Dataset and Code for Protecting communication and computing systems against poisoning attacks for Cloud Edge-Based Federated Learning System

Because of its significance and extensive usage, the KDDCUP'99 dataset is used in our edge-based intrusion detection study. The abundance of attack types and scenarios included, together with the extensive feature set and large instance repository, make this dataset an invaluable addition to the discipline. When comparing and contrasting different intrusion detection systems and approaches, this statistic is essential. The following investigation will shed light on the difficulties and restrictions imposed by the KDDCUP'99 dataset. These difficulties are exacerbated by the following factors: the age of the underlying network environment, the existence of unnecessary and uninformative characteristics, the prevalence of mislabelled instances, and the inherent bias in class distribution鉴于其重要学术价值与广泛应用，KDDCUP'99数据集被应用于本次基于边缘的入侵检测研究。该数据集涵盖了丰富的攻击类型与攻击场景，同时具备完备的特征集合与大规模样本库，因此成为该领域极具价值的研究资源。在对比与评估各类入侵检测系统与检测方法的工作中，该数据集是不可或缺的核心基准。下文将深入剖析KDDCUP'99数据集所存在的难点与局限性，而这些难点会因以下因素进一步加剧：底层网络环境的年代久远、冗余且无信息增益的特征大量存在、样本标注错误频发，以及类别分布本身存在固有偏倚。