Table 1 Existing literature on privacy issues in wireless networks and cyber-physical systems.
From: Federated learning with enhanced cryptographic security for vehicular cyber-physical systems
References | Focus Area | Approach | Contributions | Pros | Cons | Metrics |
|---|---|---|---|---|---|---|
Differential Privacy | Output perturbation during training | Introduces differential privacy techniques to protect user data during the training phase of machine learning models | Enhances privacy by adding noise; Easy to implement | May reduce model accuracy due to added noise | Privacy budget, model accuracy | |
Differential Privacy | Objective perturbation during training | Enhances privacy by adding noise to the objective function during model training, thus preserving data privacy | Strong privacy guarantees; Suitable for convex optimization problems | Limited to specific types of optimization problems | Privacy budget, model accuracy | |
Privacy in Data Aggregation | Bit-choosing algorithm | Protects user privacy in data aggregation tasks by selectively choosing bits, reducing the risk of data leakage | Reduces data leakage risk; Low computational overhead | May not be effective for all types of data | Privacy leakage probability | |
Resource Allocation and Privacy | Encryption resource allocation | Allocates encryption resources based on privacy weight and execution time, improving privacy protection during data processing | Balances between privacy and computational efficiency | May require significant computational resources | Execution time, privacy weight | |
Secure Data Transmission | Trusted authority schemes | Proposes a trusted authority scheme for secure data transmission, particularly focusing on vehicular networks | Provides strong security guarantees; Well-suited for vehicular networks | Relies on a central authority, which may become a single point of failure | Data transmission security level | |
Other Works | Resource Allocation and Data Sharing | Various AI and ML approaches | Focus on optimizing resource allocation and data sharing using AI techniques like deep reinforcement learning and deep Q-learning without addressing security and privacy concerns | Optimizes resource usage and efficiency; Enhances system performance | Often overlook security and privacy aspects | Privacy level, frequency of pseudonym changes |
Proposed Work | Data Privacy in VCPS | Federated learning with a two-phase scheme | Proposes an intelligent data transformation and collaborative data leakage detection scheme | Preserves data privacy; Effective against various types of data leakage | Complexity in implementation and coordination among multiple entities | Data utility, data leakage detection accuracy, privacy level |
