Image encryption is a critical and attractive issue in digital image processing that has gained approval and interest of many researchers in the world. A proposed hybrid encryption method was implemented by using the combination of the Nahrain chaotic map with a well-known optimised algorithm namely the grey wolf optimisation (GWO). It was noted from analysing the results of the experiments conducted on the new hybrid algorithm, that it gave strong resistance against expected statistical invasion as well as brute force. Several statistical analyses were carried out and showed that the average entropy of the encrypted images is near to its ideal information entropy.

Intelligent Transportation Systems (ITS) is one of the challenging research areas revolving around rapid communication services. Nonetheless, these networks are afflicted by congestion, routing, and security issues, particularly in urban areas, which result in delay and computational complexity issues and late message delivery or data congestion. Sensitive data are collected and transmitted over public channels, which presents challenges for security and authentication. Moreover, the security of ITS-smart vehicle communication is impacted by the complexity of discrete logarithm and factoring problems, which could make data transmission and authentication very difficult in the presence of highly scalable quantum computers. Hence, we present a novel, secure and verifiable post-quantum data transmission and authentication protocol for ITS (RAVEN), specifically designed for smart automobiles. The RAVEN scheme incorporates a discrete Gaussian distribution and lattice-based cryptosystem for enhanced performance and security. We use the formal verification method SVO logic and an informal security analysis against passive and active assaults to validate the security of mutual authentication. The use of a well-known automated AVISPA tool allowed for a further evaluation of the scheme’s security soundness. The OMNeT++ simulator is used to evaluate the efficacy of the RAVEN scheme in terms of end-to-end delay, throughput, and energy consumption. The evaluation findings demonstrate that the suggested method achieves reduced costs in terms of computation and communication compared to existing schemes but with comprehensive security that can greatly benefit ITS.

The rise of bogus news on social media necessitates robust detection techniques. This is especially important in Arabic-speaking online communities, where disinformation is common. In this study, Dragonfly-based feature selection and ensemble learning models are used to detect Arabic fake news on Twitter. Based on the concept of dragonfly foraging, the suggested method intelligently selects and prioritizes key elements in Arabic tweets. This intentional feature selection improves feature discrimination and computational efficiency. Additionally, ensemble learning models analyze the selected attributes cooperatively. The ensemble technique in this paper comprises four models: decision tree, XGBoost, gradient boosting, and LGBM, to enhance detection accuracy and robustness by utilizing these classification algorithms. Empirical tests on a large Arabic tweet dataset demonstrate the effectiveness of the methodology. The Dragonfly-based feature selection method highlights the most informative traits, thereby improving fake news detection accuracy. Ensemble learning models are more resilient to noisy and deceptive data, enhancing the system’s ability to identify fake news. Dragonfly-based feature selection and ensemble learning models enhance the identification of Arabic fake news. This approach contributes to facilitating more credible and well-informed digital dialogue within Arabic-speaking online communities, extending beyond its immediate applications. The results show that improving the ensemble models using the Dragonfly algorithm achieved the highest accuracy of 90.55 when employing XGBoost. This paper provides powerful tools to combat misinformation and promote the spread of factual information as false news continues to evolve.

The identification and prediction of coding sequences in DNA are intricate tasks that continue to captivate the field of Bioinformatics. It is a multifaceted challenge where Signal Processing methods have demonstrated a substantial impact. Extensive research has been conducted in this area, leading to the development of numerous computational techniques. While many existing review papers have offered summaries of these techniques, most have been confined to specific methodologies or viewpoints. This paper seeks to extend beyond these limitations by providing a concise but comprehensive overview of coding sequence prediction methods. Furthermore, it aims to present distilled technical guidance for potential future research. Lastly, this review offers insights into emerging trends and innovative methods that are poised to influence and enhance the direction of future studies in this area.

Understanding DNA’s structural and functional properties requires examining periodic patterns in genomic sequences. Exon differentiation in eukaryotic DNA is facilitated by the three-base periodicity (TBP) phenomenon, which is common in protein-coding areas. This TBP feature in the spectrum domain is used by digital signal processing (DSP) approaches such as the Anti-Notch Filter (ANF) and Short-Time Discrete Fourier Transform (STDFT) for exon-intron separation. Many approaches that target this particular characteristic have been developed over the last 20 years, but their efficacy is still not at its best. A thorough grasp of three-base periodicity and how it affects exon identification is necessary to improve the predicted accuracy of these techniques. Instead of introducing novel feature extraction techniques, this paper thoroughly analyzes 3-base periodicity and its importance in DSP-based exon identification techniques. It has been shown that period-3 traits are less prominent in shorter exons, which makes prediction difficult. This research also provides a comparative examination of ANF and STDFT approaches in this context.

In the field of classification, input vectors are labeled with one of several predetermined classes. Automated systems scour a wide range of industries—from e-commerce to news outlets to blogs to directories to content curators—in search of applicable applications. To make suggestions about what to read and how to find it, natural language processing (NLP) uses document classification. Due to the difficulties that big datasets with many categories can provide for conventional document categorization approaches, this research uses machine learning for multi-class document classification. The suggested approach utilizes feature extraction and machine learning approaches to solve these issues. Features such as word frequency (TF) and term frequency-inverse document frequency (TF-IDF) are retrieved from documents to represent textual content, enabling the model to classify features based on word semantics. The method utilizes the chi-square test and the Bat Algorithm to choose features optimally. Random Forest (RF), Stochastic Gradient Descent (SGD), Adaptive Boosting (AdaBoost), and Cat Boost are only a few of the cutting-edge machine-learning algorithms used in the multi-class classification approach. These algorithms, which are trained to take advantage of the retrieved features, perform well when presented with multi-class document categorization challenges. Extensive testing on various datasets has validated the efficacy of the suggested methodology, demonstrating its improved accuracy and processing efficiency compared to traditional methods. Among the various available options, the Random Forest (RF) algorithm has proven to be the most effective, with an accuracy rate of 87.57%. This method’s flexibility makes it an invaluable tool in the NLP field since it may be successfully used in various document classification tasks.

Both natural language processing and information retrieval rely on document classification. With the exponential growth of digital documents, there is an increasing demand for accurate and efficient document classification techniques. However, document classification faces several challenges, including processing vast amounts of data and extracting features from large volumes of unstructured data, which is complex. Moreover, numerous input features can be extracted, but not all of them are relevant to the classification problem. Single-classifier models often have limitations in capturing the complex patterns and nuances present in diverse document collections. In this paper, we propose a new approach to enhance document classification through the utilization of ensemble techniques. We combine multiple classifiers, including SVM, Naïve Bayes, and AdaBoost as base models, and aggregate their predictions with a neural network (NN) as a meta-model to improve overall classification performance. The suggested method aims to enhance accuracy, robustness, and generalization in classification by harnessing the diversity and complementary strengths of different classifiers. This research contributes to the field of document classification by utilizing the AdaBoost algorithm as a base model in stacking techniques applicable to document classification. The experimental results demonstrate that the proposed method outperforms the most recent methods in classifying documents in terms of accuracy, precision, and recall. The accuracy of the proposed system reached 87%, and the F1 score reached 87.8%.