feat(app): 内容安全词库 + 过滤服务 + 分享前检查 — 28 个测试全覆盖
新增文件： - sensitive_words.dart — 8 分类 ~200 条敏感词 + 谐音/形近/数字变体映射 - content_filter_service.dart — 精确匹配 + 变体匹配 + 文本预处理（去零宽/空格/符号） - content_filter_service_test.dart — 28 个测试（8分类精确/安全内容/预处理/变体/边界/词库完整性） 修改： - share_bottom_sheet.dart — 分享到班级前调用 ContentFilterService， 有敏感词时弹出警告对话框（返回修改/仍然分享），新增 contentText 参数
This commit is contained in:
iven
184
app/lib/data/services/content_filter_service.dart
Normal file
184
app/lib/data/services/content_filter_service.dart
Normal file
@@ -0,0 +1,184 @@
|
||||
// 内容安全过滤服务 — 本地敏感词检测
|
||||
//
|
||||
// 提供 checkText() 纯函数，用于在分享日记前检查文本内容是否包含敏感词。
|
||||
// 检测策略：精确匹配 + 谐音/形近/数字变体匹配。
|
||||
// 返回匹配列表，空列表表示内容安全。不自动屏蔽，由 UI 层决定提示方式。
|
||||
|
||||
import 'sensitive_words.dart';
|
||||
|
||||
/// 敏感词匹配结果
|
||||
class SensitiveWordMatch {
|
||||
/// 匹配到的敏感词原文
|
||||
final String word;
|
||||
|
||||
/// 所属分类
|
||||
final SensitiveCategory category;
|
||||
|
||||
/// 在预处理后文本中的起始位置
|
||||
final int position;
|
||||
|
||||
const SensitiveWordMatch({
|
||||
required this.word,
|
||||
required this.category,
|
||||
required this.position,
|
||||
});
|
||||
|
||||
@override
|
||||
String toString() => 'SensitiveWordMatch("$word", ${category.label}, @$position)';
|
||||
}
|
||||
|
||||
/// 内容安全过滤服务
|
||||
///
|
||||
/// 纯静态方法，无状态，可安全在任何地方调用。
|
||||
/// 性能：~200 条词 × contains() 检查，<1ms 完成。
|
||||
class ContentFilterService {
|
||||
ContentFilterService._();
|
||||
|
||||
/// 检查文本内容，返回所有匹配到的敏感词。
|
||||
///
|
||||
/// 对输入文本进行预处理（去空格/特殊符号/零宽字符/小写化），
|
||||
/// 然后遍历全量词库做精确匹配和谐音变体匹配。
|
||||
/// 返回空列表表示内容安全。
|
||||
static List<SensitiveWordMatch> checkText(String text) {
|
||||
if (text.isEmpty) return const [];
|
||||
|
||||
final normalized = _normalize(text);
|
||||
if (normalized.isEmpty) return const [];
|
||||
|
||||
final matches = <SensitiveWordMatch>[];
|
||||
final seen = <String>{}; // 去重：同一词不重复报告
|
||||
|
||||
for (final entry in kSensitiveWords.entries) {
|
||||
final category = entry.key;
|
||||
|
||||
for (final word in entry.value) {
|
||||
// 精确匹配
|
||||
final pos = normalized.indexOf(word);
|
||||
if (pos >= 0 && seen.add(word)) {
|
||||
matches.add(SensitiveWordMatch(
|
||||
word: word,
|
||||
category: category,
|
||||
position: pos,
|
||||
));
|
||||
}
|
||||
|
||||
// 谐音/变体匹配 — 将词中每个有变体映射的字替换为变体，检查是否命中
|
||||
if (_matchesWithVariants(normalized, word)) {
|
||||
if (seen.add('variant:$word')) {
|
||||
matches.add(SensitiveWordMatch(
|
||||
word: word,
|
||||
category: category,
|
||||
position: -1, // 变体匹配无法精确定位
|
||||
));
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// 整词变体匹配 — kHomophoneVariants 中的多字 key（如 "卧槽"）
|
||||
for (final variantEntry in kHomophoneVariants.entries) {
|
||||
final originalKey = variantEntry.key;
|
||||
if (originalKey.length <= 1) continue; // 单字已在上面处理
|
||||
|
||||
// 找到这个变体 key 对应的分类
|
||||
SensitiveCategory? foundCategory;
|
||||
for (final entry in kSensitiveWords.entries) {
|
||||
if (entry.value.contains(originalKey)) {
|
||||
foundCategory = entry.key;
|
||||
break;
|
||||
}
|
||||
}
|
||||
if (foundCategory == null) continue;
|
||||
|
||||
for (final variant in variantEntry.value) {
|
||||
if (variant.isEmpty) continue;
|
||||
final vPos = normalized.indexOf(variant.toLowerCase());
|
||||
if (vPos >= 0) {
|
||||
final key = 'wvariant:$originalKey:$variant';
|
||||
if (seen.add(key)) {
|
||||
matches.add(SensitiveWordMatch(
|
||||
word: originalKey,
|
||||
category: foundCategory,
|
||||
position: vPos,
|
||||
));
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
return matches;
|
||||
}
|
||||
|
||||
/// 检查文本是否包含敏感词（快捷方法）
|
||||
static bool hasSensitiveContent(String text) => checkText(text).isNotEmpty;
|
||||
|
||||
/// 获取匹配到的分类标签集合（用于 UI 展示）
|
||||
static Set<String> getMatchedCategories(List<SensitiveWordMatch> matches) {
|
||||
return matches.map((m) => m.category.label).toSet();
|
||||
}
|
||||
|
||||
/// 变体匹配：检查文本中是否出现了词的谐音/形近/数字变体版本
|
||||
///
|
||||
/// 将敏感词中每个有变体映射的字符逐一替换为变体，检查替换后的
|
||||
/// 字符串是否出现在文本中。例如 "去死" → 检查 "去4" 是否在文本中。
|
||||
static bool _matchesWithVariants(String normalizedText, String word) {
|
||||
final chars = word.split('');
|
||||
final variantChars = <List<String>>[];
|
||||
|
||||
for (final char in chars) {
|
||||
final variants = kHomophoneVariants[char];
|
||||
if (variants != null && variants.isNotEmpty) {
|
||||
// 原字符 + 所有变体
|
||||
variantChars.add([char, ...variants]);
|
||||
} else {
|
||||
variantChars.add([char]);
|
||||
}
|
||||
}
|
||||
|
||||
// 生成所有变体组合并检查
|
||||
return _checkCombinations(normalizedText, variantChars, 0, '');
|
||||
}
|
||||
|
||||
/// 递归生成变体组合并检查文本
|
||||
static bool _checkCombinations(
|
||||
String text,
|
||||
List<List<String>> variantChars,
|
||||
int index,
|
||||
String current,
|
||||
) {
|
||||
if (index == variantChars.length) {
|
||||
return text.contains(current);
|
||||
}
|
||||
for (final char in variantChars[index]) {
|
||||
if (_checkCombinations(text, variantChars, index + 1, current + char)) {
|
||||
return true;
|
||||
}
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
/// 文本预处理：去除干扰字符，统一为小写
|
||||
///
|
||||
/// 1. 去除零宽字符（U+200B~U+200F, U+FEFF）
|
||||
/// 2. 去除空格、制表符、换行
|
||||
/// 3. 去除常见特殊符号（用于绕过的 @#$%^&* 等）
|
||||
/// 4. 转小写（对英文词有效）
|
||||
static String _normalize(String text) {
|
||||
final buffer = StringBuffer();
|
||||
for (final rune in text.runes) {
|
||||
// 跳过零宽字符
|
||||
if (rune >= 0x200B && rune <= 0x200F) continue;
|
||||
if (rune == 0xFEFF) continue;
|
||||
// 跳过空白
|
||||
if (rune == 0x20 || rune == 0x09 || rune == 0x0A || rune == 0x0D) continue;
|
||||
// 跳过常见绕过符号
|
||||
if (rune == 0x2E || rune == 0x2C || rune == 0x2D || rune == 0x5F) continue; // . , - _
|
||||
if (rune == 0x21 || rune == 0x40 || rune == 0x23 || rune == 0x24) continue; // ! @ # $
|
||||
if (rune == 0x25 || rune == 0x5E || rune == 0x26 || rune == 0x2A) continue; // % ^ & *
|
||||
if (rune == 0x7E || rune == 0x60) continue; // ~ `
|
||||
|
||||
buffer.writeCharCode(rune);
|
||||
}
|
||||
return buffer.toString().toLowerCase();
|
||||
}
|
||||
}
|
||||
141
app/lib/data/services/sensitive_words.dart
Normal file
141
app/lib/data/services/sensitive_words.dart
Normal file
@@ -0,0 +1,141 @@
|
||||
// 敏感词库 — 本地静态词库常量，面向小学生场景
|
||||
//
|
||||
// 分类：暴力、色情、欺凌、毒品、赌博、政治、诈骗、粗口
|
||||
// 每个分类包含基础词 + 谐音/形近/数字变体
|
||||
// 词库为 const 编译期常量，零运行时开销
|
||||
//
|
||||
// 注意：本词库仅为 Phase 1 基础覆盖，Phase 2 将接入服务端 AI + 可更新词库。
|
||||
|
||||
/// 敏感词分类
|
||||
enum SensitiveCategory {
|
||||
violence('暴力'),
|
||||
sexual('色情'),
|
||||
bullying('欺凌'),
|
||||
drugs('毒品'),
|
||||
gambling('赌博'),
|
||||
politics('政治敏感'),
|
||||
fraud('诈骗'),
|
||||
profanity('粗口');
|
||||
|
||||
const SensitiveCategory(this.label);
|
||||
final String label;
|
||||
}
|
||||
|
||||
/// ============================================================
|
||||
/// 各分类敏感词
|
||||
/// ============================================================
|
||||
|
||||
/// 暴力类
|
||||
const _violenceWords = [
|
||||
// 直接暴力
|
||||
'杀人', '砍人', '捅人', '打死', '打死你', '弄死', '弄死你',
|
||||
'揍你', '揍死', '打死他', '砍死', '捅死',
|
||||
'杀了他', '打死他', '砍了他', '捅了他',
|
||||
'去死', '你去死', '怎么不去死',
|
||||
'割腕', '割脖子', '跳楼', '上吊',
|
||||
// 武器
|
||||
'炸弹', '手枪', '步枪', '子弹', '刀杀',
|
||||
// 自残/伤害暗示
|
||||
'自杀', '自残', '不想活',
|
||||
];
|
||||
|
||||
/// 色情类
|
||||
const _sexualWords = [
|
||||
'色情', '裸体', '裸照', '黄色', '黄片',
|
||||
'做爱', '性行为', '性交', '强奸', '强暴',
|
||||
'猥亵', '性骚扰', '偷拍',
|
||||
'发情', '骚货', '贱人',
|
||||
];
|
||||
|
||||
/// 欺凌类
|
||||
const _bullyingWords = [
|
||||
'废物', '垃圾', '蠢货', '白痴', '弱智',
|
||||
'傻子', '笨蛋', '猪头', '丑八怪',
|
||||
'滚开', '滚蛋', '闭嘴', '别烦我',
|
||||
'讨厌鬼', '没人要', '没朋友',
|
||||
'不和你玩', '不要和你玩',
|
||||
'大家不要理', '孤立',
|
||||
'偷东西', '小偷',
|
||||
];
|
||||
|
||||
/// 毒品类
|
||||
const _drugsWords = [
|
||||
'毒品', '吸毒', '贩毒', '大麻', '海洛因',
|
||||
'冰毒', '摇头丸', '可卡因', '吗啡',
|
||||
'鸦片', 'K粉', '安非他命',
|
||||
'上瘾', '毒瘾',
|
||||
];
|
||||
|
||||
/// 赌博类
|
||||
const _gamblingWords = [
|
||||
'赌博', '赌钱', '下注', '押注', '赌场',
|
||||
'买彩票', '时时彩', '六合彩',
|
||||
'百家乐', '老虎机', '扑克赌',
|
||||
'赌债', '借钱赌',
|
||||
];
|
||||
|
||||
/// 政治敏感类
|
||||
const _politicsWords = [
|
||||
'反动', '颠覆', '分裂', '暴动', '造反',
|
||||
'推翻', '政变', '游行示威',
|
||||
];
|
||||
|
||||
/// 诈骗类
|
||||
const _fraudWords = [
|
||||
'诈骗', '骗钱', '骗密码', '骗账号',
|
||||
'中奖了', '恭喜中奖', '免费领取',
|
||||
'点击链接领奖', '转账给我',
|
||||
'刷单', '兼职刷单', '高薪兼职',
|
||||
'传销', '拉人头',
|
||||
];
|
||||
|
||||
/// 粗口类
|
||||
const _profanityWords = [
|
||||
'操你', '妈的', '他妈', '去你的', '狗屎',
|
||||
'滚', '屁', '放屁', '扯淡', '王八蛋',
|
||||
'混蛋', '靠', '我去', '卧槽',
|
||||
'我靠', '我擦',
|
||||
];
|
||||
|
||||
/// 全量词库：分类 → 词列表
|
||||
const Map<SensitiveCategory, List<String>> kSensitiveWords = {
|
||||
SensitiveCategory.violence: _violenceWords,
|
||||
SensitiveCategory.sexual: _sexualWords,
|
||||
SensitiveCategory.bullying: _bullyingWords,
|
||||
SensitiveCategory.drugs: _drugsWords,
|
||||
SensitiveCategory.gambling: _gamblingWords,
|
||||
SensitiveCategory.politics: _politicsWords,
|
||||
SensitiveCategory.fraud: _fraudWords,
|
||||
SensitiveCategory.profanity: _profanityWords,
|
||||
};
|
||||
|
||||
/// ============================================================
|
||||
/// 谐音/形近/数字变体映射
|
||||
/// ============================================================
|
||||
|
||||
/// 原词 → 变体列表
|
||||
///
|
||||
/// 变体检测在预处理后的文本上运行，可以捕获常见的绕过手法：
|
||||
/// - 数字谐音: "死" → "4"
|
||||
/// - 形近替换: "傻" → "纱"
|
||||
/// - 拼音缩写: "牛逼" → "nb"
|
||||
const Map<String, List<String>> kHomophoneVariants = {
|
||||
// 暴力相关
|
||||
'死': ['4', '④', '亖', '☠'],
|
||||
'杀': ['莎', '纱', '沙'],
|
||||
'砍': ['砍人'],
|
||||
'捅': ['捅人'],
|
||||
// 欺凌相关
|
||||
'傻': ['纱', '沙', '啥'],
|
||||
'笨': [], // 无实际变体
|
||||
'蠢': ['春'],
|
||||
'废物': ['费物', '废无'],
|
||||
'垃圾': ['拉吉', '垃 圾'],
|
||||
// 粗口相关
|
||||
'操': ['草', '艹', '槽'],
|
||||
'卧槽': ['我槽', '我草', 'wc', 'WC', 'Wc'],
|
||||
'我靠': ['我 k', '我K'],
|
||||
// 欺凌
|
||||
'滚': ['衮'],
|
||||
'屁': ['辟'],
|
||||
};
|
||||
Reference in New Issue
Block a user