Table 3 Performance comparison of the proposed model with methods presented in [19] including SVM, KNN, Naïve Bayes, NN, DNN and DAE
From: Practical autoencoder based anomaly detection by using vector reconstruction error
Recall | Precision | F1-score | Accuracy | Method | Class |
|---|---|---|---|---|---|
0.509 | 0.186 | 0.272 | 0.515 | SVM | Normal |
0.997 | 0.973 | 0.985 | 0.994 | KNN | |
0.975 | 0.991 | 0.983 | 0.994 | Naïve Bayes | |
0 | 0 | 0 | 0.822 | NN | |
0 | 0 | 0 | 0.850 | DNN | |
0.805 | 0.810 | 0.795 | 0.822 | DAE | |
1 | 1 | 1 | 1 | Our proposed model | |
0.150 | 0.022 | 0.039 | 0.888 | SVM | Attacker |
0.997 | 0.979 | 0.988 | 1 | KNN | |
0.999 | 0.999 | 0.999 | 1 | Naïve Bayes | |
0 | 0 | 0 | 0.985 | NN | |
0 | 0 | 0 | 0.989 | DNN | |
0 | 0 | 0 | 0.985 | DAE | |
1 | 0.998 | 0.999 | 0.999 | Our proposed model | |
0.854 | 0.496 | 0.627 | 0.985 | SVM | Victim |
0.996 | 0.981 | 0.988 | 1 | KNN | |
0.999 | 0.999 | 0.999 | 0.999 | Naïve Bayes | |
0 | 0 | 0 | 0.985 | NN | |
0 | 0 | 0 | 0.985 | DNN | |
0 | 0 | 0 | 0.985 | DAE | |
1 | 0.998 | 0.999 | 0.999 | Our proposed model | |
0.307 | 0.064 | 0.106 | 0.689 | SVM | Unknown |
0.784 | 0.834 | 0.808 | 0.978 | KNN | |
0.880 | 0.603 | 0.715 | 0.958 | Naïve Bayes | |
0 | 0 | 0 | 0.940 | NN | |
0 | 0 | 0 | 0.967 | DNN | |
0 | 0 | 0 | 0.940 | DAE | |
1 | 0.995 | 0.997 | 0.999 | Our proposed model | |
1 | 0.754 | 0.177 | 0.318 | SVM | Suspicious |
0.980 | 0.982 | 0.981 | 0.972 | KNN | |
0.952 | 0.985 | 0.968 | 0.954 | Naïve Bayes | |
1 | 0.732 | 0.845 | 0.732 | NN | |
1 | 0.752 | 0.885 | 0.710 | DNN | |
1 | 0.732 | 0.845 | 0.732 | DAE | |
0.998 | 1 | 0.999 | 0.998 | Our proposed model |