Replace fewer TLDs when normalizing

maxmind · Mar 15, 2024 · 5b0d822 · 5b0d822
1 parent 26e43ea
commit 5b0d822
Show file tree

Hide file tree

Showing 3 changed files with 76 additions and 12 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -20,11 +20,8 @@ CHANGELOG
 * Duplicate `.com`s are now removed from email domain names when
   `hashEmail` is used. For example, `example.com.com` will become
   `example.com`.
-* Extraneous characters after `.com` are now removed from email domain
-  names when `hashEmail` is used. For example, `example.comfoo` will become
-  `example.com`.
-* Certain `.com` typos are now normalized to `.com` when `hashEmail` is
-  used. For example, `example.cam` will become `example.com`.
+* Certain TLD typos are now normalized when `hashEmail` is used. For
+  example, `example.comcom` will become `example.com`.
 * Additional `gmail.com` domain names with leading digits are now
   normalized when `hashEmail` is used. For example, `100gmail.com` will
   become `gmail.com`.

diff --git a/src/MinFraud/Util.php b/src/MinFraud/Util.php
@@ -31,6 +31,57 @@ class Util
         'putlook.com' => 'outlook.com',
     ];
 
+    /**
+     * @var array<string, string>
+     */
+    private static $typoTLDs = [
+        'comm' => 'com',
+        'commm' => 'com',
+        'commmm' => 'com',
+        'comn' => 'com',
+
+        'cbm' => 'com',
+        'ccm' => 'com',
+        'cdm' => 'com',
+        'cem' => 'com',
+        'cfm' => 'com',
+        'cgm' => 'com',
+        'chm' => 'com',
+        'cim' => 'com',
+        'cjm' => 'com',
+        'ckm' => 'com',
+        'clm' => 'com',
+        'cmm' => 'com',
+        'cnm' => 'com',
+        'cpm' => 'com',
+        'cqm' => 'com',
+        'crm' => 'com',
+        'csm' => 'com',
+        'ctm' => 'com',
+        'cum' => 'com',
+        'cvm' => 'com',
+        'cwm' => 'com',
+        'cxm' => 'com',
+        'cym' => 'com',
+        'czm' => 'com',
+
+        'col' => 'com',
+        'con' => 'com',
+
+        'dom' => 'com',
+        'don' => 'com',
+        'som' => 'com',
+        'son' => 'com',
+        'vom' => 'com',
+        'von' => 'com',
+        'xom' => 'com',
+        'xon' => 'com',
+
+        'clam' => 'com',
+        'colm' => 'com',
+        'comcom' => 'com',
+    ];
+
     /**
      * @var array<string, string>
      */
@@ -271,10 +322,16 @@ private static function cleanDomain(string $domain): string
         }
 
         $domain = preg_replace('/(?:\.com){2,}$/', '.com', $domain);
-        $domain = preg_replace('/\.com[^.]+$/', '.com', $domain);
-        $domain = preg_replace('/(?:\.(?:com|c[a-z]{1,2}m|co[ln]|[dsvx]o[mn]|))$/', '.com', $domain);
         $domain = preg_replace('/^\d+(?:gmail?\.com)$/', 'gmail.com', $domain);
 
+        $idx = strrpos($domain, '.');
+        if ($idx !== false) {
+            $tld = substr($domain, $idx + 1);
+            if (isset(self::$typoTLDs[$tld])) {
+                $domain = substr($domain, 0, $idx) . '.' . self::$typoTLDs[$tld];
+            }
+        }
+
         if (isset(self::$typoDomains[$domain])) {
             $domain = self::$typoDomains[$domain];
         }

diff --git a/tests/MaxMind/Test/MinFraud/UtilTest.php b/tests/MaxMind/Test/MinFraud/UtilTest.php
@@ -223,18 +223,18 @@ public function testMaybeHashEmail(): void
                 'input' => ['email' => ['address' => 'foo@example.comfoo']],
                 'expected' => [
                     'email' => [
-                        'address' => md5('foo@example.com'),
-                        'domain' => 'example.com',
+                        'address' => md5('foo@example.comfoo'),
+                        'domain' => 'example.comfoo',
                     ],
                 ],
             ],
             [
-                'name' => '.com typo',
+                'name' => '.cam',
                 'input' => ['email' => ['address' => 'foo@example.cam']],
                 'expected' => [
                     'email' => [
-                        'address' => md5('foo@example.com'),
-                        'domain' => 'example.com',
+                        'address' => md5('foo@example.cam'),
+                        'domain' => 'example.cam',
                     ],
                 ],
             ],
@@ -258,6 +258,16 @@ public function testMaybeHashEmail(): void
                     ],
                 ],
             ],
+            [
+                'name' => 'TLD typo',
+                'input' => ['email' => ['address' => 'foo@example.comcom']],
+                'expected' => [
+                    'email' => [
+                        'address' => md5('foo@example.com'),
+                        'domain' => 'example.com',
+                    ],
+                ],
+            ],
         ];
 
         if (\function_exists('idn_to_ascii')