[Paper Review] Abnormal Trading Detection in the NFT Market

Mingxiao Song, Yangming Liu+2 more|arXiv (Cornell University)|May 25, 2023

Financial Markets and Investment Strategies4 citations

TL;DR

This paper proposes an unsupervised machine learning approach to detect abnormal trading behaviors—particularly wash trading—in the NFT market using K-means clustering on network, monetary, and temporal features derived from on-chain data. It identifies 5.38% of NFT traders as potential wash traders, offering regulators and platforms a data-driven framework to target fraudulent activity.

ABSTRACT

The Non-Fungible-Token (NFT) market has experienced explosive growth in recent years. According to DappRadar, the total transaction volume on OpenSea, the largest NFT marketplace, reached 34.7 billion dollars in February 2023. However, the NFT market is mostly unregulated and there are significant concerns about money laundering, fraud and wash trading. The lack of industry-wide regulations, and the fact that amateur traders and retail investors comprise a significant fraction of the NFT market, make this market particularly vulnerable to fraudulent activities. Therefore it is essential to investigate and highlight the relevant risks involved in NFT trading. In this paper, we attempted to uncover common fraudulent behaviors such as wash trading that could mislead other traders. Using market data, we designed quantitative features from the network, monetary, and temporal perspectives that were fed into K-means clustering unsupervised learning algorithm to sort traders into groups. Lastly, we discussed the clustering results' significance and how regulations can reduce undesired behaviors. Our work can potentially help regulators narrow down their search space for bad actors in the market as well as provide insights for amateur traders to protect themselves from unforeseen frauds.

Motivation & Objective

• To identify abnormal trading behaviors such as wash trading in the unregulated NFT market using data-driven methods.
• To develop quantitative features from network, monetary, and temporal dimensions to characterize trader behavior.
• To apply unsupervised K-means clustering to group traders based on behavioral similarity without labeled data.
• To provide actionable insights for regulators and marketplaces to detect and mitigate fraudulent activities.
• To establish a foundation for future research on NFT market integrity and anomaly detection.

Proposed method

• Collected on-chain transaction data from the Ethereum blockchain for NFT trading activity.
• Engineered three types of features: network (in-degree, out-degree, unique in/out), monetary (transfer volume, price volatility), and temporal (transaction frequency, time intervals).
• Applied K-means clustering with multiple cluster configurations to identify distinct behavioral groups.
• Validated cluster stability by analyzing persistent groupings across different k-values, especially identifying consistently grouped outlier accounts.
• Used Etherscan to manually verify the identities and behaviors of highly anomalous accounts in the outlier cluster.
• Evaluated clustering results by analyzing behavioral patterns, transaction volume, and temporal dynamics within each cluster.

Experimental results

Research questions

• RQ1 What behavioral patterns distinguish wash traders from legitimate NFT traders in the absence of labeled data?
• RQ2 How can network, monetary, and temporal features be combined to detect abnormal trading behavior in NFT markets?
• RQ3 What proportion of NFT traders exhibit characteristics consistent with wash trading?
• RQ4 Can unsupervised clustering effectively group traders into meaningful behavioral categories, including potential wash traders?
• RQ5 How can regulators and platforms use feature-based clustering to prioritize monitoring and enforcement?

Key findings

• The study identified seven distinct clusters of NFT traders, including collectors, speculators, and two potential wash trader groups.
• A total of 13,607 accounts—representing 5.38% of the analyzed traders—were classified as potential wash traders based on high in/out-degree, frequent internal transfers, and stable market timing.
• Seven institutional and marketplace accounts consistently formed an outlier cluster due to extreme network activity, including high in-degree, out-degree, and transaction volume.
• Wash traders in Class 3 exhibited high transfer ratios and frequent intra-group trading, indicating coordinated manipulation.
• Wash traders in Class 5 showed high-volume buying during mature collection phases, suggesting strategic timing to inflate perceived value.
• Collectors (Class 0) were characterized by high-value, low-frequency trades with long holding periods and higher-than-average selling prices.