Table 6 Robustness under noise \(({\varTheta }_{noise} \%)\) across various real evaluation conditions.
From: Digital twin-assisted blockchain IoT security model using contrastive and causal learning techniques
Noise type | Severity (%) | Device | LSTM-IDS | XAI-ML-IDS | DT2SA | BCE-IoT | Causio-twinchain (proposed) |
|---|---|---|---|---|---|---|---|
Gaussian | 5 | Doorbell | 72.4 | 75.8 | 78.1 | 81.5 | 98.1 |
10 | 70.9 | 74.4 | 77.0 | 82.1 | 98.3 | ||
15 | PT Camera | 71.3 | 76.0 | 78.8 | 83.2 | 98.4 | |
20 | 73.0 | 75.1 | 79.2 | 82.7 | 98.5 | ||
Packet Drop | 5 | Thermostat | 72.1 | 74.9 | 78.6 | 82.9 | 98.0 |
10 | 69.8 | 73.6 | 77.3 | 81.8 | 98.2 | ||
15 | Baby Monitor | 71.2 | 75.4 | 79.0 | 83.5 | 98.7 | |
20 | 72.8 | 76.2 | 80.1 | 84.0 | 99.0 | ||
Temporal Jitter | 5 | Doorbell | 70.5 | 74.0 | 77.8 | 82.2 | 98.1 |
10 | 71.9 | 75.1 | 78.9 | 83.0 | 98.6 | ||
15 | PT Camera | 73.1 | 76.5 | 80.3 | 84.4 | 99.1 | |
20 | 72.4 | 75.7 | 79.5 | 83.7 | 98.8 | ||
FGSM (ε = 0.02) | 2 | Thermostat | 71.6 | 74.6 | 78.4 | 82.5 | 98.4 |
FGSM (ε = 0.05) | 5 | 70.8 | 73.8 | 77.6 | 81.9 | 98.0 | |
FGSM (ε = 0.1) | 10 | Baby Monitor | 72.2 | 75.4 | 79.0 | 82.8 | 98.9 |
Mixed Noise (Gaussian + Drop) | 5 + 5 | Doorbell | 73.3 | 76.8 | 80.6 | 84.6 | 99.2 |
10 + 10 | PT Camera | 72.0 | 75.0 | 78.7 | 83.1 | 98.5 | |
Mixed Noise (Jitter + FGSM) | 10 + 5 | Thermostat | 71.4 | 74.4 | 78.0 | 82.4 | 98.3 |
High Mixed Noise (All 15%) | 15 | PT Camera | 73.5 | 76.9 | 80.8 | 84.9 | 99.3 |
Extreme Mixed Noise (All 20%) | 20 | Baby Monitor | 72.7 | 76.0 | 79.7 | 83.8 | 98.7 |
