Low-power and Lossy Network Anomaly-Based Dataset in Internet of Things Networks

This repository provides a reproducible RPL-based IoT intrusion-detection dataset generated using the Contiki-Cooja simulation environment. The dataset covers five individual routing attacks—Blackhole, Selective Forwarding, Flooding, Decrease Rank, and Worst Parent—together with a Combined attack condition. Experiments were conducted using Centralized and Distributed network topologies under Static and Random-Waypoint Mobile conditions, with network sizes of 8, 16, and 32 nodes. Each experimental condition was independently executed five times using distinct Cooja random seeds, resulting in 360 independent simulation executions. The dataset integrates temporal, RPL routing, network-traffic, node-level Energest/Powertrace, and positional information. The repository also provides simulation configurations, attack implementations, seed and mobility information, preprocessing resources, feature-selection analysis, Random Forest evaluation results, statistical-analysis outputs, and supporting visualizations to facilitate reproducibility and independent verification. The machine-learning evaluation follows a run-grouped validation design in which complete independent simulation runs are separated between training and testing. Five feature-selection methods—Pearson correlation, Mutual Information, Chi-square, ANOVA, and Recursive Feature Elimination (RFE)—are evaluated using a common Random Forest classifier.