Table 1 Literature review of existing methods.
Author and citation | Approach | Blockchain platform | Key technologies | Advantages | Disadvantages |
|---|---|---|---|---|---|
Sezer et al.14 | PPFchain | Blockchain | sensor-IoT-based architectures | Precision, effectiveness, and improved security are provided by PPFchain | Real-time data traceability, availability, and low latency |
Li et al.15 | ADDetector, DP | NA | IoT, FL based scheme | The aggregation procedure among clients and the cloud should be protected by a FL-based system | Data transmission leakage of the model and raw data |
Ngan Van et al.,16 | PriFL-Chain, DP | Blockchain | FL based scheme, MEC | Minimize the expense of ML model training, efficiently safeguard privacy, and make use of diverse data sources | Large-scale deployment of IoT devices is difficult and expensive |
Madill et al.17 | ScaleSFL | Blockchain-based sharing- Hyperledger Fabric | FL | Maintain efficiency and security while increasing validation performance linearly | Provide the appearance of malevolent system assaults |
Cui et al.18 | BCFL, smart contract | Blockchain | FL | Minimize the training loss, and training time of classifier | Communication load of BFL is higher |
Lo et al.19 | Trustworthy federated learning based on blockchain | Blockchain | FL | Enable accountability and improve fairness | The amount of samples every client has during classification decreases as the number of customers declines |
Miao et al.20 | PBFL | Blockchain | FL | Achieves convergence and provides privacy protection. | The suggested model has not examined the communication impose of PBFL |
Tian et al.21 | RPDFL, Ring-All reduce-based data sharing | Blockchain | FL | According to security studies, RPDFL has improved accuracy, convergence, and is provably secure | Real-time applications are not supported |
Guduri et al.22 | blockchain-based lightweight encryption strategy | Blockchain | Cloud computing | Proposed system very effective proxy re-encryption mechanism with FL | Strive for credibility and embrace responsible AI concepts |
Ouyang et al.23 | FL based on blockchain | Blockchain | RFID, FL, IoT | Benefits related to decentralization, privacy, and security. | Doesn’t support for large-scale applications |
Wang et al.24 | BCFL, two-stage Stackelberg game | Blockchain | FL | Clients in determining the lesser computational time. | Malicious clients and servers have the ability to poison federated learning systems |
Yazdinejad et al.25 | AP2FL | NA | ActPerFL | Robustness, fairness, openness, and integrity of the FL process. Improved precision and successful removal of privacy leaks | Inadequate privacy protection for larger datasets |
Hu et al.26 | FL-HMChain, FL-CNN-HMChain | Blockchain | FL | Ensuring that the cooperative model training procedure is generally dependable and trustworthy. The suggested model significantly reduces the probability of privacy leaking | More attention must be paid to developing more accurate and dependable prediction models for the medical field |
Tian et al.27 | PEFL, DP | Blockchain | MFF | PEFL demonstrates better defense against various attack models. Proposed system achieves higher training efficiency while ensuring privacy security | Less support for efficiency, and reliability in data sharing |
Asad and Otoum28 | BPPFL | Blockchain | IoT, FL | Reduces computation and communication overhead. Highest model accuracy and robust privacy guarantees | Enhancing throughput and reducing latency in the blockchain’s infrastructure |
