"UAVIDS-2025: A Benchmark Dataset for Intrusion Detection in UAV Networks Using Machine Learning Techniques"

"Abstract \/ Dataset DescriptionUAVIDS-2025 is a comprehensive benchmark dataset designed for evaluating intrusion detection systems (IDS) in UAV (Unmanned Aerial Vehicle) swarm networks. The dataset was generated through extensive simulations using the NS-3.24 network simulator, with realistic UAV mobility modeled by an extended BOID algorithm. It includes 122,171 labeled flow records across five traffic categories: Normal, Blackhole, Flooding, Sybil, and Wormhole attacks.Each data sample represents a network flow characterized by 22 features, grouped into connection, traffic volume, and performance metrics. The simulations were configured with IEEE 802.11ac wireless standards, AODV routing, and a Nakagami channel model to ensure realism. This dataset enables the evaluation of machine learning-based IDS under various scenarios, including imbalanced attack distributions and swarm mobility.The dataset supports research in:Supervised\/unsupervised intrusion detectionFederated learning and decentralized securityAdversarial robustness and synthetic data generationUAVIDS-2025 is intended to provide a reproducible, scalable, and diverse testbed for the research community working on the security of UAV networks."

摘要/数据集概述
UAVIDS-2025是一款专为评估无人机集群网络入侵检测系统（Intrusion Detection Systems, IDS）而设计的综合性基准数据集。该数据集依托NS-3.24网络模拟器开展大规模仿真生成，采用扩展BOID算法（Extended BOID Algorithm）对真实无人机的移动特性进行建模。数据集包含122171条带标签的流量记录，涵盖5类流量场景：正常流量、黑洞攻击、泛洪攻击、女巫攻击（Sybil Attack）与虫洞攻击。
每条数据样本对应一条网络流量流，具备22项特征，可划分为连接特征、流量体量特征与性能指标三大类。本次仿真采用IEEE 802.11ac无线标准、AODV（自组织按需距离矢量）路由协议以及Nakagami信道模型进行配置，以保障仿真场景的真实性。本数据集可用于评估基于机器学习的入侵检测系统在多种场景下的性能，包括攻击分布不平衡与无人机集群移动等场景。
本数据集支持以下研究方向：
监督/无监督入侵检测
联邦学习与去中心化安全
对抗鲁棒性与合成数据生成
UAVIDS-2025旨在为从事无人机网络安全研究的学术共同体提供可复现、可扩展且多样化的测试平台。

搜集汇总

背景与挑战

背景概述

UAVIDS-2025是一个专为无人机网络入侵检测系统评估设计的基准数据集，包含122,171条带标签的网络流记录，涵盖正常流量和四种常见攻击类型。该数据集通过NS-3.24模拟器生成，具有22个特征，支持机器学习研究，如监督学习、联邦学习和对抗性鲁棒性分析，旨在为无人机网络安全提供一个可重复且多样化的测试平台。