A Spatiotemporal, Congestion-Context-Aware Vehicular Communication Dataset for Malicious Flooding Attack Analysis

This dataset presents a comprehensive spatiotemporal representation of vehicular communication under benign and malicious flooding conditions, simulated across both highway and metropolitan mobility environments. It contains twelve flow-level datasets, each representing one of three calibrated malicious flooding attack intensities (LOW, MEDIUM, HIGH) under the two mobility scenarios. The dataset is intended to support research on spatiotemporal anomaly detection, congestion-context-aware security, and machine-learning-based intrusion detection in Vehicular Communication Networks (VCNs). Each data instance corresponds to a single unidirectional communication flow between a transmitting and receiving vehicle. The dataset includes rich temporal features (start time, end time, duration, temporal mid-point), spatial features (positions, mid-points, distances), mobility features (radial displacement and velocity), and communication-layer metrics such as packet and byte counts, throughput, delay, jitter, hop count, delivery ratio, and loss indicators. These combined descriptors capture how movement, network load, and communication dynamics evolve jointly over time. Malicious flooding scenarios were simulated within the communication environment, with attackers transmitting elevated packet or byte volumes while remaining fully consistent with their underlying mobility behavior. This ensures that malicious flows do not exhibit unrealistic spatial artefacts and cannot be trivially separated from benign activity based solely on positional features. The resulting communication patterns provide challenging and realistic attack conditions that require context-aware and temporally informed detection methods. Full details of the flooding attack design and simulation methodology are provided in the associated research publication. This release contains only the flow-level datasets. To enable higher-level analysis, we additionally provide Jupyter notebooks demonstrating how to transform flows into (i) session-level representations (aggregating all flows exchanged between a source–destination pair) and (ii) receiver-window representations (aggregating flows received by each node within a cluster). These notebooks also illustrate how to derive congestion-aware contextual features used in our associated research (see citation in documentation). By combining realistic vehicle mobility, detailed communication metrics, calibrated attack intensities, and extensible multi-level processing scripts, this dataset offers a comprehensive foundation for research on spatiotemporal security, anomaly modelling, congestion-aware intrusion detection, and machine-learning analysis in VCNs.本数据集提供了良性与恶意泛洪攻击场景下车载通信的完整时空表征，该数据集基于高速公路与城市两种移动场景模拟生成。本数据集包含12个流级数据集，分别对应两种移动场景下三种经标定的恶意泛洪攻击强度（LOW、MEDIUM、HIGH）各一组。本数据集旨在支撑车载通信网络（Vehicular Communication Networks，简称VCNs）领域内的时空异常检测、拥塞上下文感知安全以及基于机器学习的入侵检测相关研究。 每个数据实例对应收发车辆间的单条单向通信流。本数据集包含丰富的时序特征（起始时间、终止时间、持续时长、时序中点）、空间特征（位置、空间中点、距离）、移动特征（径向位移与速度），以及通信层指标，如数据包与字节数、吞吐量、延迟、抖动、跳数、交付率与丢包标识。这些联合描述符能够刻画车辆移动、网络负载与通信动态随时间的协同演化规律。 研究团队在通信环境中模拟了恶意泛洪场景：攻击者在保持自身移动行为与原始移动轨迹完全一致的前提下，发送超出正常量级的数据包或字节流量。这一设计确保恶意通信流不会出现不符合现实的空间伪影，且无法仅通过位置特征就轻易与良性通信流区分开来。最终得到的通信模式具备挑战性与真实性，要求检测方法具备上下文感知与时序感知能力。泛洪攻击的具体设计与仿真方法的完整细节可参见相关研究论文。 本次发布仅包含流级数据集。为支持更高层级的分析任务，我们额外提供了Jupyter笔记本代码，用于演示如何将流数据转换为：(i) 会话级表征（聚合源-目的节点对之间的所有通信流），以及(ii) 接收窗口级表征（聚合集群内各节点接收的通信流）。这些代码还演示了如何提取本研究中使用的拥塞感知上下文特征（详见文档中的引用说明）。 本数据集结合了真实的车辆移动模式、精细的通信指标、标定后的攻击强度，以及可扩展的多级处理脚本，为车载通信网络领域内的时空安全、异常建模、拥塞感知入侵检测以及机器学习分析研究提供了全面的研究基础。