[Paper Review] A Deep Dive into NFT Rug Pulls

Jintao Huang, Ningyu He+3 more|arXiv (Cornell University)|May 10, 2023

Software Engineering Research4 citations

TL;DR

This paper presents the first comprehensive study of NFT rug pulls, introducing a rule-based detection system that identifies 7,487 confirmed rug pull projects—30× larger than existing datasets—and a predictive model that raises alarms for 90% of scams up to 96 hours in advance, enabling early mitigation of financial losses in the NFT ecosystem.

ABSTRACT

NFT rug pull is one of the most prominent type of scam that the developers of a project abandon it and then run away with investors' funds. Although they have drawn attention from our community, to the best of our knowledge, the NFT rug pulls have not been systematically explored. To fill the void, this paper presents the first in-depth study of NFT rug pulls. Specifically, we first compile a list of 253 known NFT rug pulls as our initial ground truth, based on which we perform a pilot study, highlighting the key symptoms of NFT rug pulls. Then, we enforce a strict rule-based method to flag more rug pulled NFT projects in the wild, and have labelled 7,487 NFT rug pulls as our extended ground truth. Atop it, we have investigated the art of NFT rug pulls, with kinds of tricks including explicit ones that are embedded with backdoors, and implicit ones that manipulate the market. To release the expansion of the scam, we further design a prediction model to proactively identify the potential rug pull projects in an early stage ahead of the scam happens. We have implemented a prototype system deployed in the real-world setting for over 5 months. Our system has raised alarms for 7,821 NFT projects, by the time of this writing, which can work as a whistle blower that pinpoints rug pull scams timely, thus mitigating the impacts.

Motivation & Objective

• To systematically investigate NFT rug pulls, a major scam type in the NFT ecosystem that has so far lacked comprehensive analysis.
• To measure the true prevalence of NFT rug pulls by building a reliable, large-scale dataset using rule-based detection on on-chain and off-chain data.
• To uncover the sophisticated techniques used in rug pulls, including backdoor exploitation and market manipulation, beyond basic pump-and-dump schemes.
• To develop a real-time prediction model that proactively flags potential rug pull projects before funds are stolen, enabling early intervention.
• To deploy a prototype warning system in the wild that has successfully raised alarms for 7,821 NFT projects, with high confirmation rates.

Proposed method

• Constructed an initial ground truth of 253 known NFT rug pulls through community-sourced data for pilot symptom analysis.
• Designed a strict rule-based detector using 14 key symptoms (e.g., sudden token transfers, abnormal minting patterns) to flag 7,487 additional rug pull projects across 173K Ethereum NFT projects.
• Categorized and analyzed 8 distinct scam tricks, including explicit backdoor exploitation and implicit market manipulation, using custom analyzers on smart contract and transaction data.
• Built a machine learning prediction model using 73 on-chain and off-chain features (e.g., transaction frequency, wallet behavior, social media activity) to detect early warning signs.
• Deployed a real-time prototype system on Ethereum since November 2022, continuously monitoring and raising alarms for suspicious projects.
• Validated results through cross-referencing with external evidence and confirmed post-hoc for 7,821 projects, exceeding the extended ground truth.

Figure 1: The activity of AniMoon on OpenSea.

Experimental results

Research questions

• RQ1 What are the key on-chain and off-chain symptoms that characterize NFT rug pulls?
• RQ2 How prevalent are NFT rug pulls in the Ethereum NFT ecosystem, and what is the true scale of such scams?
• RQ3 What sophisticated tricks do rug pull perpetrators use beyond simple pump-and-dump schemes?
• RQ4 Can early-stage signals be used to predict rug pull events before they occur?
• RQ5 How effective is a real-time warning system in identifying and mitigating NFT rug pulls in practice?

Key findings

• The study identified 7,487 NFT rug pull projects using a rule-based method, representing a 30× increase over existing community-maintained datasets.
• 84% of the 7,487 rug pull projects used at least one sophisticated trick, such as hidden backdoors or market manipulation, beyond basic pump-and-dump behavior.
• The prediction model successfully raised alarms for 90% of NFT scam events at least 96 hours before the rug pull occurred, demonstrating strong early detection capability.
• The deployed prototype system has raised warnings for 7,821 NFT projects in real-world operation, with most confirmed as scams through subsequent evidence.
• The study revealed that 84% of rug pulls involved deliberate technical or market manipulation, indicating a high level of premeditation and sophistication.
• The dataset of 7,487 confirmed rug pulls is the largest publicly available benchmark for future research on NFT scams and security.

Figure 2: The whole process of AniMoon rug pull.