Although federated learning is gaining prevalence in smart sensor networks, substantial risks to data privacy and security persist. An improper application of federated learning techniques can lead to critical privacy breaches. Practical and effective privacy-enhanced federated learning (PEPFL) is a widely used federated learning framework characterized by low communication overhead and efficient encryption and decryption processes. Initially, our analysis scrutinized security vulnerabilities within the PEPFL framework and identified an effective attack strategy. This strategy enables the server to derive private keys from content uploaded by participants, achieving a 100% success rate in extracting participants’ private information. Moreover, when the number of participants does not exceed 300, the attack time does not surpass 3.72 s. Secondly, this paper proposes a federated learning model that integrates homomorphic encryption and secret sharing. By using secret sharing among participants instead of secure multi-party computation, the amount of effective information available to servers is reduced, thereby effectively preventing servers from inferring participants’ private gradients. Finally, the scheme was validated through experiments, and it was found to significantly reduce the inherent collusion risks unique to the federated learning scenario. Moreover, even if some participants are unavailable, the reconstructable nature of secret sharing ensures that the decryption process can continue uninterrupted, allowing the remaining users to proceed with further training. Importantly, our proposed scheme exerts a negligible impact on the accuracy of model training.

Due to the difficulty of existing container scheduling algorithms to adapt to large-scale complex scenarios and meet the diverse application and load requirements, this study delves into a groundbreaking hybrid scheduling approach that melds Deep Deterministic Policy Gradient (DDPG) with a Genetic Algorithm (GA). The proposed method initially employs a container grouping policy to reduce the overhead associated with frequent inter-container calls. Subsequently, to address the computational inefficiency of large-scale scheduling, a genetic algorithm is utilized for rapid global optimization. To overcome the problem of genetic algorithms being highly susceptible to falling into local optimality, the DDPG algorithm is applied for local optimization, with a cross-mutation operation introduced to escape local optima. The experimental outcomes demonstrate that the proposed algorithm enhances cluster load balancing by 81.13%, improves the fitness function by 3.26%, reduces completion time by 19.06%, and decreases container dependency overhead by 2.75%. Furthermore, under the experimental conditions, the system performs best when the group size is 10. This research offers a novel paradigm for the development of container scheduling algorithms in distributed intelligent data clouds, advancing the field of resource management in cloud computing environments.

For federated learning on IoT devices, instability of user-side nodes can cause sudden disconnection of participants, and lead to inaccurate training results. Moreover, user disconnection can cause the entire joint training process to stall. Building on the basis of the xMK-CKKS multi-key homomorphic encryption, this paper proposes an improved scheme ftxMK-CKKS, which increases fault tolerance of federated learning and can deal with the problem of training stagnation caused by user disconnection. Experiments shows that the scheme works with negligible time and communication cost compared to the former scheme based on Shamir secret sharing.

Federated learning has become an effective method to realize collaborative learning among multiple data centers. However, the problem of privacy leakage has not been perfectly solved. We proposed a scheme based on homomorphic encryption with random privacy masks to cope with the problem of private leakage caused by honest but curious participants. Through combining Paillier encryption scheme with privacy mask, security of the learning process is enhanced with a trivial computation and communication cost. Experimental results show that the decryption time is reduced by more than half after using CRT optimization, without degrading the training precision.

Utilizing the micro-Doppler effect caused by the rotation of rotor blades to invert parameters such as spatial position and rotational speed of targets is of great significance for the detection and identification of unmanned aerial vehicles. The vortex electro-magnetic wave radar can directly obtain target azimuth information. However, most existing techniques for vortex electromagnetic wave radar are based on uniform motion models and can only extract partial parameters of the target, making it difficult to achieve precise characterization of the target. This article proposes a 3D parameter extraction algorithm for a rotor variable speed moving target. Firstly, the echo model of a rotor target is presented. Secondly, the separation of different rotor information was achieved based on the nearest neighbor criterion. Lastly, a set of equations containing the relationships between target structural, motion, and position parameters were established. By solving these equations, the extraction of target rotor length, wheelbase, rotational speed, acceleration, spatial position, attitude angle, and other parameters was achieved. The effectiveness and robustness of the proposed method were verified through simulation.