Dataset of RTT latency internet measurements using Ripe Atlas anchor nodes in Europe.

This dataset contains real-world latency measurements in internet collected from a distributed network of probing nodes in Europe, designed to enhance IP geolocation accuracy through machine learning techniques. It includes two separate datasets: a Learning Dataset for model training and a Validation Dataset for performance evaluation. 

Scenario Description 

The dataset comprises Round-Trip Time (RTT) latency internet measurements in Europe from different geographically distributed probing nodes (Monitors) to target IP addresses. The geolocation information (latitude/longitude) is provided as ground truth for both the Landmarks (training targets) and the Target Nodes (validation targets) to evaluate model accuracy. As a proof-of-concept, it is used the well-known Ripe Atlas anchor nodes inside Europe to act both as landmarks and target nodes (https://atlas.ripe.net/anchors/). 

Dataset Structure 

Each dataset contains multiple rows, where each row represents a RTT fingerprint vector consisting of latency measurements from multiple Monitors to a given target. 

1) Learning Dataset (LearningDataset_RTT_RipeAtlasEU.csv) 

Monitors deployed: 4 (distributed across different geographical locations). 

Targets: Known geographical locations (Landmarks), used for training. 

Columns: 

measure_id: Unique identifier for each measurement. 

anchor_id: ID of the target node to be geolocated.

dst_ip: IP address of the target node.

init_time: Timestamp of the measurement.

latency_m1 - latency_m4: RTT fingerprint vector consisting of latency measurements from 4 different Monitors.

latitude, longitude: Ground truth geolocation of the target node. 

2) Validation Dataset (ValidationDataset_RTT_RipeAtlasEU.csv) 

Monitors deployed: 4 (same as Learning Dataset). 

Targets: IPs used to evaluate model performance. Their actual locations are known but treated as unknown during inference. 

Columns: 

measure_id: Unique identifier for each measurement.

anchor_id: ID of the target node to be geolocated.

dst_ip: IP address of the target node.

init_time: Timestamp of the measurement.

latency_m1 - latency_m4: RTT fingerprint vector consisting of latency measurements from 4 different Monitors.

latitude_anchor, longitude_anchor: Ground truth geolocation of the target node, used to evaluate model accuracy (not used as an input for models).