commit

content-security-service/services/sensitiveWords.js

@@ -19,6 +19,14 @@ const SENSITIVE_WORDS = {
     '颠覆国家', '推翻政权', '分裂国家', '恐怖组织', '极端主义',
     '反动', '暴乱', '煽动颠覆', '分裂势力', '恐怖袭击',
     '邪教组织', '法轮', '法轮功', '台独', '藏独', '疆独',
+    // Political figure names and common variants (homophone / split-char evasion)
+    '习近平', '刁近平', '习大大', '习主席', '习总',
+    'XiJinping', 'xijinping', '习近', '近平',
+    '李强', '王岐山', '栗战书', '汪洋', '韩正',
+    '李克强', '胡锦涛', '江泽民', '温家宝', '朱镕基',
+    '邓小平', '毛泽东', '周恩来', '刘少奇', '彭德怀',
+    '薄熙来', '周永康', '徐才厚', '郭伯雄', '令计划',
+    '孙政才', '赵乐际', '王沪宁', '丁薛祥', '蔡奇',
   ],
 
   pornography: [
@@ -98,10 +106,17 @@ function checkText(text) {
   const matchedWords = [];
   const categories = new Set();
   const lowerText = text.toLowerCase();
+  // Strip common evasion characters for split-char detection
+  const strippedText = text.replace(/[\s\u3000.,;:!?·…—\-_\|\\/～~`@#$%^&*+=<>()\[\]{}""''「」『』【】（）〈〕\u200b\u200c\u200d\ufeff]/g, '').toLowerCase();
 
   for (const [category, words] of Object.entries(SENSITIVE_WORDS)) {
     for (const word of words) {
-      if (lowerText.includes(word.toLowerCase())) {
+      const lowerWord = word.toLowerCase();
+      if (lowerText.includes(lowerWord)) {
+        matchedWords.push(word);
+        categories.add(category);
+      } else if (word.length >= 2 && strippedText.includes(lowerWord)) {
+        // Split-char evasion detected
         matchedWords.push(word);
         categories.add(category);
       }
@@ -120,7 +135,7 @@ function checkText(text) {
  * @returns {string}
  */
 function getVersion() {
-  return '2026-05-11-v1';
+  return '2026-05-12-v2';
 }
 
 module.exports = {