Federated learning (FL) with differential privacy (DP) is a promising distributed learning framework for protecting user privacy. However, the introduction of DP noise leads to a decrease in the accuracy of the FL model and may even hinder convergence. Improving model performance while maintaining privacy is a significant challenge in FL with DP. Existing works primarily focus on the fine-grained allocation of the same privacy budget in each global iteration of FL, without considering the impact of varying privacy budget per round on model performance. Preliminary experiments with manual allocation of non-uniform privacy budget suggest that FL model performance with variable privacy budget outperforms those with uniform privacy budget, providing a potential basis for dynamically adjusting privacy budget per round to enhance model performance. Additionally, client data heterogeneity and device heterogeneity are also crucial factors affecting FL model performance. In this paper, we investigate the joint optimization problem of client selection and overall privacy budget allocation. To address the significant challenges posed by the coupling of optimization variables and the inability to express the objective function of the optimization problem explicitly in terms of the optimization variables, we propose a Coarse-Grained Dynamic Differential Privacy (CGDDP) scheme based on the Dueling Double Deep Q Network (D3QN), modeling the optimization task as a Markov Decision Process. We carefully design a reward function that encourages higher model accuracy while penalizing excessive privacy budget allocation. By introducing the sequential composition theorem, we prove that CGDDP provides (ε, δ)-DΡ for the entire local model. Theoretical analysis confirms that the security of CGDDP is no less than that achieved by allocating the same privacy budget per round. Extensive experiments under various configurations validate the effectiveness of CGDDP. The results show that, compared to traditional methods, the proposed scheme improves global model prediction accuracy by 5.78% and 4.8% for non-IID MNIST and FashionMNIST datasets, respectively, and by 17.78% and 7.53% for IID data distributions.

The shortest paths problem is a fundamental challenge in graph theory, with a broad range of potential applications. The algorithms based on matrix multiplication exhibits excellent parallelism and scalability, but is constrained by high memory consumption and algorithmic complexity. Traditional shortest paths algorithms are limited by priority queues, such as BFS and Dijkstra algorithm, making the improvement of their parallelism a focal issue. We propose a matrix operation-optimized algorithm, which offers improved parallelism, reduced time complexity, and lower memory consumption. The novel algorithm requires O(Ewcc(i)) and O(Swcc · Ewcc) times for single-source and all-pairs shortest paths problems, respectively, where Swcc and Ewcc denote the number of nodes and edges included in the largest weakly connected component in graph. To evaluate the effectiveness of the novel algorithm, we tested it using graphs from SuiteSparse Matrix Collection and Gunrock benchmark dataset. Our algorithm outperformed the BFS implementations from Gunrock and GAP (the previous state-of-the-art solution), achieving an average speedup of 3.769 × and 9.410 ×, respectively.

Password guessing is a primary method for password strength evaluation. Despite various password guessing models have been proposed, there is still a significant gap between their guessing effectiveness and the actual cracking capabilities of attackers. Integrating multiple models for password guessing, also known as hybrid password guessing, could better capture the cracking capabilities of real attackers. However, the reason why hybrid password guessing can enhance cracking capabilities, and how to effectively integrate multiple heterogeneous password guessing models, are still not well understood. To address these issues, this paper draws inspiration from the concept of multi-view learning. We regard the guess lists generated by various password guessing models as multiple views of the data. Through a comprehensive analysis of these guess lists, we have identified the key reason why hybrid password guessing can enhance the cracking capabilities: integrating more diverse views allows for the coverage of a wider range of heterogeneous password characteristics, and provides more detailed information on effective password distributions. Based on the these findings, we propose a new hybrid password guessing framework, namedGuessFuse.GuessFuseemploys the multi-view subset extraction module and segment splitting selection module to accurately extract and reorganize the effective password from multiple guess lists. Experimental results on six large-scale datasets demonstrate the effectiveness ofGuessFuse. By combining two (resp. five) guess lists,GuessFuseoutperforms its foremost counterparts by an average of 11.00% ~ 59.62% (resp. 4.70% ~ 17.66%) within 107guesses.GuessFusecan effectively improve the cracking success rate under a limited number of guesses, approaching the actual cracking capabilities of attackers.

With the popularity and development of the Internet of things (IoT), human life has been deeply affected. Because of the limitations of computation capability and battery capacity, it is difficult for IoT devices to support frequent and complex computing. Motivated by this challenge, many works attempt to upload tasks of IoT devices to the cloud center for computation. However, because of the limitation of distance and bandwidth, cloud computing is difficult to guarantee low latency. As a feasible solution, Mobile Edge Computing (MEC) has attracted more and more attention. Most existing works focus on the computation offloading strategy, while the task scheduling on edge servers is not studied in depth. The tasks uploaded by IoT devices are dynamic and random, and there are dependencies between these tasks. Therefore, it is difficult for edge servers to find a task scheduling scheme to minimize the task execution delay. In this paper, to solve the task scheduling problem of edge server in multi-server and multi-user MEC system, we propose a heuristic algorithm based on the following three scenarios: 1) Tasks uploaded by IoT devices is dynamic and uncertain. 2) There are dependencies between tasks. 3) The computation capability of the edge server is limited. Experimental results show that the proposed algorithm can significantly reduce the overall completion time of tasks and the average task execution delay in the edge server.

Even though role-based access control (RBAC) can tremendously help us minimize the complexity in administering users, it is still needed to realize the notion of roles at the resource level. In this paper, we propose a practical cryptographic RBAC model, called role-key hierarchy model, to support various security features including signature and encryption based on role-key hierarchy. With the help of rich algebraic structure of elliptic curve, we introduce a role-based cryptosystem construction to verify the rationality and validity of our proposed model. Also, a proof-of-concept prototype implementation and performance evaluation are discussed to demonstrate the feasibility and efficiency of our mechanisms.

Provable data possession is a technique for ensuring the integrity of data in outsourcing storage service. In this paper, we propose a cooperative provable data possession scheme in hybrid clouds to support scalability of service and data migration, in which we consider the existence of multiple cloud service providers to cooperatively store and maintain the clients' data. Our experiments show that the verification of our scheme requires a small, constant amount of overhead, which minimizes communication complexity.

With the wide use of online social networks (OSNs), the problem of data privacy has attracted much attention. Several approaches have been proposed to address this issue. One of privacy management approaches for OSN leverages a key management technique to enable a user to simply post encrypted conte