Deep Encrypted Traffic Detection: An Anomaly Detection Framework for Encryption Traffic Based on Parallel Automatic Feature Extraction
Table 2
Comparison of feature extraction efficiency of two automatic feature selection algorithms.
Amount of data
Time consumed by serial feature extraction (s)
Time consumed by parallel feature extraction (s)
360000
3984.83
1382.54
800000
8847.78
3070.24
1160000
12835.74
4453.67
Enhanced effect
When using the parallel blocked feature extraction algorithm: the time used by the parallel feature extraction characteristic is about 34% of that of the time consumed by the unblocked algorithm, and it greatly improves the feature extraction efficiency