[Paper Review] AI-powered Fraud Detection in Decentralized Finance: A Project Life Cycle Perspective

Bingqiao Luo, Zhang Zhen+4 more|arXiv (Cornell University)|Aug 30, 2023

Blockchain Technology Applications and Security4 citations

TL;DR

This paper proposes a project life cycle-based taxonomy of DeFi frauds—spanning development, introduction, growth, maturity, and decline—enabling stage-specific AI-powered fraud detection. It identifies tree-based and graph-based models as most effective, while advocating for real-time analysis, cross-chain data, and privacy-preserving AI to enhance detection in evolving DeFi ecosystems.

ABSTRACT

In recent years, blockchain technology has introduced decentralized finance (DeFi) as an alternative to traditional financial systems. DeFi aims to create a transparent and efficient financial ecosystem using smart contracts and emerging decentralized applications. However, the growing popularity of DeFi has made it a target for fraudulent activities, resulting in losses of billions of dollars due to various types of frauds. To address these issues, researchers have explored the potential of artificial intelligence (AI) approaches to detect such fraudulent activities. Yet, there is a lack of a systematic survey to organize and summarize those existing works and to identify the future research opportunities. In this survey, we provide a systematic taxonomy of various frauds in the DeFi ecosystem, categorized by the different stages of a DeFi project's life cycle: project development, introduction, growth, maturity, and decline. This taxonomy is based on our finding: many frauds have strong correlations in the stage of the DeFi project. According to the taxonomy, we review existing AI-powered detection methods, including statistical modeling, natural language processing and other machine learning techniques, etc. We find that fraud detection in different stages employs distinct types of methods and observe the commendable performance of tree-based and graph-related models in tackling fraud detection tasks. By analyzing the challenges and trends, we present the findings to provide proactive suggestion and guide future research in DeFi fraud detection. We believe that this survey is able to support researchers, practitioners, and regulators in establishing a secure and trustworthy DeFi ecosystem.

Motivation & Objective

• To address the lack of a systematic framework for classifying DeFi frauds across the full project life cycle.
• To analyze the correlation between fraud types and specific stages of DeFi project development.
• To review and evaluate existing AI-powered fraud detection techniques, including NLP, statistical modeling, and graph-based learning.
• To identify research gaps and propose future directions, especially in real-time detection, comprehensive data utilization, and privacy-preserving AI.
• To guide researchers, practitioners, and regulators in building a more trustworthy DeFi ecosystem.

Proposed method

• Proposes a novel taxonomy of DeFi frauds based on the Product Life Cycle Theory, categorizing frauds into five stages: development, introduction, growth, maturity, and decline.
• Reviews AI techniques across stages, including tree-based models (e.g., XGBoost), graph neural networks, NLP for whitepapers and social media, and statistical modeling.
• Analyzes on-chain transaction patterns, smart contract code, and off-chain data (e.g., social media, Telegram, Reddit) to detect fraud signals.
• Emphasizes cross-chain analysis and whale account behavior to uncover hidden fraud patterns beyond single-blockchain systems.
• Advocates for real-time fraud detection systems with early-warning capabilities using efficient, scalable AI models.
• Introduces privacy-preserving techniques such as zero-knowledge proofs to balance KYC compliance with user anonymity in fraud detection.

Figure 2. Big Events of Frauds in Recent 5 Years (Data from (Frontal, 2023 ) ).

Experimental results

Research questions

• RQ1 How do different types of DeFi frauds correlate with specific stages of a project’s life cycle?
• RQ2 Which AI techniques—such as NLP, graph-based models, or tree-based algorithms—are most effective for detecting fraud in each project stage?
• RQ3 What role do cross-chain data and off-chain behavioral signals (e.g., social media, whale accounts) play in enhancing fraud detection?
• RQ4 How can real-time and early-warning fraud detection systems be designed to proactively prevent losses in DeFi?
• RQ5 What are the challenges in balancing privacy, regulatory compliance (e.g., KYC), and effective fraud detection in decentralized systems?

Key findings

• A strong correlation exists between fraud types and stages of DeFi project life cycles, with distinct fraud patterns emerging in development, introduction, growth, maturity, and decline phases.
• Tree-based models (e.g., XGBoost) and graph-based models show commendable performance in detecting fraudulent transactions and smart contract anomalies.
• Over 16.7 billion USD in crypto assets were lost to DeFi frauds between 2011 and 2023, with 231 hacks, 135 security attacks, and 95 fraudulent schemes reported.
• 4% of crypto whales hold $25 billion and are disproportionately involved in criminal activities, highlighting the need for special account monitoring.
• Real-time fraud detection remains underdeveloped, with limited research focusing on efficiency and early detection, creating a critical gap for proactive intervention.
• Privacy-preserving AI and zero-knowledge proofs offer a viable path to reconcile regulatory needs with user privacy in DeFi, though challenges remain in implementation.

Figure 3. DeFi Frauds along Project Life Cycle.