Detecting Sybil Addresses in Blockchain Airdrops: A Subgraph-based Feature Propagation and Fusion Approach

1 st Qiangqiang Liu, 2 st Qian Huang, 3 st Frank Fan, 4 st Haishan Wu, 5 st Xueyan Tang · 2025

Read Detecting Sybil Addresses in Blockchain Airdrops: A Subgraph-based Feature Propagation and Fusion Approach by 1 st Qiangqiang Liu, 2 st Qian Huang, 3 st Frank Fan, 4 st Haishan Wu, 5 st Xueyan Tang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-09313.txt.

Introduction

Qiangqiang Liu and co-authors present a method for spotting fake accounts in blockchain token giveaways, in a research paper from 2025. In a Sybil attack, one actor runs many addresses to claim extra tokens. The team aims to catch those addresses from their transaction patterns.

For each address, the method builds a two-layer graph of its transactions. It then pulls out timing traits from the address's life, such as its first trade and its first fuel payment, plus amount and network traits. A learning model merges these traits into one judgment. On a set of 193,701 addresses, with 23,240 of them fake, the method beat rival tools on all main scores, each above 0.9.

Catalog record

CREATOR1 st Qiangqiang Liu, 2 st Qian Huang, 3 st Frank Fan, 4 st Haishan Wu, 5 st Xueyan Tang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-09313

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Detecting Sybil Addresses in Blockchain Airdrops: A Subgraph-based Feature Propagation and Fusion Approach 1 st Qiangqiang Liu, 2 st Qian Huang, 3 st Frank Fan, 4 st Haishan Wu, 5 st Xueyan Tang Abstract Sybil attacks pose a significant security threat to blockchain ecosystems, particularly in token airdrop events. This paper proposes a novel sybil address identification method based on subgraph feature extraction lightGBM. The method first constructs a two-layer deep transaction subgraph for each address, then extracts key event operation features according to the lifecycle of sybil addresses, including the time of first transaction, first gas acquisition, participation in airdrop activities, and last transaction. These temporal features effectively capture the consistency of sybil address behavior operations. Additionally, the method extracts amount and network structure features, comprehensively describing address behavior patterns and network topology through feature propagation and fusion. Experiments conducted on a dataset containing 193,701 addresses (including 23,240 sybil addresses) show that this method outperforms existing approaches in terms of precision, recall, F1 score, and AUC, with all metrics exceeding 0.9. The methods and results of this study can be further applied to broader blockchain security areas such as transaction manipulation identification and token liquidity risk assessment, contributing to the construction of a more secure and fair blockchain ecosystem. I Introduction Sybil attacks are a security threat in decentralized networks where attackers create multiple identities or nodes to influence system decision-making processes or consensus mechanisms. In computer networks and blockchain technology, the purpose of sybil attacks is typically to gain disproportionate influence or control, such as in voting systems, reputation systems, or airdrop events. In the cryptocurrency domain, sybil addresses refer to multiple addresses controlled by a single entity, used to masquerade as multiple independent participants to manipulate markets, abuse airdrop events, or influence decisions in decentralized autonomous organizations (DAOs). The existence of sybil addresses undermines the principle of decentralization in blockchain technology and compromises the fairness and trustworthiness of the system. Airdrops are a common cryptocurrency marketing strategy where projects distribute tokens to a wide range of users to increase their project’s visibility and user base. However, airdrop events are also susceptible to sybil attacks, where attackers create multiple addresses to collect more airdrop tokens, unfairly increasing their gains. This behavior not only infringes on the rights of other users but also potentially damages the project’s reputation and market value. Sybil addresses have garnered increasing attention in various blockchain projects (e.g., Starknet, zkSync, LayerZero), leading to the accumulation of labeled sybil address datasets. Machine learning models and graph neural networks [ 7 , 21 ] have demonstrated exceptional efficacy in identifying entity addresses and fraudulent addresses in blockchain systems. However, academic literature specifically addressing sybil address identification remains relatively scarce, with only two directly relevant works employing unsupervised methods for this purpose. Airdrop events can be categorized into short-term and long-term distributions: (1) Short-term airdrops: These involve one-time token distributions by project teams on a specific dataset. (2) Long-term airdrops: These encompass prolonged token distributions on a specific dataset, exemplified by Soulbound Token (SBT) airdrops or projects conducting multiple-phase airdrops or extended marketing campaigns. These distinct categories necessitate different analytical approaches. For short-term airdrops, unsupervised methods are more appropriate, facilitating the exploration and identification of sybil address behavioral patterns. Conversely, long-term airdrops are better suited to supervised methods, as the extended distribution period allows for the accumulation of a substantial corpus of labeled sybil addresses, enabling more precise supervised model training and subsequent identification. As we amass diverse sybil address datasets across various platforms, we gain insights into a wide spectrum of sybil behavioral patterns. To achieve optimal identification efficacy, supervised methodologies emerge as the most promising approach. This strategy leverages the rich, labeled data accumulated over time, allowing for more nuanced and accurate detection of sybil addresses across different contexts and blockchain ecosystems. Inspired by Chen et al. [ 1 ] , this paper proposes a supervised subgraph-based feature extraction algorithm, subgraph-based lightGBM, to identify sybil addresses by extracting time series of critical operations and network structural features. The contributions of this study are as follows: (1) This research presents the first application of a supervised machine learning method to sybil address identification. The proposed Subgraph-based lightGBM algorithm captures structural features of airdrop address subgraphs and extracts composite temporal features based on the sybil address lifecycle (first transaction, first gas recharge, participation in airdrop events, and last transaction time). (2) We conduct network feature analysis on the transaction subgraphs of sybil addresses and analyze the importance and contribution of these features to reveal how they capture the transaction patterns of sybil addresses. (3) We validate the proposed model’s effectiveness in describing sybil address transaction patterns using a dataset containing 193,701 addresses (including 23,240 sybil addresses). Our results demonstrate that our model outperforms the existing Trusta model. The structure of this paper is organized as follows: Section II presents a comprehensive literature review, encompassing relevant works on entity identification and sybil address detection in blockchain systems. Section III introduces our proposed Subgraph-based lightGBM algorithm, detailing its methodology and implementation. Section IV discusses our experimental results, including feature analysis and feature interpretability, demonstrating the efficacy of our approach. Section V addresses the limitations of our study, providing a critical evaluation of our methodology and results. Finally, Section VI concludes the paper with a summary of our findings and offers perspectives on future research directions in this domain. II Related Works II-A Entity Identification Entity identification, which associates multiple account addresses belonging to the same entity, has become a crucial step in revealing the real identities behind blockchain accounts and provides an important technical basis for sybil address detection. Existing entity identification methods are mainly divided into three categories: transaction-based, behavior-based, and off-chain information-based, each with its own characteristics, aiming to parse blockchain data from different angles to reveal entity relationships behind complex transaction networks. Wu et al. [ 2 ] and Xu et al. [ 3 ] used graph neural networks and graph attention networks, respectively, to improve identification accuracy. Cross-chain analysis has also made progress, with Victor and Lüders [ 4 ] proposing a cross-chain entity association method based on address reuse patterns, and Zhang et al. [ 5 ] developing a multi-modal learning framework. Jourdan et al. [ 6 ] studied the use of machine learning techniques to track cross-chain transactions. The privacy-preserving entity identification framework based on federated learning proposed by Li et al. [ 7 ] , and the privacy-preserving anti-money laundering analysis method discussed by Biryukov and Tikhomirov [ 8 ] , both demonstrate the balance between privacy and security. The research by Tzanetakis et al. [ 9 ] showcases the potential application of entity identification in regulating illegal markets. Overall, blockchain entity identification research has developed to a stage capable of handling complex network structures, conducting cross-chain analysis, and considering privacy protection. However, sybil address identification has its own specificity, focusing more on identifying malicious actors, requiring stricter judgment criteria and higher identification accuracy. Sybil addresses typically exhibit more consistent or coordinated behavior patterns and may appear concentrated within specific time periods. Therefore, when applied to sybil address identification, special attention needs to be paid to these characteristics and the constantly changing strategies

Read the complete text