IH-Korupsi: Anti-Corruption Data Forensic Toolkit

IH-Korupsi (short for Intelligent Hunting (IH) - Korupsi) is an open-source Python toolkit designed to detect financial anomalies and potential corruption using pure mathematical methods without AI/Machine Learning.

Developed by OurCreativity Edisi Coding to support transparency and financial accountability worldwide.

---

Why IH-Korupsi?

Corruption harms economies and societies. IH-Korupsi provides an educational and professional tool that can be used by:

• Government internal auditors
• Investigative journalists
• Anti-corruption researchers
• Students of data forensics
• Transparency NGOs
• Anyone committed to fighting corruption

Core Principles

1. Transparency: Every anomaly detection can be explained through mathematical formulas.
2. Auditable: No "black box" algorithms—everything is open and deterministic.
3. Zero AI Dependency: Pure statistics and mathematics to ensure results are legally defensible.
4. Open Source: Free to use, study, and improve.

---

Key Features

1. The Mathematician (Statistical Detection)

Benford's Law

Detects manipulation in financial reports. Benford’s Law states that in natural financial data, the first leading digit follows a specific logarithmic distribution. If someone fabricates numbers, the distribution will likely deviate.

Case Example: Financial reports that are manipulated tend to have an unusual spike in numbers starting with digits like 5, 6, 7, or 8.

Relative Size Factor (RSF)

Identifies unusual transactions for a specific entity. RSF compares an entity's largest transaction to the average of its other transactions.

Formula: RSF = Largest Transaction / Average of Other Transactions

Case Example: A vendor that usually receives $500–$1,000 suddenly gets a contract for $50,000.

Z-Score & IQR

Standard statistical methods to find extreme outliers in transaction data.

---

2. The Connector (Network Analysis)

Circular Trading Detection

Finds funds that return to the original sender through multiple intermediaries.

Case Example:

Department A → Vendor X → Sub-vendor Y → Consultant Z → Department A

This pattern is often used for price mark-ups or money laundering loops.

Centrality Analysis

Finds hidden key actors in a network using algorithms like PageRank and Betweenness Centrality.

---

3. The Chronologist (Time-Series Analysis)

Fiscal Cliff Dumping

Detects unusual spending spikes during the final month of the fiscal year (budget dumping).

Indicator: Ratio of December spending vs. monthly average > 2.5x.

Velocity Check

Detects inhuman transaction frequencies within a short period.

Case Example: 50 transactions in a single day for the same vendor.

---

4. String Detective (Name Duplication)

Fuzzy Name Matching

Identifies "Ghost Vendors"—entities with slightly different names but likely the same identity.

Examples:

• Global Solutions Corp vs Global Solutions Corp (double space)
• Smith & Sons Ltd vs Smith and Sons Ltd

Uses the Levenshtein Distance algorithm without external NLP libraries.

---

Installation

Prerequisites

• Python 3.10 or newer
• Windows / Linux / MacOS

Setup

1. Clone this repository:

git clone https://github.com/[username]/ih-korupsi.git
cd ih-korupsi

2. Install dependencies:

pip install -r requirements.txt

---

Usage

Mode 1: Run with Sample Data (For Learning)

Run the toolkit with synthetic data containing injected anomalies:

python main.py --type sample --output sample_report.json --html visual_report.html

The system will:

1. Generate 500 rows of synthetic transaction data.
2. Inject intentional anomalies (Benford, RSF, Fiscal Cliff).
3. Execute all detection modules.
4. Generate a JSON data report and a beautiful HTML visual report.

Mode 2: Analyze Your Own Data

CSV Format

Your CSV must include at least these columns:

• amount: Transaction value
• vendor_name: Vendor or recipient name
• vendor_id: Unique vendor ID
• date: Transaction date (YYYY-MM-DD)
• sender_id: Sender identifier
• receiver_id: Receiver identifier

Example:

python main.py --input my_data.csv --type csv --output my_results.json --html report.html

---

Understanding the Report

The JSON report is structured as follows:

{
  "metadata": {
    "total_rows": 500,
    "total_amount": 300000000,
    "currency": "IDR"
  },
  "findings": {
    "The Mathematician": { ... },
    "The Connector": { ... },
    "The Chronologist": { ... },
    "String Detective": { ... }
  }
}

Interpreting Findings

Benford's Law (MAD Score)

• MAD < 0.006: High Conformity (Normal)
• MAD 0.006–0.012: Acceptable
• MAD 0.012–0.015: Marginal (Needs attention)
• MAD > 0.015: Non-conformity (Red flag!)

RSF (Relative Size Factor)

• RSF < 5: Normal
• RSF 5–10: Needs verification
• RSF > 10: Highly suspicious

---

Important Notes & Limitations

1. Not Legal Proof: IH-Korupsi provides indicators only. Anomalies do not automatically mean corruption; they require further investigation.
2. Context Matters: Some anomalies can be legitimate (e.g., a massive infrastructure project causing a high RSF).
3. Data Quality: Garbage in, garbage out. Ensure your input data is clean.
4. Ethical Use: This toolkit is for transparency and education. Please use it responsibly.

---

About OurCreativity Edisi Coding

We are a developer community that believes technology can be a force for social good. IH-Korupsi is one of our efforts toward a more transparent and accountable future.

Slogan: "Code for Justice, Data for Transparency"

---

License

This project is licensed under the MIT License.

---

Join us in building a cleaner, more transparent world!