Added new rule on anti-spam for unicode characters

Author: tabuna

app/Services/SpamDetector.php

@@ -65,6 +65,36 @@ public function containsStopWords(): bool
         return Str::of($this->message)->contains($this->stopWords, true);
     }
 
+    /**
+     * Checks if the message contains an excessive amount of special characters.
+     * For example, the proportion of special characters should not exceed a given threshold (default is 2%).
+     *
+     * @param float $threshold
+     *
+     * @return bool
+     */
+    public function hasExcessiveUnicodeCharacters(float $threshold = 0.02): bool
+    {
+        // Length of the message including special characters
+        $withUnicode = Str::of($this->message)
+            ->replaceMatches('/^[^\p{L}\p{N}\p{Z}\p{P}]+|[^\p{L}\p{N}\p{Z}\p{P}]+$/u', '') // without start and end special characters (emoji, etc.)
+            ->length();
+
+        // Length of the message without special characters
+        $withOutUnicode = Str::of($this->message)
+            ->replaceMatches('/[^\p{L}\p{N}\p{Z}\p{P}]/u', '')
+            ->length();
+
+        // Difference in length
+        $unicodeLength = $withUnicode - $withOutUnicode;
+
+        // Proportion of special characters in the message
+        $unicodePercentage = $unicodeLength / $withUnicode;
+
+        // Check if the proportion of special characters exceeds the given threshold
+        return $unicodePercentage > $threshold;
+    }
+
     /**
      * Check if the message is spam using a Naive Bayes classifier.
      *
@@ -114,6 +144,10 @@ private function trainClassifier(Classifier $classifier, string $fileName, strin
      */
     public function isSpam()
     {
+        if ($this->hasExcessiveUnicodeCharacters()) {
+            return true;
+        }
+
         if ($this->containsStopWords()) {
             return true;
         }

tests/Unit/SpamDetectorTest.php

@@ -29,6 +29,26 @@ public static function messageProvider()
             ['Стабильный доход от 100$ Нужен только телефон', true],
             ['блокчейн в ЛС', true],
             ['Крипто инвестиции', true],
+            ['18+', true],
+            ['hamsterkombat', true],
+            ['hamster', true],
+            ['Прuвет', true],
         ];
     }
+    public function testUnicodeRules():void
+    {
+        //Long unicode characters in center of the message/words
+        $spamDetector = new SpamDetector('Прuвет всем, хoчу предлoжuть реaльный дoпoлнuтельный зaрaбoтoк!
+- От 50$ в/зa день гaрaнтuрoвaнo
+- Чaс в день твoегo временu
+- Честнo u легaльнo, НЕ НАРКОТИКИ!!
+
+Еслu ты действuтельнo зauнтересoвaн в быстрoм u честнoм зaрaбoтке , пuшu + в ЛС!!!!');
+
+        $this->assertTrue($spamDetector->hasExcessiveUnicodeCharacters());
+
+        // Long unicode characters at the end of the message/words
+        $spamDetector = new SpamDetector('🍕 Прикольно, что ты тут делаешь? 🍣🍰');
+        $this->assertFalse($spamDetector->hasExcessiveUnicodeCharacters());
+    }
 }