Table 1 Comparative summary.
Paper | Scope / contribution | Strengths | Weaknesses | IoT suitability |
|---|---|---|---|---|
Core FL algorithm | Communication-efficient baseline; simple | Sensitive to non-IID data; poor heterogeneity handling | Baseline; needs adaptation | |
Heterogeneity-aware FL | Stable under non-IID; robust convergence | Hyperparameter tuning | Improved for IoT heterogeneity | |
Cryptographic FL aggregation | Strong confidentiality | Computational/key overhead | Promising w/ optimizations | |
Federated generative models | Local synthetic data generation | GAN instability; high communication | Experimental for IoT | |
Security of federated generative models | Highlights new attack surfaces | Few practical defenses | Critical concern | |
Hierarchical energy-aware FL | Reduced energy & communication | Trust dependency on cluster heads | High | |
Energy-efficient FL | Practical strategies: pruning, selective participation | Limited cross-IoT evaluation | Moderate; implementation needed | |
Blockchain-assisted FL | Transparency; incentives | Consensus overhead | Needs lightweight permissioned design | |
Backdoor/poisoning attacks | Demonstrates attack potency | Defense trade-offs | High concern | |
Federated anomaly detection | Privacy-preserving detection | Needs augmentation & robust defenses | High | |
Secure data sharing in 6G IoT healthcare | Data integrity, blockchain auditability | Energy efficiency not analyzed | Healthcare IoT; mid-resource devices | |
IoT intrusion detection with blockchain | High detection accuracy, secure key management | High computational cost, no data privacy | Mid-tier IoT devices | |
Blockchain-enhanced IDS | Tamper-resistant, distributed support | Energy/latency overhead, no generative AI | Distributed IoT; high-resource nodes | |
Explainable DL for CPS | High accuracy, interpretable | Limited privacy, computationally intensive | Industrial IoT; high-resource devices |
