一、实时风险检测引擎
class RiskDetectionEngine: """信件内容实时风险检测""" def __init__(self): self.models = { 'safety': SafetyModel(), 'emotion': EmotionAnalysis(), 'compliance': ComplianceChecker() } self.rules = PrisonRulesDatabase() def analyze_letter(self, letter_data): risks = [] # 1. 安全风险检测 safety_risks = self.detect_security_threats(letter_data['content']) if safety_risks: risks.extend(safety_risks) # 2. 情绪危机预警 emotion_risks = self.detect_emotional_crisis(letter_data) if emotion_risks: risks.extend(emotion_risks) # 3. 合规性检查 compliance_risks = self.check_compliance(letter_data) if compliance_risks: risks.extend(compliance_risks) # 4. 风险评估 risk_level = self.calculate_risk_level(risks) return { 'risks': risks, 'risk_level': risk_level, 'recommendations': self.generate_recommendations(risks, risk_level) } def detect_security_threats(self, content): """检测安全威胁""" threats = [] # 越狱相关关键词 escape_keywords = ['逃跑', '越狱', '挖地道', '里应外合', '工具'] for keyword in escape_keywords: if keyword in content: threats.append({ 'type': 'escape_risk', 'keyword': keyword, 'severity': 'high', 'message': '检测到越狱相关敏感内容' }) # 暴力威胁 violence_keywords = ['报复', '凶器', '伤害', '袭击'] for keyword in violence_keywords: if keyword in content: threats.append({ 'type': 'violence_risk', 'keyword': keyword, 'severity': 'medium', 'message': '检测到暴力威胁内容' }) return threats二、智能预警系统
// 2. 智能预警通知系统 class RiskAlertSystem { constructor() { this.notificationChannels = { 'high': ['sms', 'push', 'email', 'system_alert'], 'medium': ['push', 'system_alert'], 'low': ['system_alert'] }; } async sendRiskAlert(letterId, risks, riskLevel) { // 生成预警消息 const alertMessage = this.formatAlertMessage(letterId, risks, riskLevel); // 选择通知渠道 const channels = this.notificationChannels[riskLevel] || ['system_alert']; // 并行发送通知 const notifications = []; for (const channel of channels) { const notification = await this.sendViaChannel(channel, alertMessage); notifications.push({ channel, status: notification.success ? 'sent' : 'failed', timestamp: new Date().toISOString() }); } // 记录预警日志 await this.logAlert({ letterId, riskLevel, risks, notifications, timestamp: new Date().toISOString() }); return { success: true, notifications }; } formatAlertMessage(letterId, risks, riskLevel) { const riskCount = risks.length; const mainRisks = risks.slice(0, 3).map(r => r.type); return { title: `信件风险预警 - ${this.getRiskLevelText(riskLevel)}`, content: `信件ID: ${letterId}\n检测到${riskCount}个风险点\n主要风险: ${mainRisks.join(', ')}`, actions: [ { text: '查看详情', action: 'view_details' }, { text: '立即审核', action: 'review_now' }, { text: '联系家属', action: 'contact_family' } ], priority: riskLevel === 'high' ? 'urgent' : 'normal' }; } async sendViaChannel(channel, message) { switch (channel) { case 'sms': return this.sendSMSAlert(message); case 'push': return this.sendPushNotification(message); case 'email': return this.sendEmailAlert(message); case 'system_alert': return this.createSystemAlert(message); default: return { success: false }; } } }三、家属侧提醒
// 3. 家属端风险提醒 class FamilyRiskReminder { public function showRiskWarnings($letterId, $risks) { $warnings = []; // 根据风险类型生成家属端提示 foreach ($risks as $risk) { $warning = $this->formatFamilyWarning($risk); $warnings[] = $warning; } // 决定显示方式 $displayType = $this->determineDisplayType($risks); return [ 'letter_id' => $letterId, 'warnings' => $warnings, 'display_type' => $displayType, 'suggestions' => $this->getSuggestions($risks), 'can_proceed' => $this->canProceed($risks) ]; } private function formatFamilyWarning($risk) { $templates = [ 'escape_risk' => [ 'title' => '内容安全提示', 'message' => '您的内容可能涉及监狱管理规定，请修改相关表述', 'suggestion' => '请勿在信件中讨论与改造无关的内容' ], 'violence_risk' => [ 'title' => '情绪疏导建议', 'message' => '检测到情绪化表达，可能影响服刑人员心理状态', 'suggestion' => '建议使用积极鼓励的语言进行沟通' ], 'compliance_risk' => [ 'title' => '格式规范提醒', 'message' => '内容格式不符合监狱通信要求', 'suggestion' => '请按照要求重新编辑信件内容' ] ]; return $templates[$risk['type']] ?? [ 'title' => '内容提醒', 'message' => '请检查信件内容', 'suggestion' => '确保内容积极健康' ]; } private function canProceed($risks) { // 高风险内容禁止发送 $highRiskCount = count(array_filter($risks, function($r) { return $r['severity'] === 'high'; })); return $highRiskCount === 0; } }四、审核侧提醒
# 4. 审核端智能提醒 class AuditorRiskAlert: def generate_audit_alert(self, letter_data, risk_analysis): """生成审核员风险提醒""" alert = { 'letter_id': letter_data['id'], 'prison_code': letter_data['prison_code'], 'inmate_id': letter_data['inmate_id'], 'family_info': self.mask_family_info(letter_data['family_info']), # 风险概览 'risk_summary': { 'total_risks': len(risk_analysis['risks']), 'high_risk_count': self.count_by_severity(risk_analysis['risks'], 'high'), 'medium_risk_count': self.count_by_severity(risk_analysis['risks'], 'medium'), 'risk_score': risk_analysis.get('risk_score', 0) }, # 重点风险项 'key_risks': self.extract_key_risks(risk_analysis['risks']), # 审核建议 'audit_suggestions': [ self.get_suggestion_for_risk(risk) for risk in risk_analysis['risks'][:3] ], # 历史参考 'similar_cases': await self.find_similar_cases(letter_data), # 紧急程度 'urgency': self.calculate_urgency(risk_analysis), # 处置建议 'action_items': self.generate_action_items(risk_analysis) } return alert def generate_action_items(self, risk_analysis): """生成处置建议""" actions = [] if risk_analysis['risk_level'] == 'high': actions.extend([ {'action': 'immediate_review', 'priority': 1}, {'action': 'notify_prison_security', 'priority': 2}, {'action': 'contact_family', 'priority': 3} ]) elif risk_analysis['risk_level'] == 'medium': actions.extend([ {'action': 'review_within_hour', 'priority': 1}, {'action': 'add_warning_note', 'priority': 2} ]) else: actions.append({'action': 'normal_review', 'priority': 1}) return actions五、风险处理日志
-- 5. 风险处理追踪表 CREATE TABLE risk_handling_logs ( log_id CHAR(32) PRIMARY KEY, letter_id VARCHAR(32) NOT NULL, -- 风险信息 risk_types JSON COMMENT '风险类型列表', risk_level ENUM('low', 'medium', 'high') NOT NULL, risk_score DECIMAL(5,2) COMMENT '风险评分', -- 处理信息 handled_by VARCHAR(32) COMMENT '处理人', handled_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP, handling_method ENUM('auto_block', 'manual_review', 'modified_resend', 'approved') NOT NULL, -- 处置详情 action_taken TEXT COMMENT '采取的措施', review_notes TEXT COMMENT '审核备注', family_notified BOOLEAN DEFAULT FALSE COMMENT '是否通知家属', -- 结果跟踪 final_status ENUM('blocked', 'sent', 'modified_sent', 'pending') NOT NULL, follow_up_required BOOLEAN DEFAULT FALSE, follow_up_deadline DATE COMMENT '跟进截止日期', -- 审计字段 created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP, updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP, -- 索引 INDEX idx_letter (letter_id), INDEX idx_risk_level (risk_level, handled_at), INDEX idx_handler (handled_by, handled_at), INDEX idx_status (final_status, created_at), -- 外键（逻辑关联） FOREIGN KEY (letter_id) REFERENCES prison_letters(letter_id) ON DELETE CASCADE ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COMMENT='信件风险处理日志'; -- 风险预警统计视图 CREATE VIEW risk_statistics AS SELECT DATE(created_at) as stat_date, risk_level, COUNT(*) as total_cases, SUM(CASE WHEN final_status = 'blocked' THEN 1 ELSE 0 END) as blocked_count, SUM(CASE WHEN final_status = 'sent' THEN 1 ELSE 0 END) as sent_count, AVG(risk_score) as avg_risk_score, GROUP_CONCAT(DISTINCT JSON_UNQUOTE(JSON_EXTRACT(risk_types, '$[*].type'))) as risk_types FROM risk_handling_logs GROUP BY stat_date, risk_level ORDER BY stat_date DESC;总结
微爱帮监狱信件风险提醒五大核心：
实时检测- AI智能风险识别，毫秒级响应
智能预警- 分级通知，多渠道提醒
家属提醒- 温和引导，合规建议
审核提醒- 处置建议，历史参考
处理日志- 全程追踪，可回溯审计
技术特点：
三层风险分级（高/中/低）
双端提醒（家属+审核）
智能处置建议
完整风险日志
价值承诺：
用智能预警守护通信安全，用温和提醒促进积极沟通。每一份风险提醒，都是对改造的支持，对家庭的保护。
)