Future work

■ Use the NSL-KDD as an unsupervised dataset, and be able to validate results

– Attention mechanisms, AE, clustering methods

■ More advanced supervised methods (DNN, CNN, LSTM)

■ Data-centric upgrade: use real data with unsupervised learning

– With data from a secure environment, AEs would perform very well
– NSL-KDD can be used for validation of the unsupervised models
– More advanced project: feature extraction/selection, data cleaning, unsupervised 

methods only

33