"SmartGridDDoS"

"This dataset contains labeled, time-windowed network-traffic features generated from ns-3 simulation traces for a fog-based smart grid communication scenario. Raw ns-3 ASCII trace files (.tr) were parsed into per-packet tabular CSVs by extracting timestamps, source\/destination IP addresses, transport protocols, ports (when available), packet sizes, flow identifiers, and inter-arrival times. The per-packet records were then aggregated into fixed-length time windows (e.g., 100 ms) per smart meter (UE), producing a machine-learning and reinforcement-learning-ready dataset where each row represents the traffic behavior of one UE during one time window.For each (UE, window), the dataset provides statistical features that characterize traffic volume and dynamics, including packet and byte rates (pps, byte rate), uplink ratio, protocol composition (TCP\/UDP\/ICMP packet counts), inter-arrival time statistics (mean, standard deviation, p90, p99, coefficient of variation), packet-size statistics (mean, standard deviation, p90), flow diversity (number of unique flows) and destination diversity (number of unique destinations). Heartbeat-related indicators are included when applicable (heartbeat rate, delivery ratio, and heartbeat jitter). Window-level labels are derived from known packet-level labels (port- and schedule-based), where a window is marked as an attack if any attack packet occurs for that UE within the window; otherwise, it is labeled benign. Optional exponential moving averages (EMA) of key rates are also included to provide smoothed state signals.The dataset is organized to support reproducible evaluation of intrusion detection and self-healing control in smart grid traffic, enabling benchmarking of classifiers and multi-agent policies using consistent, window-based features and causally defined labels."