LCORE-D: A Benchmark Dataset for Core Network Analysis

This dataset presents a controlled, fault-annotated network monitoring dataset collected from a large-scale emulated Internet Service Provider Network (ISP) deployed using GNS3. The dataset comprises 407 hours of network monitoring data, which includes both induced fault events and naturally occurring transient faults. The dataset captures both healthy and faulty network behaviour under a wide range of intentionally induced hard failures, such as single and multiple link failures, node failures, routing misconfigurations, and line card failures. In addition to these deliberate fault injections, the dataset also includes transient faults that were not intentionally induced. These transient faults emerge naturally as a consequence of network dynamics, including protocol reconvergence, brief control-plane instability, and short-duration connectivity disruptions that resolve without operator intervention. By preserving these transient events rather than filtering them out, the dataset supports research on transient fault detection, false alarm reduction, and the temporal evolution of network failures.