Table 3 Comparison of Precision rates (%).
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 | 59.38 | 66.67 | 70.49 | 97.62 | 100 | 100 | 100 | 100 | 67.74 |
Naive bayes | 59.04 | 61.54 | 64.38 | 75.00 | 100 | 100 | 100 | 100 | 100 | |
Decision trees | 76.92 | 80.00 | 78.95 | 94.12 | 100 | 100 | 100 | 100 | 83.72 | |
Random forests | 72.31 | 83.05 | 80.36 | 100 | 100 | 100 | 100 | 100 | 92.59 | |
SVM | 96.15 | 86.21 | 100 | 100 | 100 | 100 | 100 | 100 | 92.31 | |
GoogLeNet | 100 | 98.04 | 98.00 | 100 | 100 | 100 | 100 | 100 | 72.73 | |
ResNet | 100 | 100 | 100 | 100 | 100 | 100 | 100 | 100 | 100 | |
VIT | 63.64 | 94.44 | 75.47 | 100 | 100 | 100 | 100 | 100 | 84.21 | |
Swin Transformer | 85.71 | 52.94 | 63.16 | 53.26 | 53.75 | 53.16 | 57.14 | 54.55 | 66.67 | |
Unsupervised learning | K-Means | 66.67 | 68.75 | 74.07 | 59.52 | 100 | 100 | 100 | 100 | 62.50 |
K-Medoids | 66.18 | 66.67 | 74.07 | 59.52 | 100 | 100 | 100 | 100 | 62.50 | |
Hierarchical clustering | 59.76 | 59.26 | 60.24 | 60.98 | 100 | 100 | 100 | 100 | 100 | |
FCM | 66.67 | 68.25 | 70.00 | 60.47 | 100 | 100 | 100 | 100 | 60.98 | |
DBSCAN | n/a | 51.02 | n/a | 100 | 100 | 100 | n/a | n/a | n/a | |
Spectral clustering | 68.42 | 75.00 | 76.60 | 59.52 | 59.52 | 59.52 | 59.52 | 100 | 60.98 | |
Our Proposal | 96.15 | 97.92 | 98.04 | 100 | 100 | 100 | 100 | 100 | 93.33 |
