Anomaly Detection in BACnet systems
Meeting 25
Title: K-modes
Date: 29/03/2019
Time: 5 pm
Location: Zoom Conference Call
Attendees: Sofian Ghazali, Rahul Balamurugan, Zahid Kamil
Absences: None
Action Items: How to create anomalous data and visualize these clusters
Meeting Agenda:
-
Create anomalous data
-
Visualize clusters
Discussion and Notes:
The meeting was focused on how to create anomalous data. The anomalous data created should not resemble the same data that we have. However, we do not know how anomalous data looks like so we have to assume what the data could have. We assume that it would have IP addresses that are not used in the normal BACnet data that was collected from Qatar University. This is how we can check if all our hardwork on the preprocessing and the machine learning model would pay off or else we would have to start from scratch again.
Next, we also need to determine how the clusters would look like. We know that only two features can be represented in a 2-dimensioal axis and we have 76 features so we opted to move with bar graphs using the python library of seaborn.
Next meeting action items:​
-
Generate Anomalous data
Meeting Adjournment: 6:30 pm
Meeting Minutes prepared by: Zahid Kamil