Table 4 Comparison of F1 scores.
From: A framework for hardware trojan detection based on contrastive learning
| Â | Method | T100 | T700 | T800 | T500 | T1900 | T1600 | T600 | T2000 | T2100 |
|---|---|---|---|---|---|---|---|---|---|---|
Supervised learning | KNN | 0.6667 | 0.7586 | 0.7748 | 0.8913 | 1 | 1 | 1 | 1 | 0.5185 |
Naive bayes | 0.7368 | 0.75 | 0.7642 | 0.1111 | 1 | 1 | 1 | 1 | 0.1481 | |
Decision trees | 0.7843 | 0.8727 | 0.8411 | 0.9505 | 1 | 1 | 1 | 1 | 0.7742 | |
Random forests | 0.8174 | 0.8991 | 0.8491 | 0.9691 | 1 | 1 | 1 | 1 | 0.6494 | |
SVM | 0.9804 | 0.9259 | 1 | 1 | 1 | 1 | 1 | 1 | 0.381 | |
GoogLeNet | 1 | 0.9901 | 0.98 | 1 | 1 | 1 | 1 | 1 | 0.5783 | |
ResNet | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | |
VIT | 0.7717 | 0.7907 | 0.7767 | 0.6842 | 1 | 1 | 1 | 1 | 0.4638 | |
Swin Transformer | 0.2105 | 0.6667 | 0.3478 | 0.6901 | 0.6615 | 0.6512 | 0.4103 | 0.6957 | 0.3944 | |
Unsupervised learning | K-Means | 0.7273 | 0.7719 | 0.7692 | 0.5435 | 1 | 1 | 1 | 1 | 0.5556 |
K-Medoids | 0.7627 | 0.7731 | 0.7692 | 0.5435 | 1 | 1 | 1 | 1 | 0.5556 | |
Hierarchical clustering | 0.7424 | 0.7328 | 0.7519 | 0.5495 | 1 | 1 | 1 | 1 | 0.1132 | |
FCM | 0.7434 | 0.7611 | 0.7636 | 0.5591 | 1 | 1 | 1 | 1 | 0.5495 | |
DBSCAN | n/a | 0.6757 | n/a | 0.0392 | 1 | 1 | n/a | n/a | n/a | |
Spectral clustering | 0.729 | 0.7021 | 0.7423 | 0.5435 | 0.7463 | 0.7463 | 0.7463 | 0.3333 | 0.5495 | |
Our Proposal | 0.9804 | 0.9592 | 0.9901 | 1 | 1 | 1 | 1 | 1 | 0.8842 |
