Performance testing occupies an advantageous status in software testing since software performance is closely related to software quality and user experience. Performance defects, however, are often more latent with complex action mechanisms, making performance testing challenging. In this work, we invite chaos engineering and propose an automated performance testing system to discover performance defects and ensure software quality. The main idea of our system is to inject faults artificially to simulate real scenarios to see whether the system can still perform well even under bad conditions. In total, there are two main roles in our system: scenario designer and fault manager, in which the former is to design possible application scenarios of the system, while the latter is to inject faults according to the scenarios. Testing is conducted under real conditions so that defects can be figured out more easily. Experiments have demonstrated its rationality and validity.

The emerging Internet of Things (IoT) 5G smart cities are experiencing radical changes in transportation, utilities and energy by real-time interconnection, automation, and data sharing. Gone are the days when we could just protect these vulnerable assets on the physical network layer. As things are increasingly open for attack, we can no longer afford being under protected. To mitigate the above problems, this paper puts forward a lightweight scalable Hybrid Deep Learning-Based Security Architecture designed to secure IoT systems in 5G Smart City environments. The proposed model leverages DNN for the spatial feature extraction and LSTM networks to capture the temporal relationships in network traffic. Transfer learning is leveraged in the architecture to improve flexibility and detection accuracy by reutilizing prior learned knowledge and only retraining to account for new threat vectors. The model is trained and tested using benchmark datasets that cover diverse real-world attack landscapes like CICIDS-2018 and UNSW-NB15. According to the results of the experiments, the filtered model yields a low false positive rate of 1.08%, the AUC-ROC is 0.987 and exhibits good performance measures, such as an accuracy of 99.74%, precision of 98.21%, recall of 99.89%, and F1-score of 98.05%. These results are evidence of the model's real-time intrusion detection capability, making the framework a feasible and effective technique for securing the 5G-enabled IoT infrastructures of smart cities.

The Electric Vehicle Routing Problem (EVRP) is a complex combinatorial optimization problem that arises in sustainable transportation logistics. This paper presents the design and implementation of a multi-objective global-global hybrid algorithmic approach to optimize the Constrained Electric Vehicle Routing Problem. The objective of this work is to develop a hybrid multi-objective approach by integrating parameter-free algorithms and Swarm-based Optimization techniques. The performance of the proposed approach evaluated by solving the Multi-Objective constrained Electric Vehicle Routing Problem. The Multi-Objective hybrid approach is designed by integrating MOPSO and the MOJaya algorithm. The proposed approach extends the concept of non-dominated sorting with a ranking scheme. Hypervolume and IGD performance metrics are used to evaluate the performance of the proposed approach. The Multi-Objective constrained Electric Vehicle Routing Problem incorporates unique constraints such as limited battery capacity, charging station availability, and time windows. Since the problem is multi-objective, it seeks to optimize multiple conflicting objectives, including minimizing operational costs, energy consumption, and fleet size. The proposed approach succeeds in obtaining feasible solutions for the selected problem. The results obtained by the multi-objective hybrid PSO-Jaya algorithm are comparatively better than those produced by the multi-objective PSO and multi-objective Jaya algorithm in terms of Hypervolume and IGD values. The proposed approach demonstrates the potential of hybridization of two global search techniques in optimizing the multi-objective problem. The results also encourage further application of this approach to real-time multi-objective optimization problems.

This paper presents ELAREES, a task scheduling algorithm for heterogeneous multiprocessor real-time systems, designed to optimize energy savings while enhancing fault tolerance. ELAREES addresses the dual challenges of fault tolerance in task execution and communication reliability between tasks, alongside efficient power management. The algorithm employs a primary/backup strategy, assigning each task a primary execution on a low-power (LP) core and a backup on a high-performance (HP) core to ensure resilience against execution faults. Furthermore, ELAREES integrates a robust communication protocol that monitors data transmission over shared media connection buses, dynamically selecting optimal transmission paths and initiating retransmissions when necessary to mitigate communication errors. By leveraging Dynamic Voltage and Frequency Scaling (DVFS) and Dynamic Power Management (DPM) techniques, ELAREES achieves significant power savings while maintaining high system reliability. Simulation results demonstrate consistent power savings of approximately 30% across various scenarios, with only a minimal impact of 0.02% on reliability. This research contributes to the field of energy-efficient computing in real-time systems, offering a comprehensive solution for managing the trade-offs between energy consumption, execution fault tolerance, and communication reliability in heterogeneous multicore environments.

A session-based recommender system (SBRS) focuses on users' interests depending on their browsing habits to make appropriate recommendations in a working session. Nowadays, graph neural networks are a very popular approach for the Session-based Recommendation System. In existing works, a session is treated as a single time-point moving user model and avoids the complex association of the items. Moreover, the temporal component of user representation learning is disregarded by the conventional graph convolutional network (GCN); the resulting user preference model is static and unable to capture the dynamism of user preferences. Also, their works overlook the time factor. In this study, we present a time-aware Session-Based Lightweight Graph Convolutional network (SB-LGCN) that employs several GNN techniques to effectively capture both static and dynamic user preferences and minimize the complexity. Our goal is to streamline GCN's design to make it more succinct and suitable for recommendation. The Proposed Light Graph Convolutional Network (Light GCN) model includes only the most essential component in GCN neighborhood aggregation. It learns user and item embeddings by linearly propagating them on the user-item interaction graph and uses the normalized sum of the neighbors' embeddings learned at all the layers as the final embedding. Finally, a weighted sum aggregator is used to achieve the prediction. The performance is verified by extensive experiments on the SB-LGCN model for the Movie Lens and YooChoose1/64 datasets. Results indicated that the proposed model outperforms the best by accuracy with impressive training time efficiency.

Despite widespread use of cloud computing, one of its main challenges is to maintain efficient load balancing and resource allocation. Numerous well-established load balancing techniques have been implemented to optimize task scheduling. The research objectives are equitable resource distribution, overloading prevention, and response time reduction in cloud computing response delays. This study has successfully achieved a comparative performance analysis of four load balancing algorithms — Random, EMPBT-LB, Throttled, and a Proposed Load Balancing (LB) algorithm — evaluated based on their response time performance. The study emphasizes load balancing, where incoming user jobs must be distributed among several virtual machines (VMs) located across different data centers (DCs). The Proposed LB algorithm demonstrates significant improvements over conventional methods by employing a hybrid decision-making approach that incorporates previous processing data, adaptive randomness, and current workload status. The simulation results in the proposed approach (a) achieving the lowest average response time (b) eliminating maximum processing delays, and (c) maintaining high consistency under varying workloads.