Hotfix rawlog

Rawlog fix
Attack detection fix
Logs cleaned

Kubernetes

• Kubernetes deployment
• Incident detector
• Saving incidents to s3 cloud storage
• Dynamic thresholds

Model live update and Feedback

• Update the model while the pipeline is running
• Changes for the feedback dashboard

Model Development

Model development:

• AnomalyModel
• Feature importance: Shapley values calculation
• Model evaluation

Pipeline separation dev

Creation of four separate pipelines:

• Preprocess: runs on the client - for the log processing, the resulting feature vectors are sent to the prediction pipeline
• Post-process: also runs on the client - for handling the result of the prediction
• Prediction: receives the feature vectors and provides prediction
• Training: to train Isolation Forest

Initial release with IForest

Fully functional Baskerville with IForest