Research Article
Network Intrusion Detection with Nonsymmetric Deep Autoencoding Feature Extraction
Table 5
Experimental results of 11 attacks in the CIC-IDS2017 dataset.
| Attack type | Precision (%) | Recall (%) | F1 (%) | SNDAE | ANDAE | SNDAE | ANDAE | SNDAE | ANDAE |
| FTP-Patator | 97.1 | 99.87 | 98.53 | 98.95 | 97.81 | 99.4 | SSH-Patator | 96.02 | 99.42 | 95.13 | 97.85 | 95.56 | 98.63 | DoS | 99.83 | 99.79 | 99.66 | 99.67 | 99.74 | 99.73 | Heartbleed | 0 | 0 | 0 | 0 | 0 | 0 | Brute Force | 91.81 | 99.31 | 91.58 | 96.67 | 91.64 | 97.97 | XSS | 97.24 | 99.47 | 87.77 | 92.85 | 92.21 | 96.04 | Sqlinjection | 0 | 0 | 0 | 0 | 0 | 0 | Infiltration | 0 | 83.33 | 0 | 33.33 | 0 | 47.43 | Botnet | 73.01 | 92.56 | 71.81 | 75.92 | 72.32 | 83.42 | DDoS | 99.97 | 99.96 | 99.89 | 99.93 | 99.93 | 99.95 | Portscan | 99.97 | 99.99 | 99.92 | 99.97 | 99.52 | 99.98 |
|
|
The bold values are the larger of the evaluation metrics of the two models.
|