Toward Robust Differential Privacy in Adaptive Federated Learning Architectures

The convergence of vehicular technology, artificial intelligence (AI), and distributed computing has catalyzed the emergence of the Internet of Vehicles (IoV) as a cornerstone of next-generation intelligent transportation systems (ITS). By enabling vehicle-to-everything (V2X) communication, IoV supports cooperative perception, real-time decision-making, and autonomous driving. However, the reliance on large-scale, data-driven intelligence in IoV exposes systems to critical challenges, including adversarial poisoning, privacy leakage, identity forgery, and the fragility of centralized learning architectures. Federated Learning (FL) has been proposed as a promising paradigm to alleviate some of these issues by enabling distributed model training without centralizing sensitive vehicular data. Nonetheless, conventional FL remains vulnerable to security and trust limitations, particularly in dynamic vehicular environments. This thesis addresses these challenges by designing secure, privacy-preserving, and scalable FL frameworks that leverage distributed ledger technologies and cutting-edge security mechanisms.The thesis advances knowledge through four interconnected contributions. First, two novel optimization-driven poisoning attack models are introduced: PA-PSOSA and PAPSOGA, which combine particle swarm optimization with simulated annealing and genetic algorithms, respectively. These models demonstrate that even a small poisoning budget can substantially degrade global model utility under black-box and clean-label constraints, highlighting the urgency of robust defenses in vehicular FL. Second, a permissioned blockchain-enabled FL (BCFL) framework is proposed, in which consortium edge nodes running Practical Byzantine Fault Tolerance (PBFT) consensus replace the central aggregator. With blockchain integration and data validation mechanisms, this design ensures identity authentication, verifiable audit trails, and improved resilience against poisoning and Sybil attacks, while maintaining high model accuracy under adversarial conditions. Third, the framework is further enhanced to achieve inference-resistance by integrating secure aggregation (SecAgg) and differential privacy (DP), and lightweight with off-chain commitments. This design significantly reduces ledger storage requirements, increases system throughput, and mitigates inference-based privacy risks. Finally, to overcome the scalability limitations of PBFT-based BCFL, a DAG-enabled FL (DFL) framework is developed. By leveraging parallel validation, utility-score-based tip selection, and reputation-weighted aggregation, this framework significantly improves scalability, reduces communication complexity, and enhances robustness in asynchronous vehicular environments.Together, these contributions articulate a coherent progression from exposing vulnerabilities in vehicular FL to constructing secure, privacy-preserving, and scalable frameworks tailored for IoV ecosystems. The findings demonstrate that interdisciplinary integration of optimization theory, cryptography, differential privacy, and distributed ledger technologies is indispensable for trustworthy vehicular intelligence. Beyond theoretical significance, the proposed frameworks offer practical designs for deployment in safety-critical IoV environments. Future research directions include the integration of zero-knowledge proofs (ZKP) for verifiable privacy, adaptive defenses against evolving adversarial strategies, and experimental validation in real-world vehicular testbeds. Collectively, this thesis establishes a foundation for secure federated intelligence in IoV, contributing to the reliability, efficiency, and trustworthiness of next-generation ITS.

Federated learning (FL) has become a promising technology to provide edge Artificial Intelligence (AI) due to its advantages in privacy protection and reduced communication costs. However, FL is still confronted with privacy leakage issues because the sharing local model may expose the training data information. Existing works typically utilize secure aggregation techniques to eliminate privacy leakage, where local model parameters in FL are obfuscated before they are sent to the aggregator. Nevertheless, secure aggregation makes poisoning attacks more convenient given that existing anomaly detection methods mostly require access to plaintext local models. A Poisoning-Defense Secure Aggregation in FL (PDSA-FL) is proposed to enhance the privacy protection of honest clients and defend against poisoning attacks from malicious clients. Specifically, a Secure Aggregation scheme based on Random Parameters Decomposition (SARPD) is designed to protect client privacy during the FL aggregation process and eliminates the impact of dropped clients on the aggregation results. Secondly, a Poisoning Detection method based on Similarity Grouping (PDSG) is proposed to mitigate the impact of poisoning attacks on the global model of FL without leaking client model parameters. The security analysis discusses the effectiveness of the proposed PDSA-FL in terms of privacy protection. Extensive simulation results show that PDSA-FL can effectively defend against poisoning attacks, significantly improve the convergence performance of global models, and reduce the computation time of clients.

Privacy-preserving Federated Learning (FL) based on Differential Privacy (DP) protects clients’ data by adding DP noise to samples’ gradients and has emerged as a de facto standard for data privacy in FL. However, the accuracy of global models in DP-based FL may be reduced significantly when rogue clients occur who deviate from the preset DP-based FL approaches and selfishly inject excessive DP noise beyond expectations, thereby applying a smaller privacy budget in the DP mechanism to ensure a higher level of security. Existing DP-based FL fails to prevent such attacks as they are imperceptible. Under the DP-based FL system and random Gaussian noise, the local model parameters of the rogue clients and the honest clients have identical distributions. In particular, the rogue local models show a low performance, but directly filtering out lower-performance local models compromises the generalizability of global models, as local models trained on scarce data also behave with low performance in the early epoch. In this paper, we propose ReFL, a novel privacy-preserving FL system that enforces DP and avoids the accuracy reduction of global models caused by excessive DP noise of rogue clients. Based on the observation that rogue local models with excessive DP noise and honest local models trained on scarce data have different performance patterns in long-term training epochs, we propose a long-term contribution incentives scheme to evaluate clients’ reputations and identify rogue clients. Furthermore, we design a reputation-based aggregation to avoid the damage of rogue clients’ models on the global model accuracy, based on the incentive reputation. Extensive experiments demonstrate ReFL guarantees the global model accuracy performance 0.77% - 81.71% higher than existing DP-based FL methods in the presence of rogue clients.

With the continuous development of vehicular ad hoc networks (VANET) security, using federated learning (FL) to deploy intrusion detection models in VANET has attracted considerable attention. Compared to conventional centralized learning, FL retains local training private data, thus protecting privacy. However, sensitive information about the training data can still be inferred from the shared model parameters in FL. Differential privacy (DP) is sophisticated technique to mitigate such attacks. A key challenge of implementing DP in FL is that non-selectively adding DP noise can adversely affect model accuracy, while having many perturbed parameters also increases privacy budget consumption and communication costs for detection models. To address this challenge, we propose FFIDS, a FL algorithm integrating model parameter pruning with differential privacy. It employs a parameter pruning technique based on the Fisher Information Matrix to reduce the privacy budget consumption per iteration while ensuring no accuracy loss. Specifically, FFIDS evaluates parameter importance and prunes unimportant parameters to generate compact sub-models, while recording the positions of parameters in each sub-model. This not only reduces model size to lower communication costs, but also maintains accuracy stability. DP noise is then added to the sub-models. By not perturbing unimportant parameters, more budget can be reserved to retain important parameters for more iterations. Finally, the server can promptly recover the sub-models using the parameter position information and complete aggregation. Extensive experiments on two public datasets and two F2MD simulation datasets have validated the utility and superior performance of the FFIDS algorithm.

HierFedPDP:Hierarchical federated learning with personalized differential privacy

With the continuous improvement of hardware computing power, edge computing of industrial data has been gradually applied. In the past decade, the promotion of edge computing has also greatly improved the efficiency of industrial production. Compared with the conventional cloud computing, it not only saves the bandwidth consumption of data transmission, but also ensures the terminal data security to a certain extent. However, the continuous update of attack types also put forward new requirements for the privacy protection of industrial edge computing. So it should fundamentally solve the risk of industrial data leakage in the process of deep model training in edge terminal. In this article, we propose a new federated edge learning framework based on hybrid differential privacy and adaptive compression for industrial data processing. Specifically, it first completes the adaptive gradient compression preparation, then constructs the industrial federated learning model, and finally makes use of adaptive differential privacy model to optimize, so as to complete the privacy protection towards the transmission of gradient parameters in industrial environment. By optimizing the hybrid differential privacy and adaptive compression, we can better prevent the terminal data privacy against inference attacks. The experimental results show that this method is very effective in the industrial edge computing situation, and it also opens up a new direction for the effect of differential privacy in federated learning.

In this paper, the convergence time of federated learning (FL), when deployed over a realistic wireless network, is studied. In particular, a wireless network is considered in which wireless users transmit their local FL models (trained using their locally collected data) to a base station (BS). The BS, acting as a central controller, generates a global FL model using the received local FL models and broadcasts it back to all users. Due to the limited number of resource blocks (RBs) in a wireless network, only a subset of users can be selected to transmit their local FL model parameters to the BS at each learning step. Moreover, since each user has unique training data samples, the BS prefers to include all local user FL models to generate a converged global FL model. Hence, the FL training loss and convergence time will be significantly affected by the user selection scheme. Therefore, it is necessary to design an appropriate user selection scheme that can select the users who can contribute toward improving the FL convergence speed more frequently. This joint learning, wireless resource allocation, and user selection problem is formulated as an optimization problem whose goal is to minimize the FL convergence time and the FL training loss. To solve this problem, a probabilistic user selection scheme is proposed such that the BS is connected to the users whose local FL models have significant effects on the global FL model with high probabilities. Given the user selection policy, the uplink RB allocation can be determined. To further reduce the FL convergence time, artificial neural networks (ANNs) are used to estimate the local FL models of the users that are not allocated any RBs for local FL model transmission at each given learning step, which enables the BS to improve the global model, the FL convergence speed, and the training loss. Simulation results show that the proposed approach can reduce the FL convergence time by up to 56% and improve the accuracy of identifying handwritten digits by up to 3%, compared to a standard FL algorithm.

Traditional federated learning (FL) protects privacy by exchanging model updates instead of raw data. However, when applied to continuous medical behavior data, existing methods face two major challenges. First, standard FL frameworks remain vulnerable to advanced privacy attacks such as membership inference and attribute inference, since model gradients may still reveal individual behavioral patterns or sensitive medical events. Second, most differential privacy (DP) mechanisms, such as Gaussian noise injection, focus on global model protection and fail to ensure user-level trajectory privacy across the entire training process. To address these limitations, this study proposes FedHealthGuard, a user-level trajectory differential privacy federated learning framework tailored for continuous medical behavior data. FedHealthGuard introduces a trajectory-centric privacy paradigm, treating each participant’s sequence of model updates throughout the training process as a private trajecto ry. Comprehensive evaluations on three real-world continuous medical datasets (MIMIC-III, PAMAP2, ECG-PTB) show that FedHealthGuard attains strong user-level trajectory privacy (ɛ< 2.0) while preserving predictive utility. These results confirm that FedHealthGuard effectively balances privacy protection strength and predictive utility. This study provides an essential technical foundation for building the next generation of trustworthy, privacy-compliant, and high-performance distributed AI systems.

Decentralized IoT data sharing: A blockchain-based federated learning approach with joint optimizations for efficiency and privacy

Federated learning (FL) is an attractive scheme for training machine learning models across distributed edge devices without requiring raw data to be shared. However, deploying FL at the edge of the net- work raises two interrelated issues: (1) the significant communication overhead due to transmitting massive gradient vectors from resource- constrained devices and (2) the risk of privacy leakage via attacks that can invert gradients or infer training samples. Current methods treat gradient compression and differential privacy (DP) as separate concerns, leading to compounded accuracy loss. We propose AGC-DP—Adaptive Gradient Compression (AGC) with Differential Privacy (DP)—a uni- fied framework that achieves efficient communication and strong pri- vacy guarantees across diverse edge FL environments. AGC-DP also introduces a resource-aware client selection algorithm to balance con- tributions from battery level, available memory, and uplink bandwidth; an adaptive compression ratio for each client based on the optimization problem; and a privacy amplification construction that exploits sub- sampling to improve DP accounting. We provide rigorous theoretical analysis, including convergence guarantees under non-i.i.d. data dis- tributions and formal (ϵ, δ)- DP proofs. Comprehensive experiments across the CIFAR-10, HAR, and FEMNIST datasets with up to 100 edge clients show that AGC-DP reduces communication overhead by up to 91.1% compared to vanilla FedAvg, while maintaining accuracy within 1.5% of the vanilla baseline. Our results show that privacy, com- munication efficiency, and model utility can be achieved simultaneously in resource-constrained edge FL.

In this paper, we focus on the privacy preserving mechanism design for crowdsourced Federated Learning (FL), where a requester can outsource its model training task to some workers via an FL platform. A potential way to preserve the privacy of workers' local data is to leverage Differential Privacy (DP) mechanisms on local models. However, most of these studies cannot allow workers to dominate their own privacy protection levels by themselves. Thus, we propose a Personalized Privacy Preserving Mechanism, called P3M, to satisfy the heterogeneous privacy needs of workers, which consists of two parts. The first part includes a personalized privacy budget determination problem. We model it as a two-stage Stackelberg game, derive the personalized privacy budget for each worker and the optimal payment for the requester, and prove that they form a unique Stackelberg equilibrium. Second, we design a dynamic perturbation scheme to perturb model parameters. Through the theoretical analysis, we prove that P3M satisfies the desired DP property, and derive the bounds of the variance of average perturbed parameters and the convergence upper bound. This demonstrates that the global model accuracy can be controllable and P3M is endowed with the satisfactory convergence performance. In addition, we extend our problem to the scenario where the total privacy budget of all workers is limited, so as to prevent some workers from setting exorbitant privacy budgets. Under the privacy constraint, we re-determine the personalized privacy budget for each worker. Finally, exhaustive simulations of P3M are conducted based on real-world datasets, and the experimental results corroborate its effectiveness and practicability.

By taking advantage of Deep Learning (DL), Received signal strength (RSS) fingerprint-based indoor localization has attracted more attention. Training DL models require an immense amount of RSS samples, and crowdsourcing has been another way to collect data by users. Recently, some researchers apply Federated Learning (FL) in indoor localization instead of crowdsourcing for privacy protection. However, more practical issues are still not considered.For further guaranteeing privacy and deploying systems in real time, this paper proposes OPFL, a privacy-preserved online personalized federated learning framework for fingerprint-based indoor localization. In OPFL, users update model parameters and collect new labeled samples meanwhile. By adding artificial noise based on the Differential Privacy (DP) mechanism, it is harder for attackers to infer information. Besides, the deterioration of models results from DP is mitigating by applying the personalized algorithm. The experimental results show that under the noise perturbing, the performance of OPFL testing on the local test set is far better than the conventional federated averaging algorithm (FedAvg), and is even close to the centralized training.

This artilce aims to develop a differential private federated learning (FL) scheme with the least artificial noises added while minimizing the energy consumption of participating mobile devices. By observing that some communication efficient FL approaches and even the nature of wireless communications contribute to the differential privacy (DP) preservation of training data on mobile devices, in this paper, we propose to jointly leverage gradient compression techniques (i.e., gradient quantization and sparsification) and additive white Gaussian noises (AWGN) in wireless channels to develop a piggyback DP approach for FL over mobile devices. Even with the piggyback DP approach, information distortion caused by gradient compression and noise perturbation may slow down FL convergence, which in turn consumes more energy of mobile devices for local computing and model update communications. Thus, we theoretically analyze FL convergence and formulate an energy efficient FL optimization under piggyback DP, transmission power, and FL convergence constraints. Furthermore, we propose an efficient iterative algorithm where closed-form solutions for artificial DP noise and power control are derived. Extensive simulation and experimental results demonstrate the effectiveness of the proposed scheme in terms of energy efficiency and privacy preservation.

Normalization is an important but understudied challenge in privacy-related application domains such as federated learning (FL), differential privacy (DP), and differentially private federated learning (DP-FL). While the unsuitability of batch normalization for these domains has already been shown, the impact of other normalization methods on the performance of federated or differentially private models is not well-known. To address this, we draw a performance comparison among layer normalization (LayerNorm), group normalization (GroupNorm), and the recently proposed kernel normalization (KernelNorm) in FL, DP, and DP-FL settings. Our results indicate LayerNorm and GroupNorm provide no performance gain compared to the baseline (i.e. no normalization) for shallow models in FL and DP. They, on the other hand, considerably enhance the performance of shallow models in DP-FL and deeper models in FL and DP. KernelNorm, moreover, significantly outperforms its competitors in terms of accuracy and convergence rate (or communication efficiency) for both shallow and deeper models in all considered learning environments. Given these key observations, we propose a kernel normalized ResNet architecture called KNResNet-13 for differentially private learning. Using the proposed architecture, we provide new state-of-the-art accuracy values on the CIFAR-10 and Imagenette datasets, when trained from scratch.

As a novel distributed machine learning approach, federated learning (FL) is proposed to train a global model while preserving data privacy. However, some studies manifest that adversaries can still recover private information from the gradients. Differential privacy (DP) is a rigorous mathematical tool to protect records in a database against leakage. It has been widely applied in FL by perturbing the gradients. Nevertheless, while using DP in FL, the convergence performance of the global model is inevitably degraded. In this paper, we implement a DP-based FL scheme, which achieves local DP (LDP) by adding well-designed Gaussian noise on the gradients before clients upload them to the server. After that, we propose two strategies to improve the convergence performance of the DP-based FL. Both methods are realized by modifying the local objective function to limit the effect of LDP noise on convergence without degrading the privacy protection level. We then provide the detailed framework which adopts the LDP scheme and two strategies. The framework on different machine learning models is tested by simulation results, which show that our framework can improve the convergence performance up to 40% faster under different noise compared with other DP-based FL. Finally, we show the theoretical convergence guarantee of our proposed framework by first presenting the expected decrease in the global loss function for one round of training and then providing the upper convergence bound after multiple communication rounds.