Due to the increasing adoption of transfer learning approaches, Natural Language Processing (NLP) has undergone a significant transformation particularly in the field of text classification. Through the application of transfer learning, many pre-trained transformer models, such as BERT, distilBERT, and RoBERTa, have emerged and shifted attention towards NLP. These transformer models are already trained on huge corpora of unlabeled text, helping them to deeply recognize semantics, grammar, and contextual relationships. Unlike other deep learning models, transformers are not dependent on large task-specific datasets. Instead, they support pre-training information to generalize effectively across several NLP tasks like text classification. In this study, we utilized two datasets: the IMDB dataset for binary text classification and AG News dataset for multi-class classification. To achieve our objective, we have implemented five transformer models - BERT, distilBERT, RoBERTa, ALBERT and XLNET - on both datasets to train our model. Among these models, RoBERTa achieved the highest accuracy of 95.4% on the IMDB dataset and 94.2% on the AG news dataset. Although with such state-of-the-art performance, the RoBERTa model emerged as the most effective and robust technique for text classification technique as the confusion matrix also confirmed its superior precision and balanced classification abilities.

Along with the proliferation of blockchain technology, ensuring the anonymity and confidentiality of transactions has become increasingly critical. Recently, blockchain networks remain vulnerable to various attacks targeting sensitive data. These attacks can severely undermine the security of an organization’s network and lead to sensitive information leakage, compromising privacy and confidentiality. To address these challenges, we introduce RingZk, an advanced privacy-preserving framework designed to enhance the privacy and security of blockchain. RingZk employs Ring Signature to sign a transaction in a manner that conceals their identity within a group of possible signers. Additionally, RingZk leverages Zero Knowledge Proofs to cryptographically secure transaction data, guaranteeing that sensitive information remains secure. This approach not only enhances transaction security but also fortifies the network against threats. The effectiveness of RingZk is assessed through a series of experiments focused on key privacy attributes, and it is compared against existing state-of-art techniques. The results reveal how our approach surpasses current solutions and is effective in addressing privacy and security issues in the realm of blockchain transactions.

Assaults known as Distributed Denial-of-Service (DDoS) overload target servers, services, or networks with malicious data, making them unavailable to authorized users. These assaults are becoming a bigger problem, impacting governments, corporations, and web applications. DDoS assaults are predicted to become far more frequent and sophisticated when 5G and other future network technologies are introduced and expanded. An improved, effective, and secure detection system is therefore desperately needed to safeguard vital network equipment and enable the smooth rollout of 5G networks. A novel DDoS detection system that combines a composite multilayer perceptron (MLP) with an efficient feature extraction approach is developed in order to meet this difficulty. The methodology distinguishes between harmful and benign activity using a dataset created from simulated network traffic. Support Vector Machines (SVM) and K-Nearest Neighbors (KNN) algorithms were used to examine network patterns in order to classify the data. Furthermore, the system uses two different Deep Neural Networks (DNN1 and DNN2), each of which has a special architecture intended to improve the detection process’s accuracy and resilience. Comparing the suggested approach to conventional classifiers reveals remarkable performance. According to experimental data, the framework achieves a remarkable accuracy rate of 99.71% and a wonderfully low loss of 0.011. This performance demonstrates the promise of deep learning approaches to improve security of networks, since it greatly outperforms SVM and KNN classifiers. The created DDoS detection solution guarantees safe, effective, and dependable 5G network operations without creating new vulnerabilities in addition to providing high accuracy and little performance loss. In the age of sophisticated wireless communication networks, this framework is an essential first step in reducing DDoS threats.

Recognizing user emotions is one of the most important factors for improving user satisfaction in music recommendation systems. This research explains emotion-based music recommendation systems, their use cases, and technologies involved in their creation. Relating to the growing size of digital music collections and increased access to streaming services, emotional understanding of music is important for providing emotionally relevant and personalized recommendations. This research analyzes the different emotion recognition methods in music, including acoustic, lyrics, and hybrid approaches, and discusses the effectiveness of these approaches. It analyzes the influence of emotional content on user engagement, user satisfaction, and playlist desirability. Moreover, we address issues of building efficient emotion-based recommenders, such as data annotation, cultural divergences of emotions, and emotion model interpretability. This research addresses the future of emotion-informed music recommendation systems, such as the fusion of physiological information or signals and modern machine learning methods. The overarching aim is to generate music recommendations for the user that are not only relevant according to their choices but also their feelings at the given moment. Thus, this paper answers the call regarding the development of music recommenders in the contemporary world.

The application of large language models (LLMs) in automating job posting extraction and generating personalized cold emails offers a promising approach to improving the efficiency and effectiveness of recruitment and marketing processes. The increasing reliance on email marketing in sales outreach and HR assistance has created a need for more efficient and effective email generation tools. Traditional methods of crafting personalized emails can be time-consuming and labor-intensive, leading to decreased productivity and efficiency. A prompt-based approach is used to fine-tune the LLM, enabling it to accurately extract relevant information from job postings, including job title, experience, skills, and description. The model leverages the LLM's ability to process unstructured data and generate human-like text, producing cold emails tailored to the specific needs of clients. The results demonstrate that the model achieves high accuracy in extracting job posting information and generating relevant cold emails, indicating its potential to improve response rates, conversion rates and provide personalized email content.

LoRaWAN has emerged as a leading Low Power Wide Area Network (LPWAN) technology due to its long-range communication and low energy consumption. However, as network size and node density increase, challenges such as energy inefficiency, suboptimal spreading factor (SF) utilization, and reduced scalability arise, especially in multi-hop scenarios. This paper proposes an integrated framework that addresses these issues through three key innovations: an energy-aware multi-hop routing protocol, a dynamic SF allocation mechanism, and a scalability enhancement model. The proposed scheme optimizes path selection based on residual energy and link quality, effectively extending network lifetime. Concurrently, the SF allocation algorithm dynamically adapts to node density and link conditions, ensuring balanced transmission reliability. To assess scalability, we introduce a simulation-based analysis under varying node densities and traffic loads. Experimental results demonstrate significant improvements in energy efficiency, reduced packet collisions, and enhanced overall network throughput. Our approach provides a comprehensive solution for sustainable and scalable deployment of large-scale LoRaWAN-based IoT applications.