| Method | Feature learner | Classifier | Anomaly | Accuracy |
| SVE + CDBN [31] | CDBN RBM | FCNN | FDI | > 94% |
| Stacked LSTM + Bloom Filter [8] | Stacked LSTM | Softmax | Data injection Command injection Reconnaissance DoS | 92% |
| Stacked AE + Softmax [23] | SAE | Softmax | DoS Probe R2L U2R | 97.6% 86.34% 12.98% 39.62% |
| CNN/LSTM + FCNN [28] | CNN/LSTM | FCNN | 36 attacks | 92% (F1-score) |
| DAE-RBM [33] | DAE | DAE Residuals | Operating anomaly detection Fault analysis | N/A |
| SAE + MLP [27] | SAE | MLP | Data injection Comman injection Relay setting modification | 96% |
| SDAE + ELM [35] | SDAE | ELM | Operating faults | 99% |