As stated in the paper: "We choose files that contain anomalous behavior as testing set. We remove part of the dataset because some graphs are generated in exceptional accidents such as outages and shutdown of hosts"
It's unclear which files are removed, since in the code just 2 files kept, but in the paper results there are more nodes than in the 2 files used as train and test.
As stated in the paper: "We choose files that contain anomalous behavior as testing set. We remove part of the dataset because some graphs are generated in exceptional accidents such as outages and shutdown of hosts"
It's unclear which files are removed, since in the code just 2 files kept, but in the paper results there are more nodes than in the 2 files used as train and test.