Blockchained service provisioning and malicious node detection via federated learning in scalable Internet of Sensor Things networks
- Publisher:
- Elsevier
- Publication Type:
- Journal Article
- Citation:
- Computer Networks, 2022, 204, pp. 108691
- Issue Date:
- 2022-02-26
Closed Access
| Filename | Description | Size | |||
|---|---|---|---|---|---|
| 1-s2.0-S1389128621005570-main.pdf | 1.87 MB |
Copyright Clearance Process
- Recently Added
- In Progress
- Closed Access
This item is closed access and not available.
In this paper, a blockchained Beyond Fifth Generation (B5G) enabled malicious node detection model is proposed for the Internet of Sensor Things (IoSTs). Moreover, a secure service provisioning scheme using cascading encryption and feature evaluation process is also proposed for the IoSTs. The presence of malicious nodes causes severe issues in the localization and service provisioning, which discourages new entities to join the network. Therefore, it is very important to establish trust between all entities by detecting and removing such nodes. The proposed B5G enabled malicious node detection model uses federated learning for the detection of malicious nodes. The federated learning uses Support Vector Machine (SVM) and Random Forest (RF) classifiers to detect the malicious nodes. The malicious nodes are classified on the bases of their honesty and end-to-end delay. Moreover, the service provider nodes provide services to each other and get the reward. However, the service provisioning in the IoSTs has many issues like a repudiation of service providers as well as the clients. The feature evaluation and cascading encryption mechanisms are used to solve these issues. The digital signature in cascading encryption ensures the non-repudiation of the service provider. On the other hand, feature evaluation of service ensures that the client cannot repudiate about actually demanded services. Moreover, the conformance of services is also ensured by the feature evaluation process. The simulation results show the effectiveness of our proposed non-repudiation model. The SVM and RF classifiers are compared in terms of accuracy, precision, F1 score and recall. The accuracy, precision, F1 score and recall of SVM are 79%, 1, 0.8795 and 0.78, respectively. On the other hand, the accuracy, precision, F1 score and recall of RF classifier are 95%, 0.92, 0.96 and 1, respectively. The results show that RF has better accuracy than RF in malicious nodes detection.
Please use this identifier to cite or link to this item: