EDoS_Network_Slicing_K8S

The dataset includes raw resource usage and service performance metrics collected from CNFs of vCDN slices deployed on a cloud-native platform (kubernetes). The The dataset serves to train ML models for mitigating EDoS attacks by detecting anomalous resource scaling operations caused by application-layer DDoS attacks. The raw resource usage and performance metrics data have been recorded from two vCDN slices over a period of 5 days. A time series for each resource usage and performance metric was recorded using a data sampling period of 300s. The training data were collected during the first 4 days of attack-free activity. The 5th day served to create the testing dataset which includes data for normal activity as well as anomalous activity caused by application-layer DDoS attacks (specifically Hulk and Slowloris attacks) executed on different periods of the day. A training dataset and a testing dataset, containing the raw multivariate time series data, are generated for each vCDN slice's CNF with a total of 5401 and 2701 samples, respectively. For each slice, we provide the anomaly labels, which correspond to the attack periods.