Synthetic Network Log Dataset for Anomaly Detection Benchmarking

A reproducible synthetic network traffic dataset with ground-truth anomaly labels, generated for benchmarking anomaly detection algorithms. Contains 19,643 log entries spanning 72 hours across a simulated internal network, with three labeled anomaly types: volume spikes, port scanning, and beaconing (C2 check-in simulation). Each entry includes timestamp, source/destination IP, port, bytes transferred, protocol, connection status, and ground-truth label. The dataset is fully reproducible using the included generator script with --seed 42. Intended for training and evaluating anomaly detection methods where known ground truth is required.