Table 1.
A comparative analysis of the proposed work with the existing state-of-the-art works.
| Author | Year | Objective | Methodology | 1 | 2 | 3 | 4 | 5 | Pros | Cons |
|---|---|---|---|---|---|---|---|---|---|---|
| Chang et al. [17] | 2019 | Anomaly detection in ICS | K-means and convolutional autoencoder | X | X | X | X | X | The adopted approaches gives higher accuracy on a gas pipeline and water storage tank dataset | PCA and XAI has not considered |
| Alhaidari et al. [20] | 2019 | Secure supervisory control and data acquisition systems against DDoS attack | NB, RF, and J48 | X | X | ✔ | X | X | Among all three approach RF gives higher accuracy | They have focused on the feature selection |
| Elnour et al. [23] | 2020 | Hybrid attack detection scheme for water treatment plant | Isolation Forest and CNN | X | X | ✔ | X | X | The scheme detects maximum attacks, reduces the computational complexity and increases the accuracy compared with the other approaches | Not discussed about data poisoning attack on the dataset |
| Rakesh et al. [24] | 2021 | Monitor water quality using ML approach | NB, RF, and LR | X | X | X | X | X | The scheme compares the simulation results with the experimental results | Not considered feature selection |
| Khan et al. [28] | 2021 | Facilitate resource-efficient solution to the IoT application | ANN, SVM, DT, and DELM | X | X | X | ✔ | X | The scheme offer security and protection to the smart home | Not considered data positioning attack and IPFS-based storage |
| Puthal et al. [25] | 2022 | User-centric security and fake data identification for IoT-based critical infrastructure | DT | X | X | X | X | X | They proposed theoretical and experimental solution that resist brute force, DDoS, and replay attack | Not compared the DT results with other AI approaches such as SVM, RF, XGBoost, etc. |
| Narayanan et al. [26] | 2022 | Anomalies detection in smart cyber-physical systems | Automatic behavioural abstraction technique based on neural networks | X | X | ✔ | X | X | The scheme detects the maximum number of attack with 1 percent a false positive rate | They have not taken feature selection approaches and data poisoning attack |
| Ragab et al. [31] | 2022 | To secure the industrial control system | SVM, RF, Adaboost, KNN, and BDLE-CAD | X | X | X | ✔ | X | They applied chimp optimization based feature selection that increases the accuracy | Not considered data poisoning attack for dataset |
| Gu et al. [29] | 2023 | Quality control in manufacturing process | KNN, XGboost, ANN, XGB Max | X | X | X | ✔ | X | The proposed scheme prevents DoS, man in the top, DDoS, and brute force. | Not considered data poisoning attack. |
| The proposed architecture | 2023 | Secure data dissemination architecture | RF, DT, SVM, perceptron, and GaussianNB classifier | ✔ | ✔ | ✔ | ✔ | ✔ | Accurate, efficient, secure, and reliable architecture for IoT-based critical infrastructure | - |
Parameters- 1: PCA, 2: XAI, 3: Anomaly detection, 4: Conventional blockchain, 5: IPFS-based blockchain.