From 6b53d6c840140fc5df6d7638808b978d96502a35 Mon Sep 17 00:00:00 2001 From: Wojtek Kosior Date: Mon, 23 Aug 2021 11:05:51 +0200 Subject: use StreamFilter under Mozilla to prevent csp tags from blocking our injected scripts --- background/stream_filter.js | 176 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 176 insertions(+) create mode 100644 background/stream_filter.js (limited to 'background/stream_filter.js') diff --git a/background/stream_filter.js b/background/stream_filter.js new file mode 100644 index 0000000..2dce811 --- /dev/null +++ b/background/stream_filter.js @@ -0,0 +1,176 @@ +/** + * Hachette modifying a web page using the StreamFilter API + * + * Copyright (C) 2018 Giorgio Maone + * Copyright (C) 2021 Wojtek Kosior + * Redistribution terms are gathered in the `copyright' file. + * + * Derived from `bg/ResponseProcessor.js' and `bg/ResponseMetaData.js' + * in LibreJS. + */ + +/* + * IMPORTS_START + * IMPORT browser + * IMPORTS_END + */ + +function validate_encoding(charset) +{ + try { + new TextDecoder(); + return charset; + } catch(e) { + return undefined; + } +} + +function is_content_type_header(header) +{ + header.name.toLowerCase().trim() === "content-type"; +} + +const charset_reg = /;\s*charset\s*=\s*([\w-]+)/i; + +function properties_from_headers(headers) +{ + const properties = {}; + + for (const header of headers.filter(is_content_type_header)) { + const match = charset_reg.exec(header.value); + if (!properties.detected_charset && validate_encoding(match[1])) + properties.detected_charset = match[1]; + + if (/html/i.test(header.value)) + properties.html = true; + } + + return properties; +} + +const UTF8_BOM = [0xef, 0xbb, 0xbf]; +const BOMs = [ + [UTF8_BOM, "utf-8"], + [[0xfe, 0xff], "utf-16be"], + [[0xff, 0xfe], "utf-16le"] +]; + +function charset_from_BOM(data) +{ + for (const [BOM, charset] of BOMs) { + if (BOM.reduce((ac, byte, i) => ac && byte === data[i], true)) + return charset; + } + + return ""; +} + +const charset_attrs = + ['charset', 'http-equiv="content-type"', 'content*="charset"']; +const charset_meta_selector = + charset_attrs.map(a => `head>meta[${a}]`).join(", "); + +function charset_from_meta_tags(doc) +{ + for (const meta of doc.querySelectorAll(charset_meta_selector)) { + const maybe_charset = meta.getAttribute("charset"); + if (maybe_charset && validate_encoding(maybe_charset)) + return maybe_charset; + + const match = charset_reg.exec(meta.getAttribute("content")); + if (match && validate_encoding(match[1])) + return match[1]; + } + + return undefined; +} + +function create_decoder(properties, data) +{ + let charset = charset_from_BOM(data) || properties.detected_charset; + if (!charset && data.indexOf(0) !== -1) { + console.debug("Warning: zeroes in bytestream, probable cached encoding mismatch. Trying to decode it as UTF-16.", + properties); + return new TextDecoder("utf-16be"); + } + + /* Missing HTTP charset, sniffing in content... */ + /* + * TODO: I recall there is some standard saying how early in the doc the + * charset has to be specified. We could process just this part of data. + */ + const text = new TextDecoder("latin1").decode(data, {stream: true}); + properties.html = properties.html || /html/i.test(text); + + if (properties.html) { + const tmp_doc = new DOMParser().parseFromString(text, "text/html"); + charset = charset_from_meta_tags(tmp_doc); + } + + return new TextDecoder(charset || "latin1"); +} + +function filter_data(properties, event) +{ + const data = new Uint8Array(event.data); + let first_chunk = false; + if (!properties.decoder) { + first_chunk = true; + properties.decoder = create_decoder(properties, data); + properties.encoder = new TextEncoder(); + /* Force UTF-8, this is the only encoding we can produce. */ + properties.filter.write(new Uint8Array(UTF8_BOM)); + } + + let decoded = properties.decoder.decode(data); + + if (first_chunk) { + /* + * HAX! Our content scripts that execute at `document_start' will always + * run before the first script in the document, but under Mozilla some + * `' tags might already be loaded at that point. Here we inject a + * dummy ``; + const doctype_decl = /^(\s*"']*>)?/i.exec(decoded)[0]; + decoded = doctype_decl + dummy_script + + decoded.substring(doctype_decl.length); + } + + properties.filter.write(properties.encoder.encode(decoded)); + + if (properties.decoder.encoding === "utf-8") + properties.filter.disconnect(); +} + +function apply_stream_filter(details, headers, policy) +{ + if (policy.allow) + return headers; + + const properties = properties_from_headers(headers); + properties.policy = policy; + + properties.filter = + browser.webRequest.filterResponseData(details.requestId); + + properties.filter.ondata = event => filter_data(properties, event); + properties.filter.onstop = () => properties.filter.close(); + + /* + * In the future we might consider modifying the headers that specify + * encoding. For now we are not yet doing it, though. However, we + * prepend the data with UTF-8 BOM which should be enough. + */ + return headers; +} + +/* + * EXPORTS_START + * EXPORT apply_stream_filter + * EXPORTS_END + */ -- cgit v1.2.3 From 03d041ce03f630d2a28494946ae71588e36d257d Mon Sep 17 00:00:00 2001 From: Wojtek Kosior Date: Fri, 3 Sep 2021 18:46:26 +0200 Subject: only apply stream filter modifications when reasonably necessary --- background/main.js | 5 +++-- background/stream_filter.js | 47 +++++++++++++++++++++++++++++++++++++++++---- 2 files changed, 46 insertions(+), 6 deletions(-) (limited to 'background/stream_filter.js') diff --git a/background/main.js b/background/main.js index 85f8ce8..2e9fa50 100644 --- a/background/main.js +++ b/background/main.js @@ -57,9 +57,10 @@ function on_headers_received(details) return; const [pattern, settings] = query_best(storage, details.url); - const allow = !!(settings && settings.allow); + const has_payload = !!(settings && settings.components); + const allow = !has_payload && !!(settings && settings.allow); const nonce = gen_nonce(); - const policy = {allow, url, nonce}; + const policy = {allow, url, nonce, has_payload}; let headers = details.responseHeaders; let skip = false; diff --git a/background/stream_filter.js b/background/stream_filter.js index 2dce811..96b6132 100644 --- a/background/stream_filter.js +++ b/background/stream_filter.js @@ -12,6 +12,7 @@ /* * IMPORTS_START * IMPORT browser + * IMPORT is_csp_header_name * IMPORTS_END */ @@ -110,6 +111,35 @@ function create_decoder(properties, data) return new TextDecoder(charset || "latin1"); } +function may_define_csp_rules(html) +{ + const doc = new DOMParser().parseFromString(html, "text/html"); + + for (const meta of doc.querySelectorAll("head>meta[http-equiv]")) { + if (is_csp_header_name(meta.getAttribute("http-equiv"), true) && + meta.content) + return true; + } + + /* + * Even if no naughty `' tags were found, subsequent chunk of HTML + * data could add some. Before we return `false' we need to be sure we + * reached the start of `' where `' tags are no longer valid. + */ + + if (doc.documentElement.nextSibling || doc.body.nextSibling || + doc.body.childNodes.length > 1) + return false; + + if (!doc.body.firstChild) + return true; + + if (doc.body.firstChild.nodeName !== "#text") + return false; + + return /^(<\/|&#|.)$/.test(doc.body.firstChild.wholeText); +} + function filter_data(properties, event) { const data = new Uint8Array(event.data); @@ -118,13 +148,15 @@ function filter_data(properties, event) first_chunk = true; properties.decoder = create_decoder(properties, data); properties.encoder = new TextEncoder(); - /* Force UTF-8, this is the only encoding we can produce. */ - properties.filter.write(new Uint8Array(UTF8_BOM)); } let decoded = properties.decoder.decode(data); - if (first_chunk) { + /* Force UTF-8, this is the only encoding we can produce. */ + if (first_chunk) + properties.filter.write(new Uint8Array(UTF8_BOM)); + + if (first_chunk && may_define_csp_rules(decoded)) { /* * HAX! Our content scripts that execute at `document_start' will always * run before the first script in the document, but under Mozilla some @@ -133,7 +165,14 @@ function filter_data(properties, event) * will force `document_start' to happen earlier. This way our content * scripts will be able to sanitize `http-equiv' tags with CSP rules * that would otherwise stop our injected scripts from executing. + * + * As we want to only process HTML files that happen to have naughty + * `' tags in `', we use a DOMParser-based heuristic in + * `may_define_rules()'. We don't do any additional MIME sniffing as it + * is too unreliable (and our heuristic will likely mark non-HTML files + * as harmless anyway). */ + const dummy_script = ``; const doctype_decl = /^(\s*"']*>)?/i.exec(decoded)[0]; @@ -149,7 +188,7 @@ function filter_data(properties, event) function apply_stream_filter(details, headers, policy) { - if (policy.allow) + if (!policy.has_payload) return headers; const properties = properties_from_headers(headers); -- cgit v1.2.3 From 44e89d8ec71b441a431c848567f34b9a36f6b982 Mon Sep 17 00:00:00 2001 From: Wojtek Kosior Date: Thu, 9 Sep 2021 17:47:51 +0200 Subject: simplify CSP handling All page's CSP rules are now removed when a payload is to be injected. When there is no payload, CSP rules are not modified but only supplemented with Hachette's own. --- background/policy_injector.js | 30 ++++++++++--------- background/stream_filter.js | 5 ++-- common/misc.js | 68 +++++++------------------------------------ content/main.js | 57 +++++++++++++++++++----------------- 4 files changed, 60 insertions(+), 100 deletions(-) (limited to 'background/stream_filter.js') diff --git a/background/policy_injector.js b/background/policy_injector.js index 72318d4..e5af055 100644 --- a/background/policy_injector.js +++ b/background/policy_injector.js @@ -10,9 +10,8 @@ * IMPORTS_START * IMPORT sign_data * IMPORT extract_signed - * IMPORT sanitize_csp_header - * IMPORT csp_rule - * IMPORT is_csp_header_name + * IMPORT make_csp_rule + * IMPORT csp_header_regex * IMPORTS_END */ @@ -43,22 +42,25 @@ function inject_csp_headers(headers, policy) break; } + if (policy.has_payload) { + csp_headers = []; + const non_csp_headers = []; + const header_list = + h => csp_header_regex.test(h) ? csp_headers : non_csp_headers; + headers.forEach(h => header_list(h.name).push(h)); + headers = non_csp_headers; + } else { + headers.push(...csp_headers || []); + } + if (!hachette_header) { hachette_header = {name: "x-hachette"}; headers.push(hachette_header); } - csp_headers = csp_headers || - headers.filter(h => is_csp_header_name(h.name)); - - /* When blocking remove report-only CSP headers that snitch on us. */ - headers = headers.filter(h => !is_csp_header_name(h.name, !policy.allow)); - if (old_signature) headers = headers.filter(h => h.value.search(old_signature) === -1); - headers.push(...csp_headers.map(h => sanitize_csp_header(h, policy))); - const policy_str = encodeURIComponent(JSON.stringify(policy)); const signed_policy = sign_data(policy_str, new Date().getTime()); const later_30sec = new Date(new Date().getTime() + 30000).toGMTString(); @@ -76,12 +78,12 @@ function inject_csp_headers(headers, policy) hachette_data = encodeURIComponent(JSON.stringify(hachette_data)); hachette_header.value = sign_data(hachette_data, 0).join("_"); - /* To ensure there is a CSP header if required */ - if (!policy.allow) + if (!policy.allow) { headers.push({ name: "content-security-policy", - value: csp_rule(policy.nonce) + value: make_csp_rule(policy) }); + } return headers; } diff --git a/background/stream_filter.js b/background/stream_filter.js index 96b6132..3e30a4b 100644 --- a/background/stream_filter.js +++ b/background/stream_filter.js @@ -12,7 +12,7 @@ /* * IMPORTS_START * IMPORT browser - * IMPORT is_csp_header_name + * IMPORT csp_header_regex * IMPORTS_END */ @@ -116,8 +116,7 @@ function may_define_csp_rules(html) const doc = new DOMParser().parseFromString(html, "text/html"); for (const meta of doc.querySelectorAll("head>meta[http-equiv]")) { - if (is_csp_header_name(meta.getAttribute("http-equiv"), true) && - meta.content) + if (csp_header_regex.test(meta.httpEquiv) && meta.content) return true; } diff --git a/common/misc.js b/common/misc.js index 6adaf1e..6cded84 100644 --- a/common/misc.js +++ b/common/misc.js @@ -43,29 +43,19 @@ function gen_nonce(length=16) return Uint8toHex(randomData); } -/* csp rule that blocks all scripts except for those injected by us */ -function csp_rule(nonce) +/* CSP rule that blocks scripts according to policy's needs. */ +function make_csp_rule(policy) { - const rule = `'nonce-${nonce}'`; - return `script-src ${rule}; script-src-elem ${rule}; script-src-attr 'none'; prefetch-src 'none';`; + let rule = "prefetch-src 'none'; script-src-attr 'none';"; + const script_src = policy.has_payload ? + `'nonce-${policy.nonce}'` : "'none'"; + rule += ` script-src ${script_src}; script-src-elem ${script_src};`; + return rule; } /* Check if some HTTP header might define CSP rules. */ -const csp_header_names = new Set([ - "content-security-policy", - "x-webkit-csp", - "x-content-security-policy" -]); - -const report_only_header_name = "content-security-policy-report-only"; - -function is_csp_header_name(string, include_report_only) -{ - string = string && string.toLowerCase().trim() || ""; - - return (include_report_only && string === report_only_header_name) || - csp_header_names.has(string); -} +const csp_header_regex = + /^\s*(content-security-policy|x-webkit-csp|x-content-security-policy)/i; /* * Print item together with type, e.g. @@ -111,41 +101,6 @@ function parse_csp(csp) { return directives; } -/* Make CSP headers do our bidding, not interfere */ -function sanitize_csp_header(header, policy) -{ - const rule = `'nonce-${policy.nonce}'`; - const csp = parse_csp(header.value); - - if (!policy.allow) { - /* No snitching */ - delete csp['report-to']; - delete csp['report-uri']; - - delete csp['script-src']; - delete csp['script-src-elem']; - - csp['script-src-attr'] = ["'none'"]; - csp['prefetch-src'] = ["'none'"]; - } - - if ('script-src' in csp) - csp['script-src'].push(rule); - else - csp['script-src'] = [rule]; - - if ('script-src-elem' in csp) - csp['script-src-elem'].push(rule); - else - csp['script-src-elem'] = [rule]; - - const new_csp = Object.entries(csp).map( - i => `${i[0]} ${i[1].join(' ')};` - ); - - return {name: header.name, value: new_csp.join('')}; -} - /* Regexes and objects to use as/in schemas for parse_json_with_schema(). */ const nonempty_string_matcher = /.+/; @@ -161,12 +116,11 @@ const matchers = { /* * EXPORTS_START * EXPORT gen_nonce - * EXPORT csp_rule - * EXPORT is_csp_header_name + * EXPORT make_csp_rule + * EXPORT csp_header_regex * EXPORT nice_name * EXPORT open_in_settings * EXPORT is_privileged_url - * EXPORT sanitize_csp_header * EXPORT matchers * EXPORTS_END */ diff --git a/content/main.js b/content/main.js index fb334dd..a26f72d 100644 --- a/content/main.js +++ b/content/main.js @@ -16,9 +16,8 @@ * IMPORT is_chrome * IMPORT is_mozilla * IMPORT start_activity_info_server - * IMPORT csp_rule - * IMPORT is_csp_header_name - * IMPORT sanitize_csp_header + * IMPORT make_csp_rule + * IMPORT csp_header_regex * IMPORTS_END */ @@ -172,22 +171,20 @@ function block_attribute(node, attr, ns=null) const name = construct_name.join(""); seta(node, `${blocked_str}-${name}`, geta(node, name)); } -} - -function sanitize_meta(meta, policy) -{ - const value = meta.content || ""; - if (!value || !is_csp_header_name(meta.httpEquiv || "", true)) - return; - - block_attribute(meta, "content"); + rema(node, attr); } /* - * Used to disable `; + ``; const doctype_decl = /^(\s*"']*>)?/i.exec(decoded)[0]; decoded = doctype_decl + dummy_script + decoded.substring(doctype_decl.length); diff --git a/common/ajax.js b/common/ajax.js index 8082bbe..7269a8a 100644 --- a/common/ajax.js +++ b/common/ajax.js @@ -1,6 +1,7 @@ /** - * part of Hachette - * Wrapping XMLHttpRequest into a Promise. + * This file is part of Haketilo. + * + * Function: Wrapping XMLHttpRequest into a Promise. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/common/connection_types.js b/common/connection_types.js index 88c6964..3e9df56 100644 --- a/common/connection_types.js +++ b/common/connection_types.js @@ -1,5 +1,7 @@ /** - * Hachette background scripts message connection types "enum" + * This file is part of Haketilo. + * + * Function: Define an "enum" of message connection types. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/common/lock.js b/common/lock.js index 822ad1b..6cf0835 100644 --- a/common/lock.js +++ b/common/lock.js @@ -1,5 +1,7 @@ /** - * Hachette lock (aka binary semaphore aka mutex) + * This file is part of Haketilo. + * + * Function: Implement a lock (aka binary semaphore aka mutex). * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/common/message_server.js b/common/message_server.js index ea40487..c8c6696 100644 --- a/common/message_server.js +++ b/common/message_server.js @@ -1,5 +1,7 @@ /** - * Hachette message server + * This file is part of Haketilo. + * + * Function: Message server. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/common/misc.js b/common/misc.js index 6cded84..9ffb7ff 100644 --- a/common/misc.js +++ b/common/misc.js @@ -1,5 +1,7 @@ /** - * Hachette miscellaneous operations refactored to a separate file + * This file is part of Haketilo. + * + * Function: Miscellaneous operations refactored to a separate file. * * Copyright (C) 2021 Wojtek Kosior * Copyright (C) 2021 jahoti diff --git a/common/observable.js b/common/observable.js index 02f1c1b..ab3b444 100644 --- a/common/observable.js +++ b/common/observable.js @@ -1,6 +1,7 @@ /** - * part of Hachette - * Facilitate listening to events + * This file is part of Haketilo. + * + * Function: Facilitate listening to (internal, self-generated) events. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/common/once.js b/common/once.js index 098b43f..93e842f 100644 --- a/common/once.js +++ b/common/once.js @@ -1,5 +1,8 @@ /** - * Hachette feature initialization promise + * This file is part of Haketilo. + * + * Function: Wrap APIs that depend on some asynchronous initialization into + * promises. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/common/patterns.js b/common/patterns.js index ae29fcd..625be05 100644 --- a/common/patterns.js +++ b/common/patterns.js @@ -1,5 +1,7 @@ /** - * Hachette operations on page url patterns + * This file is part of Haketilo. + * + * Function: Operations on page URL patterns. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/common/sanitize_JSON.js b/common/sanitize_JSON.js index 8b86d2d..4cf1ef4 100644 --- a/common/sanitize_JSON.js +++ b/common/sanitize_JSON.js @@ -1,6 +1,7 @@ /** - * part of Hachette - * Powerful, full-blown format enforcer for externally-obtained JSON + * This file is part of Haketilo. + * + * Function: Powerful, full-blown format enforcer for externally-obtained JSON. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/common/settings_query.js b/common/settings_query.js index b54e580..7e1315e 100644 --- a/common/settings_query.js +++ b/common/settings_query.js @@ -1,5 +1,7 @@ /** - * Hachette querying page settings with regard to wildcard records + * This file is part of Haketilo. + * + * Function: Querying page settings. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/common/signing.js b/common/signing.js index 1904bcd..11cd442 100644 --- a/common/signing.js +++ b/common/signing.js @@ -1,6 +1,7 @@ /** - * part of Hachette - * Functions related to "signing" of data, refactored to a separate file. + * This file is part of Haketilo. + * + * Functions: Operations related to "signing" of data. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. @@ -16,7 +17,7 @@ /* * In order to make certain data synchronously accessible in certain contexts, - * hachette smuggles it in string form in places like cookies, URLs and headers. + * Haketilo smuggles it in string form in places like cookies, URLs and headers. * When using the smuggled data, we first need to make sure it isn't spoofed. * For that, we use this pseudo-signing mechanism. * diff --git a/common/storage_client.js b/common/storage_client.js index 2b2f495..ef4a0b8 100644 --- a/common/storage_client.js +++ b/common/storage_client.js @@ -1,5 +1,7 @@ /** - * Hachette storage through connection (client side) + * This file is part of Haketilo. + * + * Function: Storage through messages (client side). * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/common/storage_light.js b/common/storage_light.js index 067bf0c..32e3b1f 100644 --- a/common/storage_light.js +++ b/common/storage_light.js @@ -1,6 +1,7 @@ /** - * part of Hachette - * Storage manager, lighter than the previous one. + * This file is part of Haketilo. + * + * Function: Storage manager, lighter than the previous one. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/common/storage_raw.js b/common/storage_raw.js index 4c02ee4..e354b6b 100644 --- a/common/storage_raw.js +++ b/common/storage_raw.js @@ -1,6 +1,7 @@ /** - * part of Hachette - * Basic wrappers for storage API functions. + * This file is part of Haketilo. + * + * Function: Basic wrappers for storage API functions. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/common/stored_types.js b/common/stored_types.js index bfceba6..a693b1c 100644 --- a/common/stored_types.js +++ b/common/stored_types.js @@ -1,5 +1,7 @@ /** - * Hachette stored item types "enum" + * This file is part of Haketilo. + * + * Function: Define an "enum" of stored item types. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/content/activity_info_server.js b/content/activity_info_server.js index 1b69703..d1dfe36 100644 --- a/content/activity_info_server.js +++ b/content/activity_info_server.js @@ -1,7 +1,8 @@ /** - * part of Hachette - * Informing about activities performed by content script (script injection, - * script blocking). + * This file is part of Haketilo. + * + * Function: Informing the popup about what happens in the content script + * (script injection, script blocking, etc.). * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/content/main.js b/content/main.js index 6478ea0..cec9943 100644 --- a/content/main.js +++ b/content/main.js @@ -1,5 +1,7 @@ /** - * Hachette main content script run in all frames + * This file is part of Haketilo. + * + * Function: Main content script that runs in all frames. * * Copyright (C) 2021 Wojtek Kosior * Copyright (C) 2021 jahoti @@ -33,7 +35,7 @@ function extract_cookie_policy(cookie, min_time) let policy = null; const extracted_signatures = []; - for (const match of cookie.matchAll(/hachette-(\w*)=([^;]*)/g)) { + for (const match of cookie.matchAll(/haketilo-(\w*)=([^;]*)/g)) { const new_result = extract_signed(...match.slice(1, 3)); if (new_result.fail) continue; @@ -60,7 +62,7 @@ function extract_url_policy(url, min_time) const [base_url, payload, anchor] = /^([^#]*)#?([^#]*)(#?.*)$/.exec(url).splice(1, 4); - const match = /^hachette_([^_]+)_(.*)$/.exec(payload); + const match = /^haketilo_([^_]+)_(.*)$/.exec(payload); if (!match) return [null, url]; @@ -83,7 +85,7 @@ function employ_nonhttp_policy(policy) policy.nonce = gen_nonce(); const [base_url, target] = /^([^#]*)(#?.*)$/.exec(policy.url).slice(1, 3); const encoded_policy = encodeURIComponent(JSON.stringify(policy)); - const payload = "hachette_" + + const payload = "haketilo_" + sign_data(encoded_policy, new Date().getTime()).join("_"); const resulting_url = `${base_url}#${payload}${target}`; location.href = resulting_url; @@ -187,7 +189,7 @@ function sanitize_meta(meta) function sanitize_script(script) { - script.hachette_blocked_type = script.getAttribute("type"); + script.haketilo_blocked_type = script.getAttribute("type"); script.type = "text/plain"; } @@ -197,12 +199,12 @@ function sanitize_script(script) */ function desanitize_script(script) { - script.setAttribute("type", script.hachette_blocked_type); + script.setAttribute("type", script.haketilo_blocked_type); - if ([null, undefined].includes(script.hachette_blocked_type)) + if ([null, undefined].includes(script.haketilo_blocked_type)) script.removeAttribute("type"); - delete script.hachette_blocked_type; + delete script.haketilo_blocked_type; } const bad_url_reg = /^data:([^,;]*ml|unknown-content-type)/i; @@ -235,7 +237,7 @@ function start_data_urls_sanitizing(doc) */ function prevent_script_execution(event) { - if (!event.target._hachette_payload) + if (!event.target.haketilo_payload) event.preventDefault(); } @@ -336,7 +338,7 @@ if (!is_privileged_url(document.URL)) { let signatures; [policy, signatures] = extract_cookie_policy(document.cookie, min_time); for (const signature of signatures) - document.cookie = `hachette-${signature}=; Max-Age=-1;`; + document.cookie = `haketilo-${signature}=; Max-Age=-1;`; } else { const scheme = /^([^:]*)/.exec(document.URL)[1]; const known_scheme = ["file", "ftp"].includes(scheme); diff --git a/content/page_actions.js b/content/page_actions.js index 040b4ab..db7c352 100644 --- a/content/page_actions.js +++ b/content/page_actions.js @@ -1,5 +1,7 @@ /** - * Hachette handling of page actions in content scripts + * This file is part of Haketilo. + * + * Function: Handle page actions in a content script. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. @@ -60,7 +62,7 @@ function add_script(script_text) let script = document.createElement("script"); script.textContent = script_text; script.setAttribute("nonce", nonce); - script._hachette_payload = true; + script.haketilo_payload = true; document.body.appendChild(script); report_script(script_text); diff --git a/content/repo_query.js b/content/repo_query.js index 3708108..637282c 100644 --- a/content/repo_query.js +++ b/content/repo_query.js @@ -1,6 +1,7 @@ /** - * part of Hachette - * Getting available content for site from remote repositories. + * This file is part of Haketilo. + * + * Function: Getting available content for site from remote repositories. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/copyright b/copyright index 4c37eb3..fe2aed7 100644 --- a/copyright +++ b/copyright @@ -1,5 +1,5 @@ Format: https://www.debian.org/doc/packaging-manuals/copyright-format/1.0/ -Upstream-Name: Hachette +Upstream-Name: Haketilo Source: https://git.koszko.org/browser-extension/ Files: * diff --git a/html/DOM_helpers.js b/html/DOM_helpers.js index 01e2be9..4fe118d 100644 --- a/html/DOM_helpers.js +++ b/html/DOM_helpers.js @@ -1,5 +1,7 @@ /** - * Hachette operations on DOM elements + * This file is part of Haketilo. + * + * Function: Operations on DOM elements. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/html/MOZILLA_scrollbar_fix.css b/html/MOZILLA_scrollbar_fix.css index 5feb7c3..cdbd5c6 100644 --- a/html/MOZILLA_scrollbar_fix.css +++ b/html/MOZILLA_scrollbar_fix.css @@ -1,6 +1,8 @@ /** - * Hachette - * Hacky fix for vertical scrollbar width being included in child's width. + * This file is part of Haketilo. + * + * Function: Hacky fix for vertical scrollbar width being included in child's + * width. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/html/back_button.css b/html/back_button.css index 1ddc5da..b83e834 100644 --- a/html/back_button.css +++ b/html/back_button.css @@ -1,6 +1,7 @@ /** - * part of Hachette - * Style for a "back" button with a CSS arrow image. + * This file is part of Haketilo. + * + * Function: Style for a "back" button with a CSS arrow image. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/html/base.css b/html/base.css index df52f7c..517a5c1 100644 --- a/html/base.css +++ b/html/base.css @@ -1,5 +1,7 @@ /** - * Hachette base styles + * This file is part of Haketilo. + * + * Function: Base styles. * * Copyright (C) 2021 Wojtek Kosior * Copyright (C) 2021 Nicholas Johnson diff --git a/html/default_blocking_policy.js b/html/default_blocking_policy.js index 2f49bac..b6458f3 100644 --- a/html/default_blocking_policy.js +++ b/html/default_blocking_policy.js @@ -1,6 +1,7 @@ /** - * part of Hachette - * Default policy dialog logic. + * This file is part of Haketilo. + * + * Function: Logic for the dialog of default policy selection. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/html/display-panel.html b/html/display-panel.html index 3ed1b7a..ee9e767 100644 --- a/html/display-panel.html +++ b/html/display-panel.html @@ -1,12 +1,16 @@ - Hachette - page settings + Haketilo - page settings @@ -331,7 +335,7 @@ diff --git a/html/display-panel.js b/html/display-panel.js index 84c922f..c078850 100644 --- a/html/display-panel.js +++ b/html/display-panel.js @@ -1,5 +1,7 @@ /** - * Hachette display panel logic + * This file is part of Haketilo. + * + * Function: Popup logic. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/html/import_frame.js b/html/import_frame.js index c0eb2f0..ae6fab4 100644 --- a/html/import_frame.js +++ b/html/import_frame.js @@ -1,5 +1,7 @@ /** - * Hachette HTML import frame script + * This file is part of Haketilo. + * + * Function: Logic for the settings import frame. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/html/options.html b/html/options.html index 54ab9e8..2e8317c 100644 --- a/html/options.html +++ b/html/options.html @@ -1,12 +1,16 @@ - Hachette options + Haketilo options diff --git a/html/options_main.js b/html/options_main.js index 27ab0ec..f8faf9b 100644 --- a/html/options_main.js +++ b/html/options_main.js @@ -1,5 +1,7 @@ /** - * Hachette HTML options page main script + * This file is part of Haketilo. + * + * Function: Settings page logic. * * Copyright (C) 2021 Wojtek Kosior * Redistribution terms are gathered in the `copyright' file. diff --git a/icons/hachette.svg b/icons/hachette.svg deleted file mode 100644 index 6e8948d..0000000 --- a/icons/hachette.svg +++ /dev/null @@ -1,127 +0,0 @@ - - - Hatchet - - - - - - - - - - - - image/svg+xml - - Hatchet - - - David Lyons - - - - - dlyons - - - - 2017-05 - - - hatchet - ax - wood - - - Hatchet - - - - - - - - - - - - - - - - - - - - - - diff --git a/icons/hachette128.png b/icons/hachette128.png deleted file mode 100644 index 18816e9..0000000 Binary files a/icons/hachette128.png and /dev/null differ diff --git a/icons/hachette16.png b/icons/hachette16.png deleted file mode 100644 index 182ede5..0000000 Binary files a/icons/hachette16.png and /dev/null differ diff --git a/icons/hachette32.png b/icons/hachette32.png deleted file mode 100644 index ffaa84b..0000000 Binary files a/icons/hachette32.png and /dev/null differ diff --git a/icons/hachette48.png b/icons/hachette48.png deleted file mode 100644 index 1ffcd38..0000000 Binary files a/icons/hachette48.png and /dev/null differ diff --git a/icons/hachette64.png b/icons/hachette64.png deleted file mode 100644 index a02abb0..0000000 Binary files a/icons/hachette64.png and /dev/null differ diff --git a/icons/haketilo.svg b/icons/haketilo.svg new file mode 100644 index 0000000..6e8948d --- /dev/null +++ b/icons/haketilo.svg @@ -0,0 +1,127 @@ + + + Hatchet + + + + + + + + + + + + image/svg+xml + + Hatchet + + + David Lyons + + + + + dlyons + + + + 2017-05 + + + hatchet + ax + wood + + + Hatchet + + + + + + + + + + + + + + + + + + + + + + diff --git a/icons/haketilo128.png b/icons/haketilo128.png new file mode 100644 index 0000000..18816e9 Binary files /dev/null and b/icons/haketilo128.png differ diff --git a/icons/haketilo16.png b/icons/haketilo16.png new file mode 100644 index 0000000..182ede5 Binary files /dev/null and b/icons/haketilo16.png differ diff --git a/icons/haketilo32.png b/icons/haketilo32.png new file mode 100644 index 0000000..ffaa84b Binary files /dev/null and b/icons/haketilo32.png differ diff --git a/icons/haketilo48.png b/icons/haketilo48.png new file mode 100644 index 0000000..1ffcd38 Binary files /dev/null and b/icons/haketilo48.png differ diff --git a/icons/haketilo64.png b/icons/haketilo64.png new file mode 100644 index 0000000..a02abb0 Binary files /dev/null and b/icons/haketilo64.png differ diff --git a/manifest.json b/manifest.json index ce2577e..9d34732 100644 --- a/manifest.json +++ b/manifest.json @@ -1,18 +1,20 @@ +// This is the manifest file of Haketilo. +// // Copyright (C) 2021 Wojtek Kosior // Redistribution terms are gathered in the `copyright' file. { "manifest_version": 2, - "name": "Hachette", - "short_name": "Hachette", + "name": "Haketilo", + "short_name": "Haketilo", "version": "0.0.1", "author": "various", "description": "Control your \"Web\" browsing.",_GECKO_APPLICATIONS_ "icons":{ - "128": "icons/hachette128.png", - "64": "icons/hachette64.png", - "48": "icons/hachette48.png", - "32": "icons/hachette32.png", - "16": "icons/hachette16.png" + "128": "icons/haketilo128.png", + "64": "icons/haketilo64.png", + "48": "icons/haketilo48.png", + "32": "icons/haketilo32.png", + "16": "icons/haketilo16.png" }, "permissions": [ "contextMenus", @@ -29,13 +31,13 @@ "browser_action": { "browser_style": true, "default_icon": { - "128": "icons/hachette128.png", - "64": "icons/hachette64.png", - "48": "icons/hachette48.png", - "32": "icons/hachette32.png", - "16": "icons/hachette16.png" + "128": "icons/haketilo128.png", + "64": "icons/haketilo64.png", + "48": "icons/haketilo48.png", + "32": "icons/haketilo32.png", + "16": "icons/haketilo16.png" }, - "default_title": "Hachette", + "default_title": "Haketilo", "default_popup": "html/display-panel.html" }, "options_ui": { diff --git a/re-generate_icons.sh b/re-generate_icons.sh index ba0c28a..e557ad0 100755 --- a/re-generate_icons.sh +++ b/re-generate_icons.sh @@ -4,5 +4,5 @@ # Redistribution terms are gathered in the `copyright' file. for SIZE in 128 64 48 32 16; do - inkscape -z -e icons/hachette$SIZE.png -w $SIZE -h $SIZE icons/hachette.svg + inkscape -z -e icons/haketilo$SIZE.png -w $SIZE -h $SIZE icons/haketilo.svg done -- cgit v1.2.3 From 96068ada37bfa1d7e6485551138ba36600664caf Mon Sep 17 00:00:00 2001 From: Wojtek Kosior Date: Sat, 20 Nov 2021 18:29:59 +0100 Subject: replace cookies with synchronous XmlHttpRequest as policy smuggling method. Note: this breaks Mozilla port of Haketilo. Synchronous XmlHttpRequest doesn't work as well there. This will be fixed with dynamically-registered content scripts later. --- background/cookie_filter.js | 46 ------------- background/main.js | 120 +++++++++++++++++++++++++++------ background/page_actions_server.js | 32 ++------- background/policy_injector.js | 67 +++--------------- background/stream_filter.js | 6 +- build.sh | 16 +---- common/misc.js | 2 +- common/signing.js | 74 -------------------- content/activity_info_server.js | 4 +- content/main.js | 138 +++++++++++--------------------------- content/page_actions.js | 27 ++++---- dummy | 0 html/display-panel.js | 13 ++-- manifest.json | 3 +- 14 files changed, 180 insertions(+), 368 deletions(-) delete mode 100644 background/cookie_filter.js delete mode 100644 common/signing.js create mode 100644 dummy (limited to 'background/stream_filter.js') diff --git a/background/cookie_filter.js b/background/cookie_filter.js deleted file mode 100644 index 64d18b2..0000000 --- a/background/cookie_filter.js +++ /dev/null @@ -1,46 +0,0 @@ -/** - * This file is part of Haketilo. - * - * Function: Filtering request headers to remove haketilo cookies that might - * have slipped through. - * - * Copyright (C) 2021 Wojtek Kosior - * Redistribution terms are gathered in the `copyright' file. - */ - -/* - * IMPORTS_START - * IMPORT extract_signed - * IMPORTS_END - */ - -function is_valid_haketilo_cookie(cookie) -{ - const match = /^haketilo-(\w*)=(.*)$/.exec(cookie); - if (!match) - return false; - - return !extract_signed(match.slice(1, 3)).fail; -} - -function remove_haketilo_cookies(header) -{ - if (header.name !== "Cookie") - return header; - - const cookies = header.value.split("; "); - const value = cookies.filter(c => !is_valid_haketilo_cookie(c)).join("; "); - - return value ? {name: "Cookie", value} : null; -} - -function filter_cookie_headers(headers) -{ - return headers.map(remove_haketilo_cookies).filter(h => h); -} - -/* - * EXPORTS_START - * EXPORT filter_cookie_headers - * EXPORTS_END - */ diff --git a/background/main.js b/background/main.js index 40b3a9e..9cdfb97 100644 --- a/background/main.js +++ b/background/main.js @@ -17,11 +17,10 @@ * IMPORT browser * IMPORT is_privileged_url * IMPORT query_best - * IMPORT gen_nonce * IMPORT inject_csp_headers * IMPORT apply_stream_filter - * IMPORT filter_cookie_headers * IMPORT is_chrome + * IMPORT is_mozilla * IMPORTS_END */ @@ -51,34 +50,53 @@ async function init_ext(install_details) browser.runtime.onInstalled.addListener(init_ext); +/* + * The function below implements a more practical interface for what it does by + * wrapping the old query_best() function. + */ +function decide_policy_for_url(storage, policy_observable, url) +{ + if (storage === undefined) + return {allow: false}; + + const settings = + {allow: policy_observable !== undefined && policy_observable.value}; + + const [pattern, queried_settings] = query_best(storage, url); + + if (queried_settings) { + settings.payload = queried_settings.components; + settings.allow = !!queried_settings.allow && !settings.payload; + settings.pattern = pattern; + } + + return settings; +} let storage; let policy_observable = {}; -function on_headers_received(details) +function sanitize_web_page(details) { const url = details.url; if (is_privileged_url(details.url)) return; - const [pattern, settings] = query_best(storage, details.url); - const has_payload = !!(settings && settings.components); - const allow = !has_payload && - !!(settings ? settings.allow : policy_observable.value); - const nonce = gen_nonce(); - const policy = {allow, url, nonce, has_payload}; + const policy = + decide_policy_for_url(storage, policy_observable, details.url); let headers = details.responseHeaders; + + headers = inject_csp_headers(headers, policy); + let skip = false; for (const header of headers) { if ((header.name.toLowerCase().trim() === "content-disposition" && /^\s*attachment\s*(;.*)$/i.test(header.value))) skip = true; } - - headers = inject_csp_headers(headers, policy); - skip = skip || (details.statusCode >= 300 && details.statusCode < 400); + if (!skip) { /* Check for API availability. */ if (browser.webRequest.filterResponseData) @@ -88,11 +106,49 @@ function on_headers_received(details) return {responseHeaders: headers}; } -function on_before_send_headers(details) +const request_url_regex = /^[^?]*\?url=(.*)$/; +const redirect_url_template = browser.runtime.getURL("dummy") + "?settings="; + +function synchronously_smuggle_policy(details) { - let headers = details.requestHeaders; - headers = filter_cookie_headers(headers); - return {requestHeaders: headers}; + /* + * Content script will make a synchronous XmlHttpRequest to extension's + * `dummy` file to query settings for given URL. We smuggle that + * information in query parameter of the URL we redirect to. + * A risk of fingerprinting arises if a page with script execution allowed + * guesses the dummy file URL and makes an AJAX call to it. It is currently + * a problem in ManifestV2 Chromium-family port of Haketilo because Chromium + * uses predictable URLs for web-accessible resources. We plan to fix it in + * the future ManifestV3 port. + */ + if (details.type !== "xmlhttprequest") + return {cancel: true}; + + console.debug(`Settings queried using XHR for '${details.url}'.`); + + let policy = {allow: false}; + + try { + /* + * request_url should be of the following format: + * ?url= + */ + const match = request_url_regex.exec(details.url); + const queried_url = decodeURIComponent(match[1]); + + if (details.initiator && !queried_url.startsWith(details.initiator)) { + console.warn(`Blocked suspicious query of '${url}' by '${details.initiator}'. This might be the result of page fingerprinting the browser.`); + return {cancel: true}; + } + + policy = decide_policy_for_url(storage, policy_observable, queried_url); + } catch (e) { + console.warn(`Bad request! Expected ${browser.runtime.getURL("dummy")}?url=. Got ${request_url}. This might be the result of page fingerprinting the browser.`); + } + + const encoded_policy = encodeURIComponent(JSON.stringify(policy)); + + return {redirectUrl: redirect_url_template + encoded_policy}; } const all_types = [ @@ -110,18 +166,40 @@ async function start_webRequest_operations() extra_opts.push("extraHeaders"); browser.webRequest.onHeadersReceived.addListener( - on_headers_received, + sanitize_web_page, {urls: [""], types: ["main_frame", "sub_frame"]}, extra_opts.concat("responseHeaders") ); - browser.webRequest.onBeforeSendHeaders.addListener( - on_before_send_headers, - {urls: [""], types: all_types}, - extra_opts.concat("requestHeaders") + const dummy_url_pattern = browser.runtime.getURL("dummy") + "?url=*"; + browser.webRequest.onBeforeRequest.addListener( + synchronously_smuggle_policy, + {urls: [dummy_url_pattern], types: ["xmlhttprequest"]}, + extra_opts ); policy_observable = await light_storage.observe_var("default_allow"); } start_webRequest_operations(); + +const code = `\ +console.warn("Hi, I'm Mr Dynamic!"); + +console.debug("let's see how window.killtheweb looks like now"); + +console.log("killtheweb", window.killtheweb); +` + +async function test_dynamic_content_scripts() +{ + browser.contentScripts.register({ + "js": [{code}], + "matches": [""], + "allFrames": true, + "runAt": "document_start" +}); +} + +if (is_mozilla) + test_dynamic_content_scripts(); diff --git a/background/page_actions_server.js b/background/page_actions_server.js index 156a79f..74783c9 100644 --- a/background/page_actions_server.js +++ b/background/page_actions_server.js @@ -16,34 +16,12 @@ * IMPORT browser * IMPORT listen_for_connection * IMPORT sha256 - * IMPORT query_best * IMPORT make_ajax_request * IMPORTS_END */ var storage; var handler; -let policy_observable; - -function send_actions(url, port) -{ - const [pattern, queried_settings] = query_best(storage, url); - - const settings = {allow: policy_observable && policy_observable.value}; - Object.assign(settings, queried_settings); - if (settings.components) - settings.allow = false; - - const repos = storage.get_all(TYPE_PREFIX.REPO); - - port.postMessage(["settings", [pattern, settings, repos]]); - - const components = settings.components; - const processed_bags = new Set(); - - if (components !== undefined) - send_scripts([components], port, processed_bags); -} // TODO: parallelize script fetching async function send_scripts(components, port, processed_bags) @@ -116,9 +94,11 @@ async function fetch_remote_script(script_data) function handle_message(port, message, handler) { port.onMessage.removeListener(handler[0]); - let url = message.url; - console.log({url}); - send_actions(url, port); + console.debug(`Loading payload '${message.payload}'.`); + + const processed_bags = new Set(); + + send_scripts([message.payload], port, processed_bags); } function new_connection(port) @@ -134,8 +114,6 @@ async function start_page_actions_server() storage = await get_storage(); listen_for_connection(CONNECTION_TYPE.PAGE_ACTIONS, new_connection); - - policy_observable = await light_storage.observe_var("default_allow"); } /* diff --git a/background/policy_injector.js b/background/policy_injector.js index 881595b..b49ec47 100644 --- a/background/policy_injector.js +++ b/background/policy_injector.js @@ -10,77 +10,28 @@ /* * IMPORTS_START - * IMPORT sign_data - * IMPORT extract_signed * IMPORT make_csp_rule * IMPORT csp_header_regex + * Re-enable the import below once nonce stuff here is ready + * !mport gen_nonce * IMPORTS_END */ function inject_csp_headers(headers, policy) { let csp_headers; - let old_signature; - let haketilo_header; - for (const header of headers.filter(h => h.name === "x-haketilo")) { - /* x-haketilo header has format: _0_ */ - const match = /^([^_]+)_(0_.*)$/.exec(header.value); - if (!match) - continue; + if (policy.payload) { + headers = headers.filter(h => !csp_header_regex.test(h.name)); - const result = extract_signed(...match.slice(1, 3)); - if (result.fail) - continue; + // TODO: make CSP rules with nonces and facilitate passing them to + // content scripts via dynamic content script registration or + // synchronous XHRs - /* This should succeed - it's our self-produced valid JSON. */ - const old_data = JSON.parse(decodeURIComponent(result.data)); - - /* Confirmed- it's the originals, smuggled in! */ - csp_headers = old_data.csp_headers; - old_signature = old_data.policy_sig; - - haketilo_header = header; - break; + // policy.nonce = gen_nonce(); } - if (policy.has_payload) { - csp_headers = []; - const non_csp_headers = []; - const header_list = - h => csp_header_regex.test(h) ? csp_headers : non_csp_headers; - headers.forEach(h => header_list(h.name).push(h)); - headers = non_csp_headers; - } else { - headers.push(...csp_headers || []); - } - - if (!haketilo_header) { - haketilo_header = {name: "x-haketilo"}; - headers.push(haketilo_header); - } - - if (old_signature) - headers = headers.filter(h => h.value.search(old_signature) === -1); - - const policy_str = encodeURIComponent(JSON.stringify(policy)); - const signed_policy = sign_data(policy_str, new Date().getTime()); - const later_30sec = new Date(new Date().getTime() + 30000).toGMTString(); - headers.push({ - name: "Set-Cookie", - value: `haketilo-${signed_policy.join("=")}; Expires=${later_30sec};` - }); - - /* - * Smuggle in the signature and the original CSP headers for future use. - * These are signed with a time of 0, as it's not clear there is a limit on - * how long Firefox might retain headers in the cache. - */ - let haketilo_data = {csp_headers, policy_sig: signed_policy[0]}; - haketilo_data = encodeURIComponent(JSON.stringify(haketilo_data)); - haketilo_header.value = sign_data(haketilo_data, 0).join("_"); - - if (!policy.allow) { + if (!policy.allow && (policy.nonce || !policy.payload)) { headers.push({ name: "content-security-policy", value: make_csp_rule(policy) diff --git a/background/stream_filter.js b/background/stream_filter.js index e5e0827..e5d124c 100644 --- a/background/stream_filter.js +++ b/background/stream_filter.js @@ -174,8 +174,7 @@ function filter_data(properties, event) * as harmless anyway). */ - const dummy_script = - ``; + const dummy_script = ``; const doctype_decl = /^(\s*"']*>)?/i.exec(decoded)[0]; decoded = doctype_decl + dummy_script + decoded.substring(doctype_decl.length); @@ -189,11 +188,10 @@ function filter_data(properties, event) function apply_stream_filter(details, headers, policy) { - if (!policy.has_payload) + if (!policy.payload) return headers; const properties = properties_from_headers(headers); - properties.policy = policy; properties.filter = browser.webRequest.filterResponseData(details.requestId); diff --git a/build.sh b/build.sh index 936ab06..ed6a141 100755 --- a/build.sh +++ b/build.sh @@ -180,7 +180,6 @@ build_main() { mkdir -p "$BUILDDIR"/$DIR done - CHROMIUM_KEY='' CHROMIUM_UPDATE_URL='' GECKO_APPLICATIONS='' @@ -189,20 +188,7 @@ build_main() { fi if [ "$BROWSER" = "chromium" ]; then - CHROMIUM_KEY="$(dd if=/dev/urandom bs=32 count=1 2>/dev/null | base64)" - CHROMIUM_KEY=$(echo chromium-key-dummy-file-$CHROMIUM_KEY | tr / -) - touch "$BUILDDIR"/$CHROMIUM_KEY - CHROMIUM_UPDATE_URL="$UPDATE_URL" - - CHROMIUM_KEY="\n\ - // WARNING!!!\n\ - // EACH USER SHOULD REPLACE DUMMY FILE's VALUE WITH A UNIQUE ONE!!!\n\ - // OTHERWISE, SECURITY CAN BE TRIVIALLY COMPROMISED!\n\ - // Only relevant to users of chrome-based browsers.\n\ - // Users of Firefox forks are safe.\n\ - \"$CHROMIUM_KEY\"\ -" else GECKO_APPLICATIONS="\n\ \"applications\": {\n\ @@ -215,7 +201,6 @@ build_main() { sed "\ s^_GECKO_APPLICATIONS_^$GECKO_APPLICATIONS^ -s^_CHROMIUM_KEY_^$CHROMIUM_KEY^ s^_CHROMIUM_UPDATE_URL_^$CHROMIUM_UPDATE_URL^ s^_BGSCRIPTS_^$BGSCRIPTS^ s^_CONTENTSCRIPTS_^$CONTENTSCRIPTS^" \ @@ -279,6 +264,7 @@ EOF fi cp -r copyright licenses/ "$BUILDDIR" + cp dummy "$BUILDDIR" cp html/*.css "$BUILDDIR"/html mkdir "$BUILDDIR"/icons cp icons/*.png "$BUILDDIR"/icons diff --git a/common/misc.js b/common/misc.js index 9ffb7ff..5b0addb 100644 --- a/common/misc.js +++ b/common/misc.js @@ -49,7 +49,7 @@ function gen_nonce(length=16) function make_csp_rule(policy) { let rule = "prefetch-src 'none'; script-src-attr 'none';"; - const script_src = policy.has_payload ? + const script_src = policy.nonce !== undefined ? `'nonce-${policy.nonce}'` : "'none'"; rule += ` script-src ${script_src}; script-src-elem ${script_src};`; return rule; diff --git a/common/signing.js b/common/signing.js deleted file mode 100644 index 11cd442..0000000 --- a/common/signing.js +++ /dev/null @@ -1,74 +0,0 @@ -/** - * This file is part of Haketilo. - * - * Functions: Operations related to "signing" of data. - * - * Copyright (C) 2021 Wojtek Kosior - * Redistribution terms are gathered in the `copyright' file. - */ - -/* - * IMPORTS_START - * IMPORT sha256 - * IMPORT browser - * IMPORT is_mozilla - * IMPORTS_END - */ - -/* - * In order to make certain data synchronously accessible in certain contexts, - * Haketilo smuggles it in string form in places like cookies, URLs and headers. - * When using the smuggled data, we first need to make sure it isn't spoofed. - * For that, we use this pseudo-signing mechanism. - * - * Despite what name suggests, no assymetric cryptography is involved, as it - * would bring no additional benefits and would incur bigger performance - * overhead. Instead, we hash the string data together with some secret value - * that is supposed to be known only by this browser instance. Resulting hash - * sum plays the role of the signature. In the hash we also include current - * time. This way, even if signed data leaks (which shouldn't happen in the - * first place), an attacker won't be able to re-use it indefinitely. - * - * The secret shared between execution contexts has to be available - * synchronously. Under Mozilla, this is the extension's per-session id. Under - * Chromium, this is a dummy web-accessible-resource name that resides in the - * manifest and is supposed to be constructed by each user using a unique value - * (this is done automatically by `build.sh'). - */ - -function get_secret() -{ - if (is_mozilla) - return browser.runtime.getURL("dummy"); - - return chrome.runtime.getManifest().web_accessible_resources - .map(r => /^chromium-key-dummy-file-(.*)/.exec(r)).filter(r => r)[0][1]; -} - -function extract_signed(signature, signed_data) -{ - const match = /^([1-9][0-9]{12}|0)_(.*)$/.exec(signed_data); - if (!match) - return {fail: "bad format"}; - - const result = {time: parseInt(match[1]), data: match[2]}; - if (sign_data(result.data, result.time)[0] !== signature) - result.fail = "bad signature"; - - return result; -} - -/* - * Sign a given string for a given time. Time should be either 0 or in the range - * 10^12 <= time < 10^13. - */ -function sign_data(data, time) { - return [sha256(get_secret() + time + data), `${time}_${data}`]; -} - -/* - * EXPORTS_START - * EXPORT extract_signed - * EXPORT sign_data - * EXPORTS_END - */ diff --git a/content/activity_info_server.js b/content/activity_info_server.js index d1dfe36..aa92b75 100644 --- a/content/activity_info_server.js +++ b/content/activity_info_server.js @@ -42,7 +42,9 @@ function report_script(script_data) function report_settings(settings) { - report_activity("settings", settings); + const settings_clone = {}; + Object.assign(settings_clone, settings) + report_activity("settings", settings_clone); } function report_document_type(is_html) diff --git a/content/main.js b/content/main.js index cec9943..ce1ff7a 100644 --- a/content/main.js +++ b/content/main.js @@ -11,15 +11,15 @@ /* * IMPORTS_START * IMPORT handle_page_actions - * IMPORT extract_signed - * IMPORT sign_data * IMPORT gen_nonce * IMPORT is_privileged_url + * IMPORT browser * IMPORT is_chrome * IMPORT is_mozilla * IMPORT start_activity_info_server * IMPORT make_csp_rule * IMPORT csp_header_regex + * IMPORT report_settings * IMPORTS_END */ @@ -29,69 +29,6 @@ const wait_loaded = e => e.content_loaded ? Promise.resolve() : wait_loaded(document).then(() => document.content_loaded = true); -function extract_cookie_policy(cookie, min_time) -{ - let best_result = {time: -1}; - let policy = null; - const extracted_signatures = []; - - for (const match of cookie.matchAll(/haketilo-(\w*)=([^;]*)/g)) { - const new_result = extract_signed(...match.slice(1, 3)); - if (new_result.fail) - continue; - - extracted_signatures.push(match[1]); - - if (new_result.time < Math.max(min_time, best_result.time)) - continue; - - /* This should succeed - it's our self-produced valid JSON. */ - const new_policy = JSON.parse(decodeURIComponent(new_result.data)); - if (new_policy.url !== document.URL) - continue; - - best_result = new_result; - policy = new_policy; - } - - return [policy, extracted_signatures]; -} - -function extract_url_policy(url, min_time) -{ - const [base_url, payload, anchor] = - /^([^#]*)#?([^#]*)(#?.*)$/.exec(url).splice(1, 4); - - const match = /^haketilo_([^_]+)_(.*)$/.exec(payload); - if (!match) - return [null, url]; - - const result = extract_signed(...match.slice(1, 3)); - if (result.fail) - return [null, url]; - - const original_url = base_url + anchor; - const policy = result.time < min_time ? null : - JSON.parse(decodeURIComponent(result.data)); - - return [policy.url === original_url ? policy : null, original_url]; -} - -function employ_nonhttp_policy(policy) -{ - if (!policy.allow) - return; - - policy.nonce = gen_nonce(); - const [base_url, target] = /^([^#]*)(#?.*)$/.exec(policy.url).slice(1, 3); - const encoded_policy = encodeURIComponent(JSON.stringify(policy)); - const payload = "haketilo_" + - sign_data(encoded_policy, new Date().getTime()).join("_"); - const resulting_url = `${base_url}#${payload}${target}`; - location.href = resulting_url; - location.reload(); -} - /* * In the case of HTML documents: * 1. When injecting some payload we need to sanitize CSP tags before @@ -306,7 +243,7 @@ http-equiv="Content-Security-Policy" content="${make_csp_rule(policy)}"\ start_data_urls_sanitizing(doc); } -async function disable_service_workers() +async function _disable_service_workers() { if (!navigator.serviceWorker) return; @@ -315,7 +252,7 @@ async function disable_service_workers() if (registrations.length === 0) return; - console.warn("Service Workers detected on this page! Unregistering and reloading"); + console.warn("Service Workers detected on this page! Unregistering and reloading."); try { await Promise.all(registrations.map(r => r.unregister())); @@ -327,50 +264,57 @@ async function disable_service_workers() return new Promise(() => 0); } -if (!is_privileged_url(document.URL)) { - let policy_received_callback = () => undefined; - let policy; - - /* Signature valid for half an hour. */ - const min_time = new Date().getTime() - 1800 * 1000; - - if (/^https?:/.test(document.URL)) { - let signatures; - [policy, signatures] = extract_cookie_policy(document.cookie, min_time); - for (const signature of signatures) - document.cookie = `haketilo-${signature}=; Max-Age=-1;`; - } else { - const scheme = /^([^:]*)/.exec(document.URL)[1]; - const known_scheme = ["file", "ftp"].includes(scheme); - - if (!known_scheme) - console.warn(`Unknown url scheme: \`${scheme}'!`); - - let original_url; - [policy, original_url] = extract_url_policy(document.URL, min_time); - history.replaceState(null, "", original_url); - - if (known_scheme && !policy) - policy_received_callback = employ_nonhttp_policy; +/* + * Trying to use servce workers APIs might result in exceptions, for example + * when in a non-HTML document. Because of this, we wrap the function that does + * the actual work in a try {} block. + */ +async function disable_service_workers() +{ + try { + await _disable_service_workers() + } catch (e) { + console.debug("Exception thrown during an attempt to detect and disable service workers.", e); } +} - if (!policy) { - console.debug("Using fallback policy!"); - policy = {allow: false, nonce: gen_nonce()}; +function synchronously_get_policy(url) +{ + const encoded_url = encodeURIComponent(url); + const request_url = `${browser.runtime.getURL("dummy")}?url=${encoded_url}`; + + try { + var xhttp = new XMLHttpRequest(); + xhttp.open("GET", request_url, false); + xhttp.send(); + } catch(e) { + console.error("Failure to synchronously fetch policy for url.", e); + return {allow: false}; } + const policy = /^[^?]*\?settings=(.*)$/.exec(xhttp.responseURL)[1]; + return JSON.parse(decodeURIComponent(policy)); +} + +if (!is_privileged_url(document.URL)) { + const policy = synchronously_get_policy(document.URL); + if (!(document instanceof HTMLDocument)) - policy.has_payload = false; + delete policy.payload; console.debug("current policy", policy); + report_settings(policy); + + policy.nonce = gen_nonce(); + const doc_ready = Promise.all([ policy.allow ? Promise.resolve() : sanitize_document(document, policy), policy.allow ? Promise.resolve() : disable_service_workers(), wait_loaded(document) ]); - handle_page_actions(policy.nonce, policy_received_callback, doc_ready); + handle_page_actions(policy, doc_ready); start_activity_info_server(); } diff --git a/content/page_actions.js b/content/page_actions.js index db7c352..845e452 100644 --- a/content/page_actions.js +++ b/content/page_actions.js @@ -12,19 +12,17 @@ * IMPORT CONNECTION_TYPE * IMPORT browser * IMPORT report_script - * IMPORT report_settings * IMPORT report_document_type * IMPORTS_END */ -let policy_received_callback; +let policy; /* Snapshot url and content type early; these can be changed by other code. */ let url; let is_html; let port; let loaded = false; let scripts_awaiting = []; -let nonce; function handle_message(message) { @@ -38,9 +36,8 @@ function handle_message(message) scripts_awaiting.push(script_text); } } - if (action === "settings") { - report_settings(data); - policy_received_callback({url, allow: data[1].allow}); + else { + console.error(`Bad page action '${action}'.`); } } @@ -61,27 +58,27 @@ function add_script(script_text) let script = document.createElement("script"); script.textContent = script_text; - script.setAttribute("nonce", nonce); + script.setAttribute("nonce", policy.nonce); script.haketilo_payload = true; document.body.appendChild(script); report_script(script_text); } -function handle_page_actions(script_nonce, policy_received_cb, - doc_ready_promise) { - policy_received_callback = policy_received_cb; +function handle_page_actions(_policy, doc_ready_promise) { + policy = _policy; + url = document.URL; is_html = document instanceof HTMLDocument; report_document_type(is_html); doc_ready_promise.then(document_ready); - port = browser.runtime.connect({name : CONNECTION_TYPE.PAGE_ACTIONS}); - port.onMessage.addListener(handle_message); - port.postMessage({url}); - - nonce = script_nonce; + if (policy.payload) { + port = browser.runtime.connect({name : CONNECTION_TYPE.PAGE_ACTIONS}); + port.onMessage.addListener(handle_message); + port.postMessage({payload: policy.payload}); + } } /* diff --git a/dummy b/dummy new file mode 100644 index 0000000..e69de29 diff --git a/html/display-panel.js b/html/display-panel.js index c078850..4fe0173 100644 --- a/html/display-panel.js +++ b/html/display-panel.js @@ -229,14 +229,14 @@ function handle_activity_report(message) const [type, data] = message; if (type === "settings") { - let [pattern, settings] = data; + const settings = data; blocked_span.textContent = settings.allow ? "no" : "yes"; - if (pattern) { + if (settings.pattern) { pattern_span.textContent = pattern; const settings_opener = - () => open_in_settings(TYPE_PREFIX.PAGE, pattern); + () => open_in_settings(TYPE_PREFIX.PAGE, settings.pattern); view_pattern_but.classList.remove("hide"); view_pattern_but.addEventListener("click", settings_opener); } else { @@ -244,11 +244,10 @@ function handle_activity_report(message) blocked_span.textContent = blocked_span.textContent + " (default)"; } - const components = settings.components; - if (components) { - payload_span.textContent = nice_name(...components); + if (settings.payload) { + payload_span.textContent = nice_name(...settings.payload); payload_buttons_div.classList.remove("hide"); - const settings_opener = () => open_in_settings(...components); + const settings_opener = () => open_in_settings(...settings.payload); view_payload_but.addEventListener("click", settings_opener); } else { payload_span.textContent = "none"; diff --git a/manifest.json b/manifest.json index b18ea3e..7b4cb26 100644 --- a/manifest.json +++ b/manifest.json @@ -44,8 +44,7 @@ "page": "html/options.html", "open_in_tab": true }_CHROMIUM_UPDATE_URL_, - "web_accessible_resources": [_CHROMIUM_KEY_ - ], + "web_accessible_resources": ["dummy"], "background": { "persistent": true, "scripts": [_BGSCRIPTS_] -- cgit v1.2.3