ZYELL Communications Corp. is a communication company that has 30 years of experience in network service. Based on ZYELL’s real-world data that has accumulated over years and the excellent network traffic data analysis ability, we provide a new realworld network anomaly detection dataset.
The data will be time series network traffic records captured by ZYELL’s firewall and the number of data would be about 1 million. Each network traffic record is a network connection session (or a flow) and is labeled as either normal or an exact attack type.
In order to target specific types of attacks and more refined classification, we selected two main categories, i.e. Denial of Service (DOS) and Probing:
Each network traffic record has about 21 flow-based features, including Connection Duration (Seconds), Inbound/Outbound Traffic Count (Bytes), and so forth, during this network connection session. The proportion of anomalies is about 1%. The data format will be a CSV file.