enes0o0/cross-chain-sybil-detection

这是一个用于区块链网络中Sybil攻击检测的大规模标注数据集，基于跨链行为建模构建。数据集包含从五个区块链网络（Ethereum、Polygon、Arbitrum、Optimism和xDai）收集的196,477个钱包地址，具有60个工程化特征和二元Sybil标签（1表示Sybil攻击者，0表示正常钱包）。该数据集旨在支持去中心化金融（DeFi）生态系统中自动Sybil钱包检测的研究。Sybil标签的真实性来源于Hop Protocol空投活动，并通过聚类分析、时间模式检查和资金来源追踪进行验证。关键统计信息包括：总钱包数196,477个，其中Sybil钱包14,195个（占7.22%），正常钱包182,282个（占92.78%），特征数60个，列数62个（包括钱包地址、Sybil标签和60个特征）。数据集呈现明显的类别不平衡（约7.22% Sybil），这反映了现实世界欺诈检测场景的典型分布。特征组织为多个组，包括桥接交易、聚合、行为/增强、派生、共时图（60秒和300秒）、资金图和实体资金集群（EFC）等，每组包含特定数量的特征，用于描述钱包的跨链行为、风险评分、活动模式、资金拓扑等。非特征列包括唯一钱包标识符和二元标签。A large-scale, labeled dataset for detecting Sybil attacks in blockchain networks using cross-chain behavioral modeling. The dataset contains 196,477 wallet addresses collected from five blockchain networks (Ethereum, Polygon, Arbitrum, Optimism, and xDai) with 60 engineered features and binary Sybil labels. It was constructed to support research on automated Sybil wallet detection in decentralized finance (DeFi) ecosystems. Ground-truth Sybil labels were obtained from the Hop Protocol airdrop campaign, verified through clustering analysis, temporal pattern inspection, and funding source tracking. Key statistics include: total wallets 196,477, Sybil wallets 14,195 (7.22%), normal wallets 182,282 (92.78%), features 60, columns 62 (wallet_address + is_sybil_attacker + 60 features). The dataset exhibits significant class imbalance (~7.22% Sybil), representative of real-world fraud detection scenarios. Features are organized into groups such as Bridge Transaction, Aggregate, Behavioral/Enhanced, Derived, Co-Temporal Graph (60s and 300s), Funding Graph, and Entity-Funding Cluster (EFC), each with specific counts and descriptions covering cross-chain behavior, risk scores, activity patterns, funding topology, etc. Non-feature columns include unique wallet identifier and binary label.