Research Article
Few-Shot Learning-Based Network Intrusion Detection through an Enhanced Parallelized Triplet Network
Table 6
The detection results for different attack types on the UNSW_NB15 dataset.
| Type | CNN | ResNet | XGBoost | Proposed |
| Reconnaissance | 77.70 | 50.7 | 98.5 | 95.1 | Backdoor | 98.08 | 55.1 | 97.9 | 99.65 | DoS | 93.30 | 91.1 | 96.8 | 97.3 | Exploits | 88.50 | 90.6 | 91.9 | 96.3 | Analysis | 97.90 | 98.80 | 93.55 | 99.40 | Fuzzers | 73.30 | 86.4 | 71.2 | 97.10 | Generic | 99.3 | 99.6 | 99.5 | 99.7 | Worms | 70.45 | 72.72 | 70.45 | 100 | Shellcode | 67.72 | 47.35 | 94.9 | 90.48 |
|
|