Remove pixel cookie sharing detection for now

Author: ghostwords

src/js/background.js

@@ -48,8 +48,6 @@ function Badger(from_qunit) {
       manifestJson.background.persistent === false);
   }());
 
-  self.firstPartyDomainPotentiallyRequired = testCookiesFirstPartyDomain();
-
   self.widgetList = [];
   let widgetListPromise = widgetLoader.loadWidgetsFromFile(
     "data/socialwidgets.json").catch(console.error);
@@ -137,30 +135,6 @@ function Badger(from_qunit) {
     }
   }
 
-  /**
-   * Checks for availability of firstPartyDomain chrome.cookies API parameter.
-   * https://developer.mozilla.org/en-US/docs/Mozilla/Add-ons/WebExtensions/API/cookies/getAll#Parameters
-   *
-   * firstPartyDomain is required when privacy.websites.firstPartyIsolate is enabled,
-   * and is in Firefox since Firefox 59. (firstPartyIsolate is in Firefox since 58).
-   *
-   * We don't care whether firstPartyIsolate is enabled, but rather whether
-   * firstPartyDomain is supported. Assuming firstPartyDomain is supported,
-   * setting it to null in chrome.cookies.getAll() produces the same result
-   * regardless of the state of firstPartyIsolate.
-   *
-   * firstPartyDomain is not currently supported in Chrome.
-   */
-  function testCookiesFirstPartyDomain() {
-    try {
-      chrome.cookies.getAll({
-        firstPartyDomain: null
-      }, function () {});
-    } catch (ex) {
-      return false;
-    }
-    return true;
-  }
 } /* end of Badger constructor */
 
 Badger.prototype = {

src/js/heuristicblocking.js

@@ -193,178 +193,6 @@ HeuristicBlocker.prototype = {
     }
   },
 
-  /**
-   * Calls the pixel cookie sharing checking function
-   * iff the request is for an image in the top-level frame,
-   * and the request URL has querystring parameters.
-   *
-   * @param {Object} details webRequest onResponseStarted details object
-   */
-  checkForPixelCookieSharing: function (details) {
-    if (!badger.isLearningEnabled(details.tabId)) {
-      return;
-    }
-
-    if (details.type != 'image' || details.frameId !== 0 || details.url.indexOf('?') == -1) {
-      return;
-    }
-
-    let self = this,
-      tab_base = self.tabBases[details.tabId];
-    if (!tab_base) {
-      return;
-    }
-    let tab_url = self.tabUrls[details.tabId];
-
-    let request_host = (new URI(details.url)).host;
-    // CNAME uncloaking
-    if (utils.hasOwn(badger.cnameDomains, request_host)) {
-      request_host = badger.cnameDomains[request_host];
-    }
-    let request_base = getBaseDomain(request_host);
-
-    let initiator_url = getInitiatorUrl(tab_url, details);
-    if (initiator_url) {
-      tab_url = initiator_url;
-      tab_base = getBaseDomain(extractHostFromURL(initiator_url));
-    }
-
-    // ignore first-party requests
-    if (!utils.isThirdPartyDomain(request_base, tab_base)) {
-      return;
-    }
-
-    // short-circuit if we already observed this eTLD+1 tracking on this site
-    let firstParties = self.storage.getStore('snitch_map').getItem(request_base);
-    if (firstParties && firstParties.includes(tab_base)) {
-      return;
-    }
-
-    // short-circuit if we already made a decision for this FQDN
-    let action = self.storage.getBestAction(request_host);
-    if (action != constants.NO_TRACKING && action != constants.ALLOW) {
-      return;
-    }
-
-    // get all non-HttpOnly cookies for the top-level frame
-    // and pass those to the pixel cookie-share accounting function
-    let config = {
-      url: tab_url
-    };
-    if (badger.firstPartyDomainPotentiallyRequired) {
-      config.firstPartyDomain = null;
-    }
-    chrome.cookies.getAll(config, function (cookies) {
-      cookies = cookies.filter(cookie => !cookie.httpOnly);
-      if (cookies.length < 1) {
-        return;
-      }
-
-      // TODO refactor with new URI() above?
-      let searchParams = (new URL(details.url)).searchParams;
-
-      self.pixelCookieShareAccounting(tab_url, tab_base, searchParams, request_host, request_base, cookies);
-    });
-  },
-
-  /**
-   * Checks for cookie sharing: requests to third-party domains
-   * that include high entropy data from first-party cookies.
-   *
-   * Only catches plain-text verbatim sharing (b64 encoding etc. defeat it).
-   *
-   * Assumes any long string that doesn't contain URL fragments
-   * or stopwords is an identifier.
-   *
-   * Doesn't catch cookie syncing (3rd party -> 3rd party),
-   * but most of those tracking cookies should be blocked anyway.
-   */
-  pixelCookieShareAccounting: function (tab_url, tab_base, searchParams, request_host, request_base, cookies) {
-    const TRACKER_ENTROPY_THRESHOLD = 33,
-      MIN_STR_LEN = 8;
-
-    let self = this;
-
-    for (let p of searchParams) {
-      let key = p[0],
-        value = p[1];
-
-      // the argument must be sufficiently long
-      if (!value || value.length < MIN_STR_LEN) {
-        continue;
-      }
-
-      // check if this argument is derived from a high-entropy first-party cookie
-      for (let cookie of cookies) {
-        // the cookie value must be sufficiently long
-        if (!cookie.value || cookie.value.length < MIN_STR_LEN) {
-          continue;
-        }
-
-        // find the longest common substring between this arg and the cookies
-        // associated with the document
-        let substrings = utils.findCommonSubstrings(cookie.value, value) || [];
-        for (let s of substrings) {
-          // ignore the substring if it's part of the first-party URL. sometimes
-          // content servers take the url of the page they're hosting content
-          // for as an argument. e.g.
-          // https://example-cdn.com/content?u=http://example.com/index.html
-          if (tab_url.indexOf(s) != -1) {
-            continue;
-          }
-
-          // elements of the user agent string are also commonly included in
-          // both cookies and arguments; e.g. "Mozilla/5.0" might be in both.
-          // This is not a special tracking risk since third parties can see
-          // this info anyway.
-          if (navigator.userAgent.indexOf(s) != -1) {
-            continue;
-          }
-
-          // Sometimes the entire url and then some is included in the
-          // substring -- the common string might be "https://example.com/:true"
-          // In that case, we only care about the information around the URL.
-          if (s.indexOf(tab_url) != -1) {
-            s = s.replace(tab_url, "");
-          }
-
-          // During testing we found lots of common values like "homepage",
-          // "referrer", etc. were being flagged as high entropy. This searches
-          // for a few of those and removes them before we go further.
-          let lower = s.toLowerCase();
-          lowEntropyQueryValues.forEach(function (qv) {
-            let start = lower.indexOf(qv);
-            if (start != -1) {
-              s = s.replace(s.substring(start, start + qv.length), "");
-            }
-          });
-
-          // at this point, since we might have removed things, make sure the
-          // string is still long enough to bother with
-          if (s.length < MIN_STR_LEN) {
-            continue;
-          }
-
-          // compute the entropy of this common substring. if it's greater than
-          // our threshold, record the tracking action and exit the function.
-          let entropy = utils.estimateMaxEntropy(s);
-          if (entropy > TRACKER_ENTROPY_THRESHOLD) {
-            log("Found high-entropy cookie share from", tab_base, "to", request_host,
-              ":", entropy, "bits\n  cookie:", cookie.name, '=', cookie.value,
-              "\n  arg:", key, "=", value, "\n  substring:", s);
-            self._recordPrevalence(request_host, request_base, tab_base);
-
-            // record pixel cookie sharing
-            badger.storage.recordTrackingDetails(
-              request_base, tab_base, 'pixelcookieshare');
-
-            return;
-          }
-        }
-      }
-    }
-  },
-
   /**
    * Wraps _recordPrevalence for use outside of webRequest listeners.
    *
@@ -660,51 +488,6 @@ var lowEntropyCookieValues = {
   "zu":8
 };
 
-const lowEntropyQueryValues = [
-  "https",
-  "http",
-  "://",
-  "%3A%2F%2F",
-  "www",
-  "url",
-  "undefined",
-  "impression",
-  "session",
-  "homepage",
-  "client",
-  "version",
-  "business",
-  "title",
-  "get",
-  "site",
-  "name",
-  "category",
-  "account_id",
-  "smartadserver",
-  "front",
-  "page",
-  "view",
-  "first",
-  "visit",
-  "platform",
-  "language",
-  "automatic",
-  "disabled",
-  "landing",
-  "entertainment",
-  "amazon",
-  "official",
-  "webvisor",
-  "anonymous",
-  "across",
-  "narrative",
-  "\":null",
-  "\":false",
-  "\":\"",
-  "\",\"",
-  "\",\"",
-];
-
 /**
  * Extract cookies from onBeforeSendHeaders
  *
@@ -820,9 +603,6 @@ function startListeners() {
       badger.heuristicBlocking.checkForTrackingCookies(details);
     }
 
-    // check for pixel cookie sharing if the response appears to be for an image pixel
-    badger.heuristicBlocking.checkForPixelCookieSharing(details);
-
   }, {urls: ["http://*/*", "https://*/*"]}, extraInfoSpec);
 }
 

src/js/utils.js

@@ -175,48 +175,6 @@ function estimateMaxEntropy(str) {
   return max_bits;
 }
 
-// Adapted from https://gist.github.com/jaewook77/cd1e3aa9449d7ea4fb4f
-// Find all common substrings more than 8 characters long, using DYNAMIC
-// PROGRAMMING
-function findCommonSubstrings(str1, str2) {
-  /*
-   Let D[i,j] be the length of the longest matching string suffix between
-   str1[1]..str1[i] and a segment of str2 between str2[1]..str2[j].
-   If the ith character in str1 doesn’t match the jth character in str2, then
-   D[i,j] is zero to indicate that there is no matching suffix
-   */
-
-  // we only care about strings >= 8 chars
-  let D = [], LCS = [], LCS_MIN = 8;
-
-  // runs in O(M x N) time!
-  for (let i = 0; i < str1.length; i++) {
-    D[i] = [];
-    for (let j = 0; j < str2.length; j++) {
-      if (str1[i] == str2[j]) {
-        if (i == 0 || j == 0) {
-          D[i][j] = 1;
-        } else {
-          D[i][j] = D[i-1][j-1] + 1;
-        }
-
-        // store all common substrings longer than the minimum length
-        if (D[i][j] == LCS_MIN) {
-          LCS.push(str1.substring(i-D[i][j]+1, i+1));
-        } else if (D[i][j] > LCS_MIN) {
-          // remove the shorter substring and add the new, longer one
-          LCS.pop();
-          LCS.push(str1.substring(i-D[i][j]+1, i+1));
-        }
-      } else {
-        D[i][j] = 0;
-      }
-    }
-  }
-
-  return LCS;
-}
-
 function oneSecond() {
   return 1000;
 }
@@ -578,7 +536,6 @@ let utils = {
   explodeSubdomains,
   fetchResource,
   filter,
-  findCommonSubstrings,
   firstPartyProtectionsEnabled,
   getHostFromDomainInput,
   hasOwn,

src/manifest.json

@@ -21,7 +21,6 @@
     "webRequest",
     "webRequestBlocking",
     "storage",
-    "cookies",
     "privacy"
   ],
   "background": {

src/tests/tests/utils.js

@@ -818,30 +818,6 @@ QUnit.module("Utils", function (/*hooks*/) {
       "Domains that begin with http are valid entries");
   });
 
-  // Tests algorithm used in the pixel tracking heuristic
-  // It should return a common substring between two given values
-  QUnit.test("findCommonSubstrings", assert => {
-
-    assert.deepEqual(
-      utils.findCommonSubstrings('www.foo.bar', 'www.foob.ar'),
-      [],
-      "substrings under the length threshold of 8 are ignored"
-    );
-
-    assert.equal(
-      utils.findCommonSubstrings('foobar.com/foo/fizz/buzz/bar', 'foobar.com/foo/bizz/fuzz/bar')[0],
-      'foobar.com/foo/',
-      "returns longest matching value from the pair of URLs"
-    );
-
-    assert.deepEqual(
-      utils.findCommonSubstrings('foobar.com/fizz/buzz/bar/foo', 'foobar.com/fizzbuzz/buzz/bar/foo'),
-      ['foobar.com/fizz', "zz/buzz/bar/foo"],
-      "returns multiple substrings if multiple are present in comparison"
-    );
-
-  });
-
   // used in pixel tracking heuristic, given a string the estimateMaxEntropy function
   // will return the estimated entropy value from it, based on logic parsing the string's length,
   // and classes of character complication included in the string