Preprocessed CICIDS2017 Dataset and Reproducible Implementation for Sequence-Based Intrusion Detection using LSTM and Neural ODE Models

This dataset contains the complete experimental resources used for developing and evaluating deep learning–based intrusion detection models on the CICIDS2017 dataset. The objective of this work is to detect network intrusions using sequential deep learning architectures. The study transforms raw network traffic records into fixed-length temporal sequences and evaluates three models: Long Short-Term Memory (LSTM) Euler-based Neural ODE model Runge–Kutta 4 (RK4)-based Neural ODE model本数据集包含了在CICIDS2017数据集上开发与评估基于深度学习的入侵检测模型所需的全部实验资源。 本研究的目标为采用序列深度学习架构实现网络入侵检测。本研究将原始网络流量记录转换为固定长度的时序序列，并对三种模型展开评估： 长短期记忆网络（Long Short-Term Memory，LSTM） 基于欧拉法的神经常微分方程（Neural ODE）模型 基于四阶龙格-库塔（Runge–Kutta 4，RK4）法的神经常微分方程模型