Table 1 Summary of the BC-enabled XAI framework for cyber threat detection.
Ref. | Techniques | Metrics | Findings |
|---|---|---|---|
PSO, RF, DT, KNN, SMOTE-Tomek, LIME, SHAP | Accuracy, Precision, Recall, F1-Score | The proposed IDS outperforms existing methods, ensuring robust WSN security. | |
Curriculum Learning, LIME, NN, Staged Learning, Quantization & Pruning, Ensemble Model (RF, XGBoost) | Accuracy | The framework presents high accuracy and robustness, ensuring reliable IoT network security. | |
Ensemble Learning, TL, Feature Engineering, LIME, SHAP | Accuracy, Detection Rate Increase, Training Time Reduction | The framework improves detection accuracy, adaptability, and transparency for IDS in diverse environments. | |
XC-TDF, Adversarial Training, Regularization, XAI | Accuracy, Precision, Recall, F1-Score, MCC | XC-TDF enhances robustness, transparency, and accuracy, resisting adversarial attacks and noise. | |
X-NET, X-AI, NB, LR, Perceptron | Accuracy, Precision, Recall, F1-Score, ROC | X-NET enhances security and performance for remote patient monitoring utilizing X-AI models. | |
ML, XAI, SHAP, LIME, LLM, Gemini, OPENAI | Standard Metrics | The framework improves IoT attack detection and response by utilizing XAI and LLM for better accuracy and insights. | |
Consumer IoT, XAI, IDS, BC, SHAP, Python-based Framework | Benchmark Metrics | The framework improves IoT security in smart cities utilizing XAI, SHAP, and BC. | |
BC, XAI, C-PoA, PSLSTM, Multi-head attention | Accuracy, Precision, Recall, F1-Score | Enhances decision support for cybersecurity analysts in smart healthcare systems. | |
RF, SVM, CNN-LSTM, SHAP, LIME | Accuracy, Precision, Recall, F1-Score | The model attained highest accuracy of 99.9%; XAI enhanced model interpretability. | |
Adaptive CNN-GRU, DL, Botnet Attack Classification | Accuracy, Loss, Precision, Recall | AttackNet attained 99.75% accuracy, outperforming existing methods; future work will address real-time scalability and dataset diversity. | |
GRANet, HBSF, Swarm Intelligence Tuning | Precision, Recall, Detection Rate | The framework presents superior detection. | |
Hybrid CNN-BiLSTM, TL | Testing and Training Accuracy, Loss | The model attains 99.52% accuracy, surpassing existing methods. | |
DL, IDS, SHAP | Accuracy, Precision, Recall, Loss | The model attained high detection performance with interpretability. | |
ML, DL | Security Enhancement, Vulnerability Identification | The review identifies gaps and suggests combined techniques to enhance IoT security. | |
BC, Smart Contracts, NIZKP, IPFS, IDS | Security Efficiency, Privacy Preservation, Storage Cost Reduction, System Scalability | The model improves healthcare security and privacy with cost savings. | |
BiLSTM, S-AAM, SHAP, XAI | Accuracy, Dataset Efficiency, Interpretability | Achieved 99.92% and 96.54% accuracy. | |
AGLSTM, RCHO, SMOTE | Accuracy, Sensitivity and Specificity Rate, Training Percentage, K-Fold Validation | AGLSTM performs well across datasets. | |
GRU-GWO, LSTM-GWO | Detection Accuracy, Dimensionality Reduction (DR), Computational Efficiency | Enhanced accuracy and efficiency with GRU-GWO and LSTM-GWO models. |
