From c58393bf0b9e2a86abee916c3cef3361f2741567 Mon Sep 17 00:00:00 2001 From: Anne van Kesteren Date: Thu, 27 Oct 2022 11:49:12 +0200 Subject: [PATCH] Define opaque-response blocking This is good enough for early review, but there are a number of issues that still need resolving: https://github.com/annevk/orb/labels/mvp. There are also some inline TODO comments. A PR against HTML is needed to ensure it passes the appropriate metadata for media element and classic script requests. We might also want to depend on HTML for parsing JavaScript. --- fetch.bs | 312 +++++++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 305 insertions(+), 7 deletions(-) diff --git a/fetch.bs b/fetch.bs index 47eb6b489..f31d5d657 100644 --- a/fetch.bs +++ b/fetch.bs @@ -37,6 +37,9 @@ urlPrefix:https://w3c.github.io/hr-time/#;spec:hr-time urlPrefix:https://tc39.es/ecma262/#;type:dfn;spec:ecma-262 url:realm;text:realm url:sec-list-and-record-specification-type;text:Record + url:sec-parsetext;text:ParseText + url:prod-Script;text:Script + url:script-record;text:Script Record
@@ -1970,6 +1973,17 @@ Unless stated otherwise, it is false.
 
 

This flag is for exclusive use by HTML's render-blocking mechanism. [[!HTML]] +

A request has an associated +no-cors media request state ... + +

This is for exclusive use by the opaque-response-safelist check. + +

A request has an associated +no-cors JavaScript fallback encoding (an encoding). Unless +stated otherwise, it is UTF-8. + +

This is for exclusive use by the opaque-response-safelist check. +


A request has an associated @@ -3042,6 +3056,285 @@ run these steps: +

Opaque-response blocking

+ +
+

Opaque-response blocking, also known as ORB, is a network filter that blocks access + to opaque filtered responses. These responses would likely would not have been useful to the + fetching party. Blocking them reduces information leakage to potential attackers. + +

Essentially, CSS, JavaScript, images, and media (audio and video) can be requested across + origins without the CORS protocol. And unfortunately except for CSS there is no MIME type + enforcement. This algorithm aims to block as many responses as possible that are not one of these + types (or are newer variants of those types) to avoid leaking their contents through side channels. + +

The network filter combines pro-active blocking based on response headers, sniffing a limited + set of bytes, and ultimately falls back to a full parse due to unfortunate (lack of) design + decisions in the early days of the web platform. As a result there are still quite a few responses + whose secrets can end up being revealed to attackers. Web developers are strongly encouraged to use + the `Cross-Origin-Resource-Policy` response header to defend them. +

+ + +

The opaque-response-safelist check

+ +

The opaque-response-safelist check, given a request request +and a response response, is to run these steps: + +

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

  2. Let nosniff be the result of determining nosniff given + response's header list. + +

  3. +

    If mimeType is not failure, then: + +

      +
    1. If mimeType is an opaque-response-safelisted MIME type, then return + true. + +

    2. If mimeType is an opaque-response-blocklisted-never-sniffed MIME type, + then return false. + +

    3. If response's status is 206 and mimeType is an + opaque-response-blocklisted MIME type, then return false. + +

    4. If nosniff is true and mimeType is an + opaque-response-blocklisted MIME type or its essence is + "text/plain", then return false. +

    + +
  4. If request's no-cors media request state is + "subsequent", then return true. + +

  5. If response's status is 206 and + validate a partial response given 0 and response returns invalid, then return + false. + + +

  6. Let bytes be the result of running + obtain a copy of the first 1024 bytes of response given response. + +

  7. If bytes is failure, then return false. + +

  8. +

    If the audio or video type pattern matching algorithm given bytes does not + return undefined, then: + +

      +
    1. If requests's no-cors media request state is not + "initial", then return false. + +

    2. If response's status is not 200 or 206, then return false. + +

    3. Return true. +

    + +
  9. If requests's no-cors media request state is not + "N/A", then return false. + +

  10. If the image type pattern matching algorithm given bytes does not return + undefined, then return true. + +

  11. +

    If nosniff is true, then return false. + +

    This check is made late as unfortunately images and media are always sniffed. + +

  12. If response's status is not an ok status, then return + false. + +

  13. +

    If mimeType is failure, then return true. + +

    This could be improved at somewhat significant cost. See + annevk/orb #28. + +

  14. If mimeType's essence starts with + "audio/", "image/", or "video/", then return false. + +

  15. Return determine if response is JavaScript and not JSON given response. +

+ +
+ +

To obtain a copy of the first 1024 bytes of response, given a response +response, run these steps: + +

    +
  1. Let first1024Bytes be null. + +

  2. +

    In parallel: + +

      +
    1. Let bytes be the empty byte sequence. + +

    2. Let transformStream be a new {{TransformStream}}. + +

    3. +

      Let transformAlgorithm given a chunk be these steps: + +

        +
      1. Enqueue chunk in transformStream. + +

      2. +

        If first1024Bytes is null, then: + +

          +
        1. Let chunkBytes be + a copy of the bytes held by + chunk. + +

        2. Append chunkBytes to bytes. + +

        3. +

          If bytes's length is greater than 1024, then: + +

            +
          1. Truncate bytes from the end so that it only contains 1024 bytes. + +

          2. Set first1024Bytes to bytes. +

          +
        +
      + +
    4. Let flushAlgorithm be this step: if first1024Bytes is null, then set + first1024Bytes to bytes. + +

    5. Set up transformStream with + transformAlgorithm set to + transformAlgorithm and flushAlgorithm set + to flushAlgorithm. + +

    6. Set response's body's stream to the result + of response's body's stream + piped through transformStream. +

    + +
  3. Wait until first1024Bytes is non-null or response's + body's stream is errored. + +

  4. If first1024Bytes is null, then return failure. + +

  5. Return first1024Bytes. +
+ +
+ +

To determine if response is JavaScript and not JSON given a response +response, run these steps:

+ +
    +
  1. Let responseBodyBytes be null. + +

  2. +

    Let processBody given a byte sequence bytes be these steps: + +

      +
    1. Set responseBodyBytes to bytes. + +

    2. Set response's body to the body + of the result of safely extracting bytes. +

    + +
  3. Let processBodyError be this step: set responseBodyBytes to failure. + +

  4. Fully read response's body given processBody + and processBodyError. + +

  5. Wait for responseBodyBytes to be non-null. + +

  6. If responseBodyBytes is failure, then return false. + +

  7. Assert: responseBodyBytes is a byte sequence. + +

  8. +

    If parse JSON bytes to a JavaScript value given responseBodyBytes does not + throw, then return false. If it throws, catch the exception and ignore it. + +

    If there is an exception, response is not JSON. If there is not, it is. + +

  9. Let potentialMIMETypeForEncoding be the result of extracting a MIME type + given response's header list. + +

  10. +

    Let encoding be the result of legacy extracting an encoding given + potentialMIMETypeForEncoding and request's + no-cors JavaScript fallback encoding. + +

    Equivalently to fetch a classic script, this ignores the + MIME type essence. + +

  11. Let sourceText be the result of decoding + responseBodyBytes given encoding. + +

  12. If ParseText(sourceText, Script) returns a Script Record, + then return true. + + +

  13. Return false. +

+ + +

New MIME type sets

+ +

The definitions in this section are solely for the purpose of abstracting parts of the +opaque-response-safelist check. They are not suited for usage elsewhere. + +

An opaque-response-safelisted MIME type is a JavaScript MIME type or a +MIME type whose essence is "text/css" or +"image/svg+xml". + +

An opaque-response-blocklisted MIME type is an HTML MIME type, +JSON MIME type, or XML MIME type. + +

An opaque-response-blocklisted-never-sniffed MIME type is a MIME type +whose essence is one of: + +

+ +

HTTP extensions

@@ -4846,19 +5139,23 @@ these steps:
  • Set response and actualResponse to the result of running HTTP-network-or-cache fetch given fetchParams. -

  • -

    If request's response tainting is "cors" and a - CORS check for request and response returns failure, then return a - network error. +

  • If request's response tainting is "opaque", + response's status is not a redirect status, and the + opaque-response-safelist check given request and response returns + false, then return a network error. -

    As the CORS check is not to be applied to - responses whose status is 304 or 407, or responses - from a service worker for that matter, it is applied here. +

  • If request's response tainting is "cors" and + the CORS check for request and response returns failure, then return + a network error.

  • If the TAO check for request and response returns failure, then set request's timing allow failed flag. +

    As the opaque-response-safelist check, CORS check, and + TAO check are not to be applied to responses whose status + is 304 or 407, or to responses from a service worker, they are applied here. +

  • If either request's response tainting or response's type is "opaque", and the @@ -8421,6 +8718,7 @@ Mohamed Zergaoui, Mohammed Zubair Ahmed, Moritz Kneilmann, Ms2ger, +Nathan Froyd, Nico Schlömer, Nicolás Peña Moreno, Nidhi Jaju,