Investigation and rectification of NIDS datasets and standardized feature set derivation for network attack detection with graph neural networks