PCA is a dimensionality reduction technique commonly used in machine learning and data analysis. While PCA is primarily used for dimensionality reduction, it can also be employed for feature selection, albeit indirectly. Steps:
- Wireshark was used to create a .pcap file from network traffic. Then this dataset was created.
- CICIFlowMeter converted 84 features from a .pcap file to a .csv file.
- PCA calculated the top 10 features based on the variance shown in the table below. PCA was calculated using the’sklearn’ library of Python.
| Serial | Features | Variance (%) |
|---|---|---|
| 1 | Protocol | 18.446 |
| 2 | Flow Duration | 14.115 |
| 3 | Tot Fwd Pkts | 11.918 |
| 4 | Tot Bwd Pkts | 9.539 |
| 5 | TotLen Fwd Pkts | 4.649 |
| 6 | TotLen Bwd Pkts | 3.627 |
| 7 | Fwd Pkt Len Max | 3.564 |
| 8 | Fwd Pkt Len Min | 3.350 |
| 9 | Fwd Pkt Len Mean | 3.298 |
| 10 | Fwd Pkt Len Std | 2.785 |