Table 23.
Comparative analysis with recent State-of-the-Art Works.
| Approach (Year, Citation) | Key Techniques | Dataset(s) | Metrics | Blockchain | Adaptive/Transfer Learning | IOC from Unstructured Text | Threat Diversity | Focus vs. Ours |
|---|---|---|---|---|---|---|---|---|
| OTI-IoT (Aguru & Erukala, 2024)6 | Blockchain-based operational threat intelligence | Simulated IoT data | Detection 95%, FPR 3% | Yes | No | No | Multi-vector DDoS | Blockchain-based OTI for DDoS |
| Collaborative Threat Intelligence (Nazir et al., 2024)26 | Blockchain + ML ensemble | CIC-IDS2017 | Accuracy 93%, Detection 91% | Yes | Partial | No | General IoT attacks | Collaborative blockchain-ML IDS |
| Leveraging ML for Industry 4.0 (Yu et al., 2024)14 | ML & DL models | KDD Cup 1999, others | Accuracy 91%, Detection 89% | No | No | No | Industry 4.0 threats | ML challenges & resilience |
| Intelligent Hybrid IoT IDS (Elsedimy & AboHashish, 2025)31 | Fuzzy C-means + Sperm Whale Algorithm + Hybrid ML | CIC-IDS2017 | Accuracy 94%, F1 92% | No | No | No | General IoT attacks | Metaheuristic-based IDS |
| Optimized IDS with GAO-XGBoost & ECC (Nandanwar & Katarya, 2025)39 | GAO-optimized XGBoost; ECC-integrated blockchain | IoT traffic flows | Detection ≈ 98% | Yes (ECC) | No | No | Multi-class IoT attacks | Flow-based IDS + secure storage |
| Hybrid Blockchain-Based IDS (Nandanwar & Katarya, 2025)40 | Hybrid blockchain securing IDS | Heterogeneous IoT | High integrity/accuracy | Yes (hybrid) | Partial | No | General IoT intrusions | Decentralized IDS security |
| Securing Industry 5.0 CPS (Nandanwar & Katarya, 2025)41 | XAI-enhanced deep learning | CPS-specific datasets | High accuracy + explainability | No | No | No | CPS intrusions | Explainable CPS-IDS |
| Privacy-Preserving IDS with Blockchain + FL (Nandanwar & Katarya, 2024)[45 | Federated Learning + hybrid blockchain | IIoT/IoT | 95–98% accuracy | Yes | Yes (FL) | No | Multi-class IoT | Distributed privacy-aware IDS |
| TL-BiLSTM IoT (Nandanwar & Katarya, 2024)43 | Transfer-learning BiLSTM | IoT botnet datasets | High detection on botnets | No | Yes (transfer) | No | Botnet-specific | Botnet prediction |
| Proposed Framework | BERT–spaCy–regex hybrid NLP; confidence-weighted ensemble (BERT/LSTM/NB); lightweight ledger | CIC-IDS2017, UNSW-NB15 | IOC extraction: Acc ≈ 95%, F1 ≈ 95.7%; Traffic classification: Acc/Prec/Rec/F1 ≈ 94.7–94.8% | Yes (lightweight) | Yes (incremental) | Yes | Binary DoS + multi-class cross-validation | Unstructured CTI extraction + secure sharing |