Convolutional neural networks (CNNs) and self-attention (SA) have demonstrated remarkable success in low-level vision tasks, such as image super-resolution, deraining, and dehazing. The former excels in acquiring local connections with translation equivariance, while the latter is better at capturing long-range dependencies. However, both CNNs and Transformers suffer from individual limitations, such as limited receptive field and weak diversity representation of CNNs during low efficiency and weak local relation learning of SA. To this end, we propose a multi-scale fusion and decomposition network (MFDNet) for rain perturbation removal, which unifies the merits of these two architectures while maintaining both effectiveness and efficiency. To achieve the decomposition and association of rain and rain-free features, we introduce an asymmetrical scheme designed as a dual-path mutual representation network that enables iterative refinement. Additionally, we incorporate high-efficiency convolutions throughout the network and use resolution rescaling to balance computational complexity with performance. Comprehensive evaluations show that the proposed approach outperforms most of the latest SOTA deraining methods and is versatile and robust in various image restoration tasks, including underwater image enhancement, image dehazing, and low-light image enhancement. The source codes and pretrained models are available at https://github.com/qwangg/MFDNet .

In mobile crowdsensing activities, it is usually necessary for participants to upload sensing data and related locations. The existing location privacy-preserving mechanisms cannot well protect a user's trajectory privacy because attackers can mine the user's trajectory features through data analysis techniques. Aiming at the trajectory privacy protection problem, this article proposes a differential location privacy-preserving mechanism based on trajectory obfuscation (LPMT). LPMT first extracts the stay points as the features of a trajectory based on the sliding window algorithm, and then obfuscates each stay point to a target obfuscation subregion through the exponential mechanism, and finally performs the Laplace sampling in the target obfuscation subregion to obtain the obfuscated GPS points. Compared with the baseline mechanisms, LPMT can reduce data quality loss by more than 20% while providing the same level of obfuscation quality, which indicates that LPMT has the advantages of strong security and high quality of service.

In smart cities, intelligent learning environment is an important application scenario, and class attendance checking is an important measure to urge students to attend on time and ensure the quality of learning. Aiming at the existing problems in class attendance checking, such as low efficiency and easy to cheat, this paper proposes a student attendance management method named AMMoC (Attendance Management Method based on Crowdsensing). AMMoC includes two phases, i.e., the initialization phase and the authentication phase. In the initialization phase, a teacher sends an attendance checking request to the server. After receiving the request, the server sends a request to tell students to submit their location information, and then forms the student location map once the server receives all the response from students. In the authentication phase, the server verifies the truth of the location information by sending requests to several students to count the number of students. The authentication phase includes two modules, i.e., the task assignment module and the attendance verification module. In the task assignment module, AMMoC first finds the optimized sequence of subregions and verifiers by using the Monte Carlo algorithm, and then requires the verifiers to count the number of students in the subregion. Finally, the statistics results will be verified in the attendance verification module. Experiment comparisons and analyses show that AMMoC has the advantages of good anti-cheating performance, fast speed, and little disturbance to class, and is suitable for attendance checking applications in classroom environment.

Location privacy is one critical factor concerned by mobile users of Location Based Service (LBS). Location privacy of mobile users includes two parts: current position information and trajectory information. The problem of position anonymity in Ad Hoc & sensor networks has been widely studied during past years. Nevertheless, most of existing privacy preserving solutions either need a third part server or require a large number of users. This paper proposes a simple and distributed trajectory preserving strategy, called Virtual Avatar (VAvatar). VAvater adopts a virtual path programming resolution, which does not need any third-part server (agent), to achieve trajectory information anonymity for continuous querying users of LBS. On the other hand, attackers may have strong ability to make mobility predictions by analyzing spatio-temporal relation of historical communication data. In order to prevent users from such threatening, VAvatar takes following two strategies: noisy location selecting strategy and query scheduling strategy to interrupt spacial and temporal relationships between position data. Theoretical analysis and experiments show that VAvatar provides efficient protection of trajectory privacy for mobile users with tolerable energy cost.

Emerging energy harvesting technology can harvest and convert ambient energy into electrical power. It is a quite effective way to extend the lifetime of Wireless Multihop Networks (WMNs), and has been widely applied in WMNs. Meanwhile, the interference-free link scheduling is an indispensable operation in WMNs to reduce energy waste caused by link collision. In the energy harvesting WMNs with link scheduling, a few works take into account the impact of non-ideal battery on energy efficiency and network lifetime. This paper studies the harvested energy-saving link scheduling problem in the energy harvesting WMNs under the imperfect storage efficiency. We establish the energy model to index the link weight with the harvested energy consumed on it. Furthermore, we propose a Centralized Scheduling (CS) algorithm, a Centralized Scheduling with Minimum interference weights (CSM) algorithm and a distributed one to activate the links with the maximum weights so as to maximize the utilization of the harvested energy and to minimize the energy stored in the battery. Finally, this paper conducts the simulation experiment to verify the effectiveness of our algorithms and the correctness of our theoretical analysis.