Phishing presents a widespread and perilous menace within the domain of cybercrime, functioning as a rapid and uncomplicated means for cybercriminals to get confidential personal data. Phishing attacks specifically target unsuspecting persons with the intention of obtaining sensitive information, such as usernames, passwords, and financial credentials. Phishing websites frequently incorporate indicators inside their textual content and browser cookies. In order to address this issue, the suggested method makes use of the UCI Phishing Websites Data and implements categorization based on Extreme Learning Machine (ELM). In contrast to gradient-based algorithms, the ELM approach employs a solitary hidden layer of neurons that include input weights produced randomly. This design facilitates effective feature learning from data with a high number of dimensions, eliminating the need for iterative changes. The evaluation yielded a remarkable accuracy of 95.34% when compared to other machine learning techniques such as Random Forest Classification and Logistic Regression.

In this work, we present a thorough methodology that makes use of advanced machine learning algorithms and natural language processing tools to identify fake news. We used Latent Dirichlet Allocation (LDA) for topic modeling in order to identify latent topics within the articles. To depict the topics in a lower-dimensional space, dimensionality reduction techniques like T-Distributed Stochastic Neighbor embedding (t-SNE) are used. To discern between authentic and fake news stories, classification algorithms undergo training using topic representations. We assess the effectiveness of proposed method and obtain 95.64% accuracy, 1.0293 AUC, 0.9347 recall, 0.9595 precision, and 0.9466 F1-score. These outcomes demonstrate how well our methodology works to address this important problem.

In this research study, DNN are used to classify the images.With 30 various varieties of Balls used for this paper, the input data is mostly concentrated on the Balls category. Deep neural networks were chosen as the best alternative due to their high percentage of accuracy during the training phase. The percentage accuracy of image classification is used to describe the results. When the accuracy is around 89%, fine-tuning is used to increase the accuracy to 99% We spent a lot of time understanding the dataset and determining the shape and accuracy of each feature.collected information on 30 different feature types. 5 shapes were trained on each of the 3495 images in the 30 subdirectories for each type of feature.

The 3GPP (Rel.12 and Rel.16) introduces and explores the dual connectivity (DC). It is one of the most important future technologies adopted by 5G. It enhances the throughput and mobility robustness by enabling simultaneous dual connections with macro and small-cell evolved NodeBs (eNBs). While aggregating spectrum with the dual connections, DC induces two power-saving mechanisms of discontinuous reception (DRX). This results into offset synchronizations that decrease the feasibility of saving power. Therefore, the conventional DRX mechanism is not suitable for DC. Hence, we have proposed a synchronized adaptive DRX sleep (SAS) scheme for DRX that synchronizes its short-/long-sleep durations to make it suitable for active and background traffic. Next, we have evaluated the performance of SAS scheme with semi-Markov model in terms of DRX sleep time ratio and average delays. Moreover, in-depth analytical and simulation analyses of the DRX performance with DC have also been conducted. The accuracy of our model has been verified on the basis of a wide range of parameters. The proposed SAS scheme performs more effectively than the conventional DRX scheme for DC under different traffic types.

Future-generation radio access networks (RAN) are projected to fulfill the diverse requirements of user equipment (UE) by adopting a heterogeneous network (HetNet) environment. Necessary integration of different radio access technologies (RAT), such as 2G, 3G, 4G, wireless local area network (WLAN), and visible light communication (VLC) is inevitable. Moreover, UEs equipped with diverse requirements will be capable of accessing some or all the RATs. The complex HetNet environment with diverse requirements of UEs will present many challenges. The HetNet is likely to suffer severely from load imbalance among the base stations (BSs) from inheriting the traditional user association scheme such as max-SINR (signal-to-interference-plus-noise ratio)/max-RSSI (received signal strength indicator), unless some sophisticated schemes are invented. In this paper, a novel scheme is devised for a joint-user association for load balancing, where BSs are densely deployed and UEs typically have a certain degree of mobility. Unlike most of the present works, a dynamic network is considered where the position and channel condition of the UEs are not fixed. We develop two complex and distributed association schemes based on probability and d-choices, while carefully considering both loads of the BSs and SINR experienced by the UEs. Numerical results validate the efficiency of the proposed schemes by showing a received data-rate fairness among UEs and an improvement in the UE's minimum received data rate.

Future-generation radio access networks (RAN) are projected to fulfill the diverse requirements of user equipment (UE) by adopting a heterogeneous network (HetNet) environment. Necessary integration of different radio access technologies (RAT), such as 2G, 3G, 4G, wireless local area network (WLAN), and visible light communication (VLC) is inevitable. Moreover, UEs equipped with diverse requirements will be capable of accessing some or all the RATs. The complex HetNet environment with diverse requirements of UEs will present many challenges. The HetNet is likely to suffer severely from load imbalance among the base stations (BSs) from inheriting the traditional user association scheme such as max-SINR (signal-to-interference-plus-noise ratio)/max-RSSI (received signal strength indicator), unless some sophisticated schemes are invented. In this paper, a novel scheme is devised for a joint-user association for load balancing, where BSs are densely deployed and UEs typically have a certain degree of mobility. Unlike most of the present works, a dynamic network is considered where the position and channel condition of the UEs are not fixed. We develop two complex and distributed association schemes based on probability and d-choices, while carefully considering both loads of the BSs and SINR experienced by the UEs. Numerical results validate the efficiency of the proposed schemes by showing a received data-rate fairness among UEs and an improvement in the UE’s minimum received data rate.

Automation of logistics in pickup and delivery in hospital shows a great helping hand to the active workers. The application needs to allocate jobs to robots. The objective of the application is to produce an efficient system with the minimum total movement distance of all robots distance. In this paper, we propose two schemes, deep Hungarian (d-Hungarian) and deep Voronoi (d-Voronoi) job scheduling schemes. The d-Hungarian scheme is designed for one type of environment where the location of jobs is known prior whereas prior knowledge of jobs is not required in the d-Voronoi scheme. Simulation results show that our schemes efficiently reduce the traveled distance.