IoT/IoMT Attack Detection on CICIoT 2023 (test)
99.78Mean F1 ScoreSHapley Additive exPlanations (SHAP)-based attribution fingerprinting
Evaluation results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SHapley Additive exPlanations (SHAP)-based attribution fingerprintingAttacks=PGD2025.11 | 99.78 | 99.78 | 99.55 | 100 | 99.93 | 99.81 | 99.55 | 100 | 0.45 | 0 | 10,000 | 9,955 | 45 | 0 | |
| SHapley Additive exPlanations (SHAP)-based attribution fingerprintingAttacks=FGSM2025.11 | 99.51 | 99.52 | 99.55 | 99.48 | 99.87 | 99.72 | 99.55 | 99.48 | 0.45 | 0.52 | 9,948 | 9,955 | 45 | 52 | |
| SHapley Additive exPlanations (SHAP)-based attribution fingerprintingAttacks=DeepFool2025.11 | 99.29 | 99.3 | 99.55 | 99.04 | 99.86 | 99.72 | 99.55 | 99.04 | 0.45 | 0.96 | 9,904 | 9,955 | 45 | 96 | |
| LDMDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 99 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LDMDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 99 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DMDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 97 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAEDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 96 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Adversarially Trained ModelAttacks=PGD2025.11 | 94.58 | 94.72 | 97.2 | 92.1 | 99.47 | 99.21 | 97.35 | 92.49 | 2.65 | 7.9 | 9,210 | 9,735 | 265 | 790 | |
| GANDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 94 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Adversarially Trained ModelAttacks=FGSM2025.11 | 93.3 | 93.56 | 97.13 | 89.77 | 98.98 | 98.56 | 97.35 | 90.49 | 2.65 | 10.23 | 8,977 | 9,735 | 265 | 1,023 | |
| DMDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 92 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SMOTEDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 89 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAEDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 86 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GANDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 83 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Adversarially Trained ModelAttacks=DeepFool2025.11 | 78.62 | 81.92 | 96.16 | 66.49 | 96.91 | 96.26 | 97.35 | 74.39 | 2.65 | 33.51 | 6,649 | 9,735 | 265 | 3,351 | |
| Traffic-MoEparadigm=Pre-training and Fine-tuning2026.01 | 78.24 | 85.88 | 80.07 | 77.01 | — | — | — | — | — | — | — | — | — | — | |
| NetGPTparadigm=Pre-training and Fine-tuning2026.01 | 69.62 | 76.84 | 77.56 | 67.09 | — | — | — | — | — | — | — | — | — | — | |
| SMOTEDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 66 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TrafficFormerparadigm=Pre-training and Fine-tuning2026.01 | 65.94 | 76.36 | 71.72 | 64.11 | — | — | — | — | — | — | — | — | — | — | |
| DMDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 53 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ET-BERTparadigm=Pre-training and Fine-tuning2026.01 | 52.17 | 74.55 | 54.16 | 51.72 | — | — | — | — | — | — | — | — | — | — | |
| AppScannerparadigm=Machine Learning (ML)2026.01 | 46.19 | 52.66 | 55.35 | 44.71 | — | — | — | — | — | — | — | — | — | — | |
| BaselineDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=None (Imbalanced)2026.01 | 46 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LDMDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 44 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FS-Netparadigm=Deep Learning (DL)2026.01 | 40.94 | 49.03 | 52.65 | 40.46 | — | — | — | — | — | — | — | — | — | — | |
| SMOTEDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 35 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAEDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 34 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GANDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=Augmented to class balance2026.01 | 33 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FlowPrintparadigm=Machine Learning (ML)2026.01 | 19.89 | 45.78 | 31.66 | 20.48 | — | — | — | — | — | — | — | — | — | — | |
| BaselineDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=None (Imbalanced)2026.01 | 17 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BaselineDetection Model=Ensemble (subspace method, 100 cycles), Data Balancing=None (Imbalanced)2026.01 | 0 | — | — | — | — | — | — | — | — | — | — | — | — | — |