Research Article
IoT Security Detection Method Based on Multifeature and Multineural Network Fusion
| Eigenvalue | Evaluation indicators | RNN (recurrent neural network) (%) | LSTM (Long short-term memory) neural network (%) | GRU (gated neural network) (%) |
| Normal (normal) | TPR (true positive rate) | 99.785652 | 99.70653 | 99.785632 | FPR (false positive rate) | 0.02629 | 0.02623 | 0.02645 | Precision | 99.924538 | 99.536 | 99.924538 |
| Key (whether the key is compromised) | TPR (true positive rate) | 55.706935 | 55.84048 | 55.825778 | FPR (false positive rate) | 0.24754 | 0.0365363 | 0.365378 | Precision | 97.470417 | 96.322675 | 96.321742 |
| FR (spoofing the router to join the network) | TPR (true positive rate) | 94.716305 | 95.237292 | 94.703649 | FPR (false positive rate) | 0.2544174 | 0.2869203 | 2.533927 | Precision | 77.965484 | 75.980211 | 78.030935 |
| Authentication (presence or absence of authentication mechanisms) | TPR (true positive rate) | 97.770954 | 97.751925 | 97.995848 | FPR (false positive rate) | 0.2576831 | 0.2512793 | 2.889814 | Precision | 89.793535 | 90.01485 | 88.742412 |
| NE (network data encryption status) | TPR (true positive rate) | 96.295142 | 96.866101 | 97.52915 | FPR (false positive rate) | 0.2466249 | 0.2996873 | 3.521875 | Precision | 84.666841 | 82.114505 | 79.797605 |
| ACL (access control) | TPR (true positive rate) | 80.545442 | 80.666809 | 82.123224 | FPR (false positive rate) | 0.079273 | 0.010239 | 0.35294 | Precision | 96.534611 | 95.576045 | 86.475718 |
| NODE (identify foreign nodes) | TPR (true positive rate) | 42.371681 | 8.106195 | 14.584071 | FPR (false positive rate) | 1.402514 | 0.0259163 | 0.452776 | Precision | 14.449541 | 14.764668 | 15.158205 |
| NOR (detects nonrepudiation of network node communications) | TPR (true positive rate) | 25.150763 | 31.997162 | 1.631784 | FPR (false positive rate) | 0.795674 | 0.104477 | 0.054798 | Precision | 14.926316 | 14.55543 | 14.110429 |
| Aggregate | acc (accuracy) | 90.137387 | 90.13333 | 90.1364139 |
|
|