Offsim_Dataset

This dataset, named OffSim, has been specifically designed to support research on fraud detection in offline Central Bank Digital Currency (CBDC) payment systems. OffSim is a synthetic, large-scale dataset generated using a custom multi-agent simulator that extends the widely used PaySim framework. While PaySim replicates transactional behavior in mobile money systems such as M-Pesa, OffSim introduces novel features tailored to offline transaction scenarios, including synchronization delay, peer-to-peer indicators, merchant trust scores, and simulated fraud vectors such as double spending and timestamp tampering.The dataset reflects the operational constraints of offline environments, such as limited connectivity, device-level storage, and delayed ledger reconciliation. It provides labeled examples of both legitimate and fraudulent transactions, enabling robust benchmarking of fraud detection models. The dataset is ideal for training and evaluating hybrid detection architectures that combine rule-based systems with machine learning techniques, especially under extreme class imbalance and resource-constrained deployment settings.OffSim promotes reproducible research in secure digital payments and supports the development of explainable and efficient fraud detection algorithms tailored to emerging CBDC infrastructure

本数据集命名为OffSim，专为支持离线中央银行数字货币（Central Bank Digital Currency, CBDC）支付系统中的欺诈检测研究而设计。OffSim是一款合成式大规模数据集，依托定制化多智能体模拟器生成，该模拟器扩展了当前广泛使用的PaySim框架。尽管PaySim复刻了诸如M-Pesa这类移动货币系统的交易行为，但OffSim新增了专为离线交易场景定制的全新特性，涵盖同步延迟、点对点（peer-to-peer）标识、商户信任评分，以及双重支付、时间戳篡改等模拟欺诈向量。该数据集还原了离线环境下的各类运营约束，例如有限的网络连接、设备级存储限制以及延迟的账本对账流程。数据集提供了合法交易与欺诈交易的标注样本，可用于对欺诈检测模型开展稳健的基准性能测试。该数据集十分适用于训练与评估融合了基于规则系统与机器学习技术的混合检测架构，尤其适配极端类别不平衡与资源受限的部署场景。OffSim可推动安全数字支付领域的可复现研究，并助力针对新兴CBDC基础设施开发可解释且高效的欺诈检测算法。