unsafe-hashed-attributes' 'sha256-jzgBGA4UWFFmpOBq0JpdsySukE1FrEN5bUpoK8Z29fY='

The capabilities 'unsafe-hashed-attributes' provides is useful for legacy sites, but should be avoided for modern sites. In particular, note that hashes allow a particular script to execute, but do not ensure that it executes in the way a developer intends. If an interesting capability is exposed as an inline event handler (say <a onclick="transferAllMyMoney()">Transfer</a>), then that script becomes available for an attacker to inject as <script>transferAllMyMoney()</script>. Developers should be careful to balance the risk of allowing specific scripts to execute against the deployment advantages that allowing inline event handlers might provide.

8.4. Allowing external JavaScript via hashes

In [CSP2], hash source expressions could only match inlined script, but now that Subresource Integrity is widely deployed, we can expand the scope to enable externalized JavaScript as well.

If multiple sets of integrity metadata are specified for a script, the request will match a policy’s hash-sources if and only if each item in a script's integrity metadata matches the policy.

MegaCorp, Inc. wishes to allow two specific scripts on a page in a way that ensures that the content matches their expectations. They do so by setting the following policy:
Content-Security-Policy: script-src 'sha256-abc123' 'sha512-321cba'

In the presence of that policy, the following script elements would be allowed to execute because they contain only integrity metadata that matches the policy:

<script integrity="sha256-abc123" ...></script>
<script integrity="sha512-321cba" ...></script>
<script integrity="sha256-abc123 sha512-321cba" ...></script>

While the following script elements would not execute because they contain valid metadata that does not match the policy (even though other metadata does match):

<script integrity="sha384-xyz789" ...></script>
<script integrity="sha384-xyz789 sha512-321cba" ...></script>
<script integrity="sha256-abc123 sha384-xyz789 sha512-321cba" ...></script>

Metadata that is not recognized (either because it’s entirely invalid, or because it specifies a not-yet-supported hashing algorithm) does not affect the behavior described here. That is, the following elements would be allowed to execute in the presence of the above policy, as the additional metadata is invalid and therefore wouldn’t allow a script whose content wasn’t listed explicitly in the policy to execute:

<script integrity="sha256-abc123 sha1024-abcd" ...></script>
<script integrity="sha512-321cba entirely-invalid" ...></script>
<script integrity="sha256-abc123 not-a-hash-at-all sha512-321cba" ...></script>

9. Implementation Considerations

9.1. Vendor-specific Extensions and Addons

Policy enforced on a resource SHOULD NOT interfere with the operation of user-agent features like addons, extensions, or bookmarklets. These kinds of features generally advance the user’s priority over page authors, as espoused in [HTML-DESIGN].

Moreover, applying CSP to these kinds of features produces a substantial amount of noise in violation reports, significantly reducing their value to developers.

Chrome, for example, excludes the chrome-extension: scheme from CSP checks, and does some work to ensure that extension-driven injections are allowed, regardless of a page’s policy.

10. IANA Considerations

10.1. Directive Registry

The Content Security Policy Directive registry should be updated with the following directives and references [RFC7762]:

base-uri

This document (see §6.2.1 base-uri)

child-src

This document (see §6.1.1 child-src)

connect-src

This document (see §6.1.2 connect-src)

default-src

This document (see §6.1.3 default-src)

disown-opener

This document (see §6.2.4 disown-opener)

font-src

This document (see §6.1.4 font-src)

form-action

This document (see §6.3.1 form-action)

frame-ancestors

This document (see §6.3.2 frame-ancestors)

frame-src

This document (see §6.1.5 frame-src)

img-src

This document (see §6.1.6 img-src)

manifest-src

This document (see §6.1.7 manifest-src)

media-src

This document (see §6.1.8 media-src)

object-src

This document (see §6.1.9 object-src)

plugin-types

This document (see §6.2.2 plugin-types)

report-uri

This document (see §6.4.1 report-uri)

report-to

This document (see §6.4.2 report-to)

sandbox

This document (see §6.2.3 sandbox)

script-src

This document (see §6.1.10 script-src)

style-src

This document (see §6.1.11 style-src)

worker-src

This document (see §6.1.12 worker-src)

10.2. Headers

The permanent message header field registry should be updated with the following registrations: [RFC3864]

10.2.1. Content-Security-Policy

Header field name
Content-Security-Policy
Applicable protocol
http
Status
standard
Author/Change controller
W3C
Specification document
This specification (See §3.1 The Content-Security-Policy HTTP Response Header Field)

10.2.2. Content-Security-Policy-Report-Only

Header field name
Content-Security-Policy-Report-Only
Applicable protocol
http
Status
standard
Author/Change controller
W3C
Specification document
This specification (See §3.2 The Content-Security-Policy-Report-Only HTTP Response Header Field)

11. Acknowledgements

Lots of people are awesome. For instance:

Conformance

Document conventions

Conformance requirements are expressed with a combination of descriptive assertions and RFC 2119 terminology. The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “MAY”, and “OPTIONAL” in the normative parts of this document are to be interpreted as described in RFC 2119. However, for readability, these words do not appear in all uppercase letters in this specification.

All of the text of this specification is normative except sections explicitly marked as non-normative, examples, and notes. [RFC2119]

Examples in this specification are introduced with the words “for example” or are set apart from the normative text with class="example", like this:

This is an example of an informative example.

Informative notes begin with the word “Note” and are set apart from the normative text with class="note", like this:

Note, this is an informative note.

Conformant Algorithms

Requirements phrased in the imperative as part of algorithms (such as "strip any leading space characters" or "return false and abort these steps") are to be interpreted with the meaning of the key word ("must", "should", "may", etc) used in introducing the algorithm.

Conformance requirements phrased as algorithms or specific steps can be implemented in any manner, so long as the end result is equivalent. In particular, the algorithms defined in this specification are intended to be easy to understand and are not intended to be performant. Implementers are encouraged to optimize.

Index

Terms defined by this specification

Terms defined by reference

References

Normative References

[CSP3]
Mike West. Content Security Policy Level 3. URL: https://www.w3.org/TR/CSP3/
[CSS-CASCADE-4]
Elika Etemad; Tab Atkins Jr.. CSS Cascading and Inheritance Level 4. URL: https://www.w3.org/TR/css-cascade-4/
[CSSOM]
Simon Pieters; Glenn Adams. CSS Object Model (CSSOM). URL: https://www.w3.org/TR/cssom-1/
[DOM]
Anne van Kesteren. DOM Standard. Living Standard. URL: https://dom.spec.whatwg.org/
[ECMA262]
Brian Terlson; Allen Wirfs-Brock. ECMAScript® Language Specification. URL: https://tc39.github.io/ecma262/
[FETCH]
Anne van Kesteren. Fetch Standard. Living Standard. URL: https://fetch.spec.whatwg.org/
[HTML]
Anne van Kesteren; et al. HTML Standard. Living Standard. URL: https://html.spec.whatwg.org/multipage/
[INFRA]
Anne van Kesteren; Domenic Denicola. Infra Standard. Living Standard. URL: https://infra.spec.whatwg.org/
[MIMESNIFF]
Gordon P. Hemsley. MIME Sniffing Standard. Living Standard. URL: https://mimesniff.spec.whatwg.org/
[REPORTING]
Ilya Gregorik; Mike West. Reporting API. URL: https://wicg.github.io/reporting/
[RFC2045]
N. Freed; N. Borenstein. Multipurpose Internet Mail Extensions (MIME) Part One: Format of Internet Message Bodies. November 1996. Draft Standard. URL: https://tools.ietf.org/html/rfc2045
[RFC2119]
S. Bradner. Key words for use in RFCs to Indicate Requirement Levels. March 1997. Best Current Practice. URL: https://tools.ietf.org/html/rfc2119
[RFC3492]
A. Costello. Punycode: A Bootstring encoding of Unicode for Internationalized Domain Names in Applications (IDNA). March 2003. Proposed Standard. URL: https://tools.ietf.org/html/rfc3492
[RFC3864]
G. Klyne; M. Nottingham; J. Mogul. Registration Procedures for Message Header Fields. September 2004. Best Current Practice. URL: https://tools.ietf.org/html/rfc3864
[RFC3986]
T. Berners-Lee; R. Fielding; L. Masinter. Uniform Resource Identifier (URI): Generic Syntax. January 2005. Internet Standard. URL: https://tools.ietf.org/html/rfc3986
[RFC4648]
S. Josefsson. The Base16, Base32, and Base64 Data Encodings. October 2006. Proposed Standard. URL: https://tools.ietf.org/html/rfc4648
[RFC5234]
D. Crocker, Ed.; P. Overell. Augmented BNF for Syntax Specifications: ABNF. January 2008. Internet Standard. URL: https://tools.ietf.org/html/rfc5234
[RFC5988]
M. Nottingham. Web Linking. October 2010. Proposed Standard. URL: https://tools.ietf.org/html/rfc5988
[RFC7034]
D. Ross; T. Gondrom. HTTP Header Field X-Frame-Options. October 2013. Informational. URL: https://tools.ietf.org/html/rfc7034
[RFC7230]
R. Fielding, Ed.; J. Reschke, Ed.. Hypertext Transfer Protocol (HTTP/1.1): Message Syntax and Routing. June 2014. Proposed Standard. URL: https://tools.ietf.org/html/rfc7230
[RFC7231]
R. Fielding, Ed.; J. Reschke, Ed.. Hypertext Transfer Protocol (HTTP/1.1): Semantics and Content. June 2014. Proposed Standard. URL: https://tools.ietf.org/html/rfc7231
[RFC7762]
M. West. Initial Assignment for the Content Security Policy Directives Registry. January 2016. Informational. URL: https://tools.ietf.org/html/rfc7762
[SERVICE-WORKERS-1]
Alex Russell; et al. Service Workers 1. URL: https://www.w3.org/TR/service-workers-1/
[SHA2]
FIPS PUB 180-4, Secure Hash Standard. URL: http://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.180-4.pdf
[SRI]
Devdatta Akhawe; et al. Subresource Integrity. URL: https://www.w3.org/TR/SRI/
[URL]
Anne van Kesteren. URL Standard. Living Standard. URL: https://url.spec.whatwg.org/
[WebIDL]
Cameron McCormack; Boris Zbarsky; Tobie Langel. Web IDL. URL: https://heycam.github.io/webidl/

Informative References

[APPMANIFEST]
Marcos Caceres; et al. Web App Manifest. URL: https://www.w3.org/TR/appmanifest/
[BEACON]
Ilya Grigorik; et al. Beacon. URL: https://www.w3.org/TR/beacon/
[CSP2]
Mike West; Adam Barth; Daniel Veditz. Content Security Policy Level 2. URL: https://www.w3.org/TR/CSP2/
[CSS-ABUSE]
Chris Evans. Generic cross-browser cross-domain theft. 28 December 2009. URL: https://scarybeastsecurity.blogspot.com/2009/12/generic-cross-browser-cross-domain.html
[EVENTSOURCE]
Ian Hickson. Server-Sent Events. URL: https://www.w3.org/TR/eventsource/
[FILEDESCRIPTOR-2015]
filedescriptor. CSP 2015. 23 November 2015. URL: https://blog.innerht.ml/csp-2015/#danglingmarkupinjection
[H5SC3]
Mario Heiderich. H5SC Minichallenge 3: "Sh*t, it's CSP!". URL: https://github.com/cure53/XSSChallengeWiki/wiki/H5SC-Minichallenge-3:-%22Sh*t,-it%27s-CSP!%22
[HTML-DESIGN]
Anne Van Kesteren; Maciej Stachowiak. HTML Design Principles. URL: https://www.w3.org/TR/html-design-principles/
[MIX]
Mike West. Mixed Content. URL: https://www.w3.org/TR/mixed-content/
[TIMING]
Paul Stone. Pixel Perfect Timing Attacks with HTML5. URL: http://www.contextis.com/documents/2/Browser_Timing_Attacks.pdf
[UISECURITY]
Brad Hill. User Interface Security and the Visibility API. URL: https://www.w3.org/TR/UISecurity/
[UPGRADE-INSECURE-REQUESTS]
Mike West. Upgrade Insecure Requests. URL: https://www.w3.org/TR/upgrade-insecure-requests/
[WEBSOCKETS]
Ian Hickson. The WebSocket API. 20 September 2012. CR. URL: https://www.w3.org/TR/websockets/
[XHR]
Anne van Kesteren. XMLHttpRequest Standard. Living Standard. URL: https://xhr.spec.whatwg.org/
[XSLT]
James Clark. XSL Transformations (XSLT) Version 1.0. 16 November 1999. REC. URL: https://www.w3.org/TR/xslt

IDL Index

enum SecurityPolicyViolationEventDisposition {
  "enforce", "report"
};

[Constructor(DOMString type, optional SecurityPolicyViolationEventInit eventInitDict)]
interface SecurityPolicyViolationEvent : Event {
    readonly    attribute DOMString      documentURI;
    readonly    attribute DOMString      referrer;
    readonly    attribute DOMString      blockedURI;
    readonly    attribute DOMString      violatedDirective;
    readonly    attribute DOMString      effectiveDirective;
    readonly    attribute DOMString      originalPolicy;
    readonly    attribute DOMString      sourceFile;
    readonly    attribute DOMString      sample;
    readonly    attribute SecurityPolicyViolationEventDisposition      disposition;
    readonly    attribute unsigned short statusCode;
    readonly    attribute long           lineNumber;
    readonly    attribute long           columnNumber;
};

dictionary SecurityPolicyViolationEventInit : EventInit {
    DOMString      documentURI;
    DOMString      referrer;
    DOMString      blockedURI;
    DOMString      violatedDirective;
    DOMString      effectiveDirective;
    DOMString      originalPolicy;
    DOMString      sourceFile;
    DOMString      sample;
    SecurityPolicyViolationEventDisposition      disposition;
    unsigned short statusCode;
    long           lineNumber;
    long           columnNumber;
};

Issues Index

unsafe-hashed-attributes is a work in progress. <https://github.com/w3c/webappsec-csp/issues/13>
Is this kind of thing specified anywhere? I didn’t see anything that looked useful in [ECMA262].
How, exactly, do we get the status code? We don’t actually store it anywhere.
This concept is missing from W3C’s Workers. <https://github.com/w3c/html/issues/187>
Stylesheet loading is not yet integrated with Fetch in W3C’s HTML. <https://github.com/whatwg/html/issues/198>
Stylesheet loading is not yet integrated with Fetch in WHATWG’s HTML. <https://github.com/whatwg/html/issues/968>
This hook is missing from W3C’s HTML. <https://github.com/w3c/html/issues/547>
W3C’s HTML is not based on Fetch, and does not have a process a navigate response algorithm into which to hook. <https://github.com/w3c/html/issues/548>
HostEnsureCanCompileStrings() does not include the string which is going to be compiled as a parameter. We’ll also need to update HTML to pipe that value through to CSP. <https://github.com/tc39/ecma262/issues/938>
This needs to be better explained.
Do something interesting to the execution context in order to lock down interesting CSSOM algorithms. I don’t think CSSOM gives us any hooks here, so let’s work with them to put something reasonable together.
Not sure this is the right model. We need to ensure that we take care of the inverse as well, and there might be a cleverer syntax that could encompass both a document’s opener, and a document’s openees. disown-openee is weird. Maybe disown 'opener' 'openee'? Do we need origin restrictions on either/both?
What should this do in an iframe? Anything?
Spell this out in more detail as part of defining X-Frame-Options integration with the process a navigate response algorithm. <https://github.com/whatwg/html/issues/1230>
Should we use ancestor-source-list (basically, origins as opposed to paths?) It doesn’t appear that blocking navigation targets is any worse than blocking any other request type with regard to leakage. Given the redirect behavior, this devolves to an origin check in the presence of a malicious party anyway...
This processing is meant to mitigate the risk of dangling markup attacks that steal the nonce from an existing element in order to load injected script. It is fairly expensive, however, as it requires that we walk through all attributes and their values in order to determine whether the script should execute. Here, we try to minimize the impact by doing this check only for script elements when a nonce is present, but we should probably consider this algorithm as "at risk" until we know its impact. <https://github.com/w3c/webappsec-csp/issues/98>
Work in progress. <https://github.com/w3c/webappsec-csp/issues/13>