s1366560/anti-corruption-v2

Anti-Corruption Investigation v2 (Enhanced)

Overview

This enhanced version of the anti-corruption investigation tool provides comprehensive analysis of chat logs and communications to detect corruption patterns and build detailed relationship networks.

Features

Core Analysis (from v1)

• Pattern matching for corruption indicators
• Financial corruption detection
• Power abuse identification
• Secret meeting detection
• Collusion pattern recognition

New in v2 - Enhanced Relationship Analysis

1. Multi-hop Relationship Detection (多跳关系检测)

Find indirect connections between persons through 2-3 hops, revealing hidden relationships that are not directly visible.

2. Relationship Evolution Tracking (关系演变追踪)

Track how relationships develop over time, identifying escalation patterns and changes in communication intensity.

3. Power Structure Analysis (权力结构分析)

Identify hierarchy and power centers within the network, mapping official-business relationships and influence patterns.

4. Collusion Ring Detection (串通团伙检测)

Find closed groups (triangles) of colluding individuals with high mutual suspicious activity.

5. Timeline Analysis (时间线分析)

Analyze temporal patterns of suspicious activities, identifying peak activity periods and suspicious time patterns.

6. Money Flow Tracing (资金流向追踪)

Track financial transaction patterns mentioned in communications, identifying key money handlers and transaction flows.

7. Enhanced Person Profiles

Include network metrics:

• Centrality score (中心性分数)
• Betweenness score (桥梁性分数)
• Influence score (影响力分数)
• Risk score with multi-factor calculation

Usage

Command Line Interface

# Basic corruption pattern analysis
python anti_corruption_v2.py analyze data.jsonl report.json

# Enhanced social network analysis
python anti_corruption_v2.py social-network data.jsonl social_network.json --text-report report.txt

# Timeline analysis
python anti_corruption_v2.py timeline data.jsonl timeline.json --text-report timeline_report.txt

# Money flow analysis
python anti_corruption_v2.py money-flow data.jsonl money_flows.json --text-report money_report.txt

# Full comprehensive analysis
python anti_corruption_v2.py full data.jsonl output/

Python API

from anti_corruption_v2 import EnhancedSocialNetworkAnalyzer, TimelineAnalyzer, MoneyFlowAnalyzer

# Load messages
messages = [...]  # List of message dictionaries

# Social network analysis
analyzer = EnhancedSocialNetworkAnalyzer(messages)
results = analyzer.analyze()

# Access results
profiles = results['person_profiles']
intermediaries = results['intermediaries']
communities = results['communities']
collusion_rings = results['collusion_rings']
money_flows = results['money_flows']

Input Format

JSONL Format

{"timestamp": "2024-01-15 14:30:00", "sender": "张三", "receiver": "李四", "content": "那笔钱已经准备好了"}
{"timestamp": "2024-01-15 14:32:00", "sender": "李四", "receiver": "张三", "content": "好的，老地方见"}

TXT Format

[2024-01-15 14:30:00] 张三 -> 李四: 那笔钱已经准备好了
[2024-01-15 14:32:00] 李四 -> 张三: 好的，老地方见

Output Structure

Social Network Analysis

{
  "person_profiles": {
    "张三": {
      "name": "张三",
      "message_count": 150,
      "contact_count": 8,
      "risk_score": 7.5,
      "risk_level": "🔴 高风险",
      "influence_score": 8.2,
      "centrality_score": 0.75,
      "betweenness_score": 0.45,
      "primary_role": "official",
      "corruption_patterns": {
        "financial_corruption": 12,
        "power_abuse": 8
      }
    }
  },
  "network_statistics": {
    "total_persons": 25,
    "total_relationships": 68,
    "network_density": 0.23,
    "risk_distribution": {"high": 5, "medium": 8, "low": 12}
  },
  "intermediaries": [...],
  "communities": [...],
  "influence_ranking": [...],
  "multi_hop_relationships": [...],
  "relationship_evolution": [...],
  "power_structure": {...},
  "collusion_rings": [...],
  "timeline_events": [...],
  "money_flows": [...]
}

Detection Patterns

Financial Corruption (资金往来)

• Direct: 转账、汇款、账户、资金、钱款、回扣、贿赂、好处费
• Semantic: 东西准备好了、表示一下、心意、感谢费

Power Abuse (权力滥用)

• Direct: 特殊照顾、通融一下、开绿灯、违规操作、打招呼
• Semantic: 帮忙看看、关照一下、特事特办、按惯例

Secret Meetings (秘密会面)

• Direct: 老地方、私下见面、秘密会面、不要告诉别人
• Semantic: 见面聊、当面谈、出来坐坐、不方便在这里说

Collusion (串通勾结)

• Direct: 统一口径、对好供词、串通、删除记录、销毁证据
• Semantic: 保持一致、统一说法、删除吧、别留记录

Money Laundering (洗钱)

• Direct: 洗白、过账、走账、开票、公户、私户

Role Detection

• official (官员/公务员): 局长、处长、科长、领导、干部、审批
• business (商人/企业主): 老板、经理、董事长、公司、企业、项目
• intermediary (中介/掮客): 中介、介绍人、牵线、搭桥、有关系
• family (家属/亲戚): 老婆、丈夫、父亲、母亲、亲戚

Risk Scoring

Risk Levels

• 🔴 High Risk (高风险): Score >= 6
• 🟠 Medium Risk (中风险): Score >= 3
• 🟢 Low Risk (低风险): Score < 3

Risk Factors

1. Suspicious message ratio (30%)
2. Pattern diversity (15%)
3. Late night activity (15%)
4. Network position - centrality (10%)
5. Network position - betweenness (15%)
6. Role combinations - official + business (15%)

Example Workflow

# 1. Run full analysis
python anti_corruption_v2.py full example_data.jsonl ./output/

# 2. Check generated reports
cat output/social_network_report.txt
cat output/timeline_report.txt
cat output/money_flow_report.txt

# 3. Analyze specific aspects
python anti_corruption_v2.py social-network example_data.jsonl network.json --text-report network.txt

Requirements

• Python 3.8+
• Standard library only (no external dependencies)

Files

• anti_corruption_v2.py - Main analysis tool
• SKILL.md - This documentation
• example_data.jsonl - Sample data for testing

Version History

• v2.0.0 (Current): Enhanced relationship analysis with multi-hop detection, evolution tracking, power structure analysis, collusion ring detection, timeline analysis, money flow tracing
• v1.0.0: Basic corruption pattern detection