Cryptocurrency Pump-and-Dump Scheme Detection System

Overview

This project aims to address one of the most pressing challenges in the cryptocurrency market: pump-and-dump schemes. These fraudulent activities manipulate cryptocurrency prices, causing significant financial losses for investors. Our solution leverages machine learning and real-time data analysis to detect and predict these schemes before they reach their devastating conclusion.

---

Team Members

• Javad
• Khalil
• Aashna
• Phuc

We are Georgia Tech Computer Science students :)

---

Problem Statement

Cryptocurrency markets are highly susceptible to manipulation, with 25% of all new crypto tokens in 2022 created specifically for pump-and-dump schemes. These schemes exploit psychological factors like FOMO (Fear of Missing Out), fake social proof, and greed, leading to massive financial losses for investors. Recent examples, such as the Argentina cryptocurrency scandal, highlight the urgent need for a solution.

---

Solution

We developed a neural network-based system that detects pump-and-dump schemes by analyzing social media activity and market capitalization data. Our system identifies suspicious patterns and provides early warnings to regulators or investors.

Key Features

1. Real-Time Social Media Analysis:

   • Utilizes the LunarCrush API to monitor platforms like Reddit, X (Twitter), YouTube, and TikTok.
   • Tracks engagement metrics, post frequency, and sentiment analysis.

2. Market Data Integration:

   • Correlates social media data with CoinMarketCap market capitalization data.
   • Identifies unusual price movements and trading volumes.

3. Neural Network Model:

   • Trained on historical pump-and-dump cases to recognize patterns.
   • Detects subtle indicators of manipulation, such as coordinated social media campaigns and sudden price spikes.

4. Real-World Application:

   • Successfully identified the Valor token as a potential pump-and-dump scheme days before significant price movements.
   • Provides actionable insights for regulatory authorities or investors.

---

Technical Implementation

Data Sources

1. LunarCrush API:

   • Provides social media engagement metrics for cryptocurrencies.
   • Tracks mentions, sentiment, and influencer activity.

2. CoinMarketCap API:

   • Supplies real-time market data, including price, volume, and market capitalization.

Neural Network Architecture

• Input Layer:

  • Social media engagement metrics (e.g., post frequency, sentiment score).
  • Market data (e.g., price, volume, market cap).

• Hidden Layers:

  • Multiple dense layers with ReLU activation for feature extraction.
  • Dropout layers to prevent overfitting.

• Output Layer:

  • Binary classification (pump-and-dump scheme or legitimate activity).
  • Sigmoid activation function for probability output.

• Training:

  • Dataset: Historical data of known pump-and-dump schemes.
  • Loss Function: Binary cross-entropy.
  • Optimizer: Adam.

System Workflow

1. Data Collection:

   • Fetch social media and market data using APIs.

2. Preprocessing:

   • Normalize and clean data.
   • Extract relevant features (e.g., sentiment score, price change rate).

3. Prediction:

   • Feed processed data into the neural network.
   • Generate predictions and confidence scores.

4. Alert System:

   • Trigger alerts for suspicious activity.
   • Provide detailed reports for regulators or investors.

---

Potential Applications

1. Regulatory Use:

   • Alert authorities about potential fraud, insider trading, or money laundering.
   • Help maintain market integrity and protect investors.

2. Investment Strategy:

   • Enable short-selling opportunities before market crashes.
   • Provide insights for risk-averse investors.

---

Future Work

• Expand data sources to include additional platforms (e.g., Telegram, Discord).
• Improve model accuracy by incorporating more historical data.
• Develop a user-friendly dashboard for real-time monitoring.

---

How to Use

1. Clone the repository:

   git clone https://github.com/yourusername/crypto-pump-dump-detection.git

2. Install dependencies

3. Run the system:

   python app.py
   streamlist run app.py

---

Contributing

We welcome contributions! Please fork the repository and submit a pull request with your changes.

---

Contact

For questions or collaborations, please contact:
Email: kzina6@gatech.edu
GitHub: https://github.com/fantabnina