Mitigate spoofing attempt using Latin letters.

components/url_formatter/idn_spoof_checker.cc

+// Copyright 2017 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+//
+#include "components/url_formatter/idn_spoof_checker.h"
+
+#include "base/numerics/safe_conversions.h"
+#include "base/strings/string_split.h"
+#include "base/strings/string_util.h"
+#include "base/threading/thread_local_storage.h"
+#include "net/base/lookup_string_in_fixed_set.h"
+#include "third_party/icu/source/common/unicode/schriter.h"
+#include "third_party/icu/source/i18n/unicode/translit.h"
+
+namespace url_formatter {
+
+namespace {
+base::ThreadLocalStorage::StaticSlot tls_index = TLS_INITIALIZER;
+
+void OnThreadTermination(void* regex_matcher) {
+  delete reinterpret_cast<icu::RegexMatcher*>(regex_matcher);
+}
+
+#include "components/url_formatter/top_domains/alexa_skeletons-inc.cc"
+// All the domains in the above file have 3 or fewer labels.
+const size_t kNumberOfLabelsToCheck = 3;
+
+bool LookupMatchInTopDomains(base::StringPiece skeleton) {
+  DCHECK(skeleton.back() != '.');

[Peter Kasting, 2017/05/10 22:38:46]

Nit: DCHECK_NE('.', skeleton.back())?

[jungshik at Google, 2017/05/14 09:36:22]

Done.

+  auto labels = base::SplitStringPiece(skeleton, ".", base::KEEP_WHITESPACE,
+                                       base::SPLIT_WANT_ALL);
+
+  if (labels.size() > kNumberOfLabelsToCheck) {
+    labels.erase(labels.begin(),
+                 labels.begin() + labels.size() - kNumberOfLabelsToCheck);
+  }
+
+  while (labels.size() > 1) {
+    std::string partial_skeleton = base::JoinString(labels, ".");
+    if (net::LookupStringInFixedSet(
+            kDafsa, arraysize(kDafsa), partial_skeleton.data(),
+            partial_skeleton.length()) != net::kDafsaNotFound)
+      return true;
+    labels.erase(labels.begin());
+  }
+  return false;
+}
+
+}  // namespace
+
+IDNSpoofChecker::IDNSpoofChecker() {
+  UErrorCode status = U_ZERO_ERROR;
+  checker_ = uspoof_open(&status);
+  if (U_FAILURE(status)) {
+    checker_ = nullptr;
+    return;
+  }
+
+  // At this point, USpoofChecker has all the checks enabled except
+  // for USPOOF_CHAR_LIMIT (USPOOF_{RESTRICTION_LEVEL, INVISIBLE,
+  // MIXED_SCRIPT_CONFUSABLE, WHOLE_SCRIPT_CONFUSABLE, MIXED_NUMBERS, ANY_CASE})
+  // This default configuration is adjusted below as necessary.
+
+  // Set the restriction level to moderate. It allows mixing Latin with another
+  // script (+ COMMON and INHERITED). Except for Chinese(Han + Bopomofo),
+  // Japanese(Hiragana + Katakana + Han), and Korean(Hangul + Han), only one
+  // script other than Common and Inherited can be mixed with Latin. Cyrillic
+  // and Greek are not allowed to mix with Latin.
+  // See http://www.unicode.org/reports/tr39/#Restriction_Level_Detection
+  uspoof_setRestrictionLevel(checker_, USPOOF_MODERATELY_RESTRICTIVE);
+
+  // Sets allowed characters in IDN labels and turn on USPOOF_CHAR_LIMIT.
+  SetAllowedUnicodeSet(&status);
+
+  // Enable the return of auxillary (non-error) information.
+  // We used to disable WHOLE_SCRIPT_CONFUSABLE check explicitly, but as of
+  // ICU 58.1, WSC is a no-op in a single string check API.
+  int32_t checks = uspoof_getChecks(checker_, &status) | USPOOF_AUX_INFO;
+  uspoof_setChecks(checker_, checks, &status);
+
+  // Four characters handled differently by IDNA 2003 and IDNA 2008. UTS46
+  // transitional processing treats them as IDNA 2003 does; maps U+00DF and
+  // U+03C2 and drops U+200[CD].
+  deviation_characters_ = icu::UnicodeSet(
+      UNICODE_STRING_SIMPLE("[\\u00df\\u03c2\\u200c\\u200d]"), status);
+  deviation_characters_.freeze();
+
+  // Latin letters outside ASCII. 'Script_Extensions=Latin' is not necessary
+  // because additional characters pulled in with scx=Latn are not included in
+  // the allowed set.
+  non_ascii_latin_letters_ =
+      icu::UnicodeSet(UNICODE_STRING_SIMPLE("[[:Latin:] - [a-zA-Z]]"), status);
+  non_ascii_latin_letters_.freeze();
+
+  // The following two sets are parts of |dangerous_patterns_|.
+  kana_letters_exceptions_ = icu::UnicodeSet(
+      UNICODE_STRING_SIMPLE("[\\u3078-\\u307a\\u30d8-\\u30da\\u30fb-\\u30fe]"),
+      status);
+  kana_letters_exceptions_.freeze();
+  combining_diacritics_exceptions_ =
+      icu::UnicodeSet(UNICODE_STRING_SIMPLE("[\\u0300-\\u0339]"), status);
+  combining_diacritics_exceptions_.freeze();
+
+  // These Cyrillic letters look like Latin. A domain label entirely made of
+  // these letters is blocked as a simplified whole-script-spoofable.
+  cyrillic_letters_latin_alike_ =
+      icu::UnicodeSet(icu::UnicodeString("[асԁеһіјӏорԛѕԝхуъЬҽпгѵѡ]"), status);
+  cyrillic_letters_latin_alike_.freeze();
+
+  cyrillic_letters_ =
+      icu::UnicodeSet(UNICODE_STRING_SIMPLE("[[:Cyrl:]]"), status);
+  cyrillic_letters_.freeze();
+
+  // This set is used to determine whether or not to apply a slow

[Peter Kasting, 2017/05/10 22:38:46]

Since you're both moving this code to this file and modifying it, it's hard to
tell what the functional vs. nonfunctional diff from this change is.  Consider
splitting off a separate initial CL that just moves the code out to its own file
without modifying it (more than is necessary), then having this CL only make
functional changes.

[jungshik at Google, 2017/05/14 09:36:22]

Ok. A file reorg CL is  https://codereview.chromium.org/2877973003 .

+  // transliteration to remove diacritics to a given hostname before the
+  // confusable skeleton calculation for comparison with top domain names. If
+  // it has any character outside the set, the expensive step will be skipped
+  // because it cannot match any of top domain names.
+  // The last ([\u0300-\u0339] is a shorthand for "[:Identifier_Status=Allowed:]
+  // & [:Script_Extensions=Inherited:] - [\\u200C\\u200D]". The latter is a
+  // subset of the former but it does not matter because hostnames with
+  // characters outside the latter set would be rejected in an earlier step.
+  lgc_letters_n_ascii_ = icu::UnicodeSet(
+      UNICODE_STRING_SIMPLE("[[:Latin:][:Greek:][:Cyrillic:][0-9\\u002e_"
+                            "\\u002d][\\u0300-\\u0339]]"),
+      status);
+  lgc_letters_n_ascii_.freeze();
+
+  // Used for diacritics-removal before the skeleton calculation. Add
+  // "ł > l; ø > o; đ > d" that are not handled by "NFD; Nonspacing mark
+  // removal; NFC". On top of that, supplement the Unicode confusable list by
+  // replacing {U+043A (к), U+0138(ĸ), U+03BA(κ)}, U+04CF (ӏ) and U+043F(п) by
+  // 'k', 'l' and 'n', respectively.
+  // TODO(jshin): Revisit "ł > l; ø > o" mapping.
+  UParseError parse_error;
+  transliterator_.reset(icu::Transliterator::createFromRules(
+      UNICODE_STRING_SIMPLE("DropAcc"),
+      icu::UnicodeString("::NFD; ::[:Nonspacing Mark:] Remove; ::NFC;"
+                         " ł > l; ø > o; đ > d; ӏ > l; [кĸκ] > k; п > n;"),
+      UTRANS_FORWARD, parse_error, status));
+  DCHECK(U_SUCCESS(status))
+      << "Spoofchecker initalization failed due to an error: "
+      << u_errorName(status);
+}
+
+IDNSpoofChecker::~IDNSpoofChecker() {
+  uspoof_close(checker_);
+}
+
+bool IDNSpoofChecker::SafeToDisplayAsUnicode(base::StringPiece16 label,
+                                             bool is_tld_ascii) {
+  UErrorCode status = U_ZERO_ERROR;
+  int32_t result =
+      uspoof_check(checker_, label.data(),
+                   base::checked_cast<int32_t>(label.size()), NULL, &status);
+  // If uspoof_check fails (due to library failure), or if any of the checks
+  // fail, treat the IDN as unsafe.
+  if (U_FAILURE(status) || (result & USPOOF_ALL_CHECKS))
+    return false;
+
+  icu::UnicodeString label_string(FALSE, label.data(),
+                                  base::checked_cast<int32_t>(label.size()));
+
+  // A punycode label with 'xn--' prefix is not subject to the URL
+  // canonicalization and is stored as it is in GURL. If it encodes a deviation
+  // character (UTS 46; e.g. U+00DF/sharp-s), it should be still shown in
+  // punycode instead of Unicode. Without this check, xn--fu-hia for
+  // 'fu<sharp-s>' would be converted to 'fu<sharp-s>' for display because
+  // "UTS 46 section 4 Processing step 4" applies validity criteria for
+  // non-transitional processing (i.e. do not map deviation characters) to any
+  // punycode labels regardless of whether transitional or non-transitional is
+  // chosen. On the other hand, 'fu<sharp-s>' typed or copy and pasted
+  // as Unicode would be canonicalized to 'fuss' by GURL and is displayed as
+  // such. See http://crbug.com/595263 .
+  if (deviation_characters_.containsSome(label_string))
+    return false;
+
+  // If there's no script mixing, the input is regarded as safe without any
+  // extra check unless it falls into one of three categories:
+  //   - contains Kana letter exceptions
+  //   - the TLD is ASCII and the input is made entirely of Cyrillic letters
+  //     that look like Latin letters.
+  //   - it has combining diacritic marks.
+  // Note that the following combinations of scripts are treated as a 'logical'
+  // single script.
+  //  - Chinese: Han, Bopomofo, Common
+  //  - Japanese: Han, Hiragana, Katakana, Common
+  //  - Korean: Hangul, Han, Common
+  result &= USPOOF_RESTRICTION_LEVEL_MASK;
+  if (result == USPOOF_ASCII)
+    return true;
+  if (result == USPOOF_SINGLE_SCRIPT_RESTRICTIVE &&
+      kana_letters_exceptions_.containsNone(label_string) &&
+      combining_diacritics_exceptions_.containsNone(label_string)) {
+    // Check Cyrillic confusable only for ASCII TLDs.
+    return !is_tld_ascii || !IsMadeOfLatinAlikeCyrillic(label_string);
+  }
+
+  // Additional checks for |label| with multiple scripts, one of which is Latin.
+  // Disallow non-ASCII Latin letters to mix with a non-Latin script.
+  // Note that the non-ASCII Latin check should not be applied when the entire
+  // label is made of Latin. Checking with lgc_letters set here should be fine
+  // because script mixing of LGC is already rejected.
+  if (non_ascii_latin_letters_.containsSome(label_string) &&
+      !lgc_letters_n_ascii_.containsAll(label_string))
+    return false;
+
+  if (!tls_index.initialized())
+    tls_index.Initialize(&OnThreadTermination);
+  icu::RegexMatcher* dangerous_pattern =
+      reinterpret_cast<icu::RegexMatcher*>(tls_index.Get());
+  if (!dangerous_pattern) {
+    // Disallow the katakana no, so, zo, or n, as they may be mistaken for
+    // slashes when they're surrounded by non-Japanese scripts (i.e. scripts
+    // other than Katakana, Hiragana or Han). If {no, so, zo, n} next to a
+    // non-Japanese script on either side is disallowed, legitimate cases like
+    // '{vitamin in Katakana}b6' are blocked. Note that trying to block those
+    // characters when used alone as a label is futile because those cases
+    // would not reach here.
+    // Also disallow what used to be blocked by mixed-script-confusable (MSC)
+    // detection. ICU 58 does not detect MSC any more for a single input string.
+    // See http://bugs.icu-project.org/trac/ticket/12823 .
+    // TODO(jshin): adjust the pattern once the above ICU bug is fixed.
+    // - Disallow U+30FB (Katakana Middle Dot) and U+30FC (Hiragana-Katakana
+    //   Prolonged Sound) used out-of-context.
+    // - Dislallow U+30FD/E (Katakana iteration mark/voiced iteration mark)
+    //   unless they're preceded by a Katakana.
+    // - Disallow three Hiragana letters (U+307[8-A]) or Katakana letters
+    //   (U+30D[8-A]) that look exactly like each other when they're used in a
+    //   label otherwise entirely in Katakna or Hiragana.
+    // - Disallow U+0585 (Armenian Small Letter Oh) and U+0581 (Armenian Small
+    //   Letter Co) to be next to Latin.
+    // - Disallow Latin 'o' and 'g' next to Armenian.
+    // - Disallow combining diacritical mark (U+0300-U+0339) after a non-LGC
+    //   character. Other combining diacritical marks are not in the allowed
+    //   character set.
+    dangerous_pattern = new icu::RegexMatcher(
+        icu::UnicodeString(
+            "[^\\p{scx=kana}\\p{scx=hira}\\p{scx=hani}]"
+            "[\\u30ce\\u30f3\\u30bd\\u30be]"
+            "[^\\p{scx=kana}\\p{scx=hira}\\p{scx=hani}]|"
+            "[^\\p{scx=kana}\\p{scx=hira}]\\u30fc|^\\u30fc|"
+            "[^\\p{scx=kana}][\\u30fd\\u30fe]|^[\\u30fd\\u30fe]|"
+            "^[\\p{scx=kana}]+[\\u3078-\\u307a][\\p{scx=kana}]+$|"
+            "^[\\p{scx=hira}]+[\\u30d8-\\u30da][\\p{scx=hira}]+$|"
+            "[a-z]\\u30fb|\\u30fb[a-z]|"
+            "^[\\u0585\\u0581]+[a-z]|[a-z][\\u0585\\u0581]+$|"
+            "[a-z][\\u0585\\u0581]+[a-z]|"
+            "^[og]+[\\p{scx=armn}]|[\\p{scx=armn}][og]+$|"
+            "[\\p{scx=armn}][og]+[\\p{scx=armn}]|"
+            "[^\\p{scx=latn}\\p{scx=grek}\\p{scx=cyrl}][\\u0300-\\u0339]",
+            -1, US_INV),
+        0, status);
+    tls_index.Set(dangerous_pattern);
+  }
+  dangerous_pattern->reset(label_string);
+  return !dangerous_pattern->find();
+}
+
+bool IDNSpoofChecker::SimilarToTopDomains(base::StringPiece16 hostname) {
+  size_t hostname_length = hostname.length() - (hostname.back() == '.' ? 1 : 0);
+  icu::UnicodeString ustr_host(FALSE, hostname.data(), hostname_length);
+  // If input has any characters outside Latin-Greek-Cyrillic and [0-9._-],
+  // there is no point in getting rid of diacritics because combining marks
+  // attached to non-LGC characters are already blocked.
+  if (lgc_letters_n_ascii_.span(ustr_host, 0, USET_SPAN_CONTAINED) ==
+      ustr_host.length())
+    transliterator_.get()->transliterate(ustr_host);
+
+  UErrorCode status = U_ZERO_ERROR;
+  icu::UnicodeString ustr_skeleton;
+  uspoof_getSkeletonUnicodeString(checker_, 0, ustr_host, ustr_skeleton,
+                                  &status);
+  if (U_FAILURE(status))
+    return false;
+  std::string skeleton;
+  ustr_skeleton.toUTF8String(skeleton);
+  return LookupMatchInTopDomains(skeleton);
+}
+
+bool IDNSpoofChecker::IsMadeOfLatinAlikeCyrillic(
+    const icu::UnicodeString& label_string) {
+  // Collect all the Cyrillic letters in |label_string| and see if they're
+  // a subset of |cyrillic_letters_latin_alike_|.

[Peter Kasting, 2017/05/10 22:38:46]

Nit: The sentence above is the sort of thing I'd expect to see on the function
declaration instead.

[jungshik at Google, 2017/05/14 09:36:22]

Added to the header.

+  // A shortcut of defining cyrillic_letters_latin_alike_ to include [0-9] and
+  // [_-] and checking if the set contains all letters of |label_string|
+  // would work in most cases, but not if a label has non-letters outside
+  // ASCII.
+  icu::UnicodeSet cyrillic_in_label;
+  icu::StringCharacterIterator it(label_string);
+  for (it.setToStart(); it.hasNext();) {
+    const UChar32 c = it.next32PostInc();
+    if (cyrillic_letters_.contains(c))
+      cyrillic_in_label.add(c);
+  }
+  return !cyrillic_in_label.isEmpty() &&
+         cyrillic_letters_latin_alike_.containsAll(cyrillic_in_label);
+}
+
+void IDNSpoofChecker::SetAllowedUnicodeSet(UErrorCode* status) {
+  if (U_FAILURE(*status))
+    return;
+
+  // The recommended set is a set of characters for identifiers in a
+  // security-sensitive environment taken from UTR 39
+  // (http://unicode.org/reports/tr39/) and
+  // http://www.unicode.org/Public/security/latest/xidmodifications.txt .
+  // The inclusion set comes from "Candidate Characters for Inclusion
+  // in idenfiers" of UTR 31 (http://www.unicode.org/reports/tr31). The list
+  // may change over the time and will be updated whenever the version of ICU
+  // used in Chromium is updated.
+  const icu::UnicodeSet* recommended_set =
+      uspoof_getRecommendedUnicodeSet(status);
+  icu::UnicodeSet allowed_set;
+  allowed_set.addAll(*recommended_set);
+  const icu::UnicodeSet* inclusion_set = uspoof_getInclusionUnicodeSet(status);
+  allowed_set.addAll(*inclusion_set);
+
+// Five aspirational scripts are taken from UTR 31 Table 6 at
+// http://www.unicode.org/reports/tr31/#Aspirational_Use_Scripts .
+// Not all the characters of aspirational scripts are suitable for
+// identifiers. Therefore, only characters belonging to
+// [:Identifier_Type=Aspirational:] (listed in 'Status/Type=Aspirational'
+// section at
+// http://www.unicode.org/Public/security/latest/xidmodifications.txt) are
+// are added to the allowed set. The list has to be updated when a new
+// version of Unicode is released. The current version is 9.0.0 and ICU 60
+// will have Unicode 10.0 data.
+#if U_ICU_VERSION_MAJOR_NUM < 60
+  const icu::UnicodeSet aspirational_scripts(
+      icu::UnicodeString(
+          // Unified Canadian Syllabics
+          "[\\u1401-\\u166C\\u166F-\\u167F"
+          // Mongolian
+          "\\u1810-\\u1819\\u1820-\\u1877\\u1880-\\u18AA"
+          // Unified Canadian Syllabics
+          "\\u18B0-\\u18F5"
+          // Tifinagh
+          "\\u2D30-\\u2D67\\u2D7F"
+          // Yi
+          "\\uA000-\\uA48C"
+          // Miao
+          "\\U00016F00-\\U00016F44\\U00016F50-\\U00016F7E"
+          "\\U00016F8F-\\U00016F9F]",
+          -1, US_INV),
+      *status);
+  allowed_set.addAll(aspirational_scripts);
+#else
+#error "Update aspirational_scripts per Unicode 10.0"
+#endif
+
+  // U+0338 is included in the recommended set, while U+05F4 and U+2027 are in
+  // the inclusion set. However, they are blacklisted as a part of Mozilla's
+  // IDN blacklist (http://kb.mozillazine.org/Network.IDN.blacklist_chars).
+  // U+2010 is in the inclusion set, but we drop it because it can be confused
+  // with an ASCII U+002D (Hyphen-Minus).
+  // U+0338 and U+2027 are dropped; the former can look like a slash when
+  // rendered with a broken font, and the latter can be confused with U+30FB
+  // (Katakana Middle Dot). U+05F4 (Hebrew Punctuation Gershayim) is kept,
+  // even though it can look like a double quotation mark. Using it in Hebrew
+  // should be safe. When used with a non-Hebrew script, it'd be filtered by
+  // other checks in place.
+  allowed_set.remove(0x338u);   // Combining Long Solidus Overlay
+  allowed_set.remove(0x2010u);  // Hyphen
+  allowed_set.remove(0x2027u);  // Hyphenation Point
+
+#if defined(OS_MACOSX)
+  // The following characters are reported as present in the default macOS
+  // system UI font, but they render as blank. Remove them from the allowed
+  // set to prevent spoofing.
+  // Tibetan characters used for transliteration of ancient texts:
+  allowed_set.remove(0x0F8Cu);
+  allowed_set.remove(0x0F8Du);
+  allowed_set.remove(0x0F8Eu);
+  allowed_set.remove(0x0F8Fu);
+#endif
+
+  uspoof_setAllowedUnicodeSet(checker_, &allowed_set, status);
+}
+
+}  // namespace url_formatter