基于储备证明的以太坊中心化交易所地址分类链上行为特征
On-Chain Behavioral Signatures for Classifying Centralized Exchange Addresses on Ethereum Using Proof-of-Reserves
Seongchan Park · Hyunho Jeon · Byungsoo Kang · Gaeun Kim · Minseob So · Seongyong Kim
中文摘要
中心化交易所（CEX）是传统金融与数字资产市场之间的关键通道，但识别其区块链地址仍然困难。现有方法常依赖不透明的归属标签，或将交易所视为次要分类目标，对未见交易所的泛化能力考察不足。本研究利用储备证明（PoR）披露构建分类框架，涵盖来自 15 家 CEX 和五个非 CEX 类别的 3,033,857 个以太坊地址。提取 58 个链上行为特征，在留一交易所（leave-one-exchange-out）协议下比较 LightGBM 与逻辑回归（LR）。在 30 至 730 天的五个回看窗口上，LightGBM 始终优于 LR：使用选定的 17 个特征时，宏平均 PR-AUC 从 0.6611 提升至 0.9594，而 LR 在 730 天时为 0.7697。稳健性检验显示，在不同划分与留出交易所上重复特征选择和超参数优化时性能相当，交易所均衡重加权进一步提升了宏平均 PR-AUC。SHAP 分析和特征组消融表明，gas 行为、ERC-20 代币活动和内部交易是所构建数据集中的关键预测信号。这些发现支持用于区块链监控的可解释 CEX 地址筛查，同时发布的特征数据集可支持可复现的后续研究。
关键要点
- 01现有方法依赖不透明标签或把交易所当次要目标，对未见交易所泛化不足
- 02利用 PoR 披露构建 15 家 CEX 与五类非 CEX 共 3,033,857 个以太坊地址数据集
- 03提取 58 个链上行为特征，留一交易所协议下 LightGBM 宏 PR-AUC 达 0.9594，优于 LR 的 0.7697
- 04SHAP 与消融显示 gas 行为、ERC-20 活动和内部交易是关键预测信号
- 05交易所均衡重加权可提升性能，发布特征数据集支持可复现研究
解读
尚无解读。
原始英文摘要
arXiv:2610.41888v1 Announce Type: cross Abstract: Centralized exchanges (CEXs) are key gateways between traditional finance and digital asset markets, yet identifying their blockchain addresses remains challenging. Existing approaches often rely on opaque attribution labels or treat exchanges as secondary classification targets, leaving generalization to unseen exchanges insufficiently examined. This study develops a classification framework using Proof-of-Reserves (PoR) disclosures, with 3,033,857 Ethereum addresses from 15 CEXs and five Non-CEX categories. We derive 58 on-chain behavioral features and evaluate LightGBM against logistic regression (LR) under a leave-one-exchange-out protocol. Across five lookback windows (30-730 days), LightGBM consistently outperformed LR, with macro PR-AUC increasing from 0.6611 to 0.9594 using the selected 17 features, compared with 0.7697 for LR at 730 days. Robustness checks showed comparable performance when feature selection and hyperparameter optimization were repeated across splits and held-out exchanges, while exchange-balanced reweighting improved macro PR-AUC. SHAP analysis and feature-group ablation identified gas behavior, ERC-20 token activity, and internal transactions as key predictive signals in the constructed dataset. These findings support interpretable CEX address screening for blockchain monitoring, while the released feature dataset enables reproducible follow-up research.