The recent integration of vehicle-to-everything (V2X) communication, Internet of Things (IoT), and smart grid infrastructures in the transportation domain has resulted in significant security threats owing to the high degree of inter-connectedness, various data sources, and the need to ensure timely responses. Existing security techniques are mostly based on centralized approaches or domain-centric techniques. These approaches have resulted in scalability limitations, compromised data privacy, and lower robustness to new zero-day cyber threats. Although recent approaches have adopted the idea of federated learning to address the issue of data privacy, existing techniques have mostly been based on single domain-centric approaches. In order to deal with the aforesaid challenges, a new framework named ‘Cross-Domain Federated Adversarial Learning,’ abbreviated as ‘CFAL,’ has been conceived with the aim of facilitating unified intrusion detection for V2X, IoT, and Smart Grid Systems. The most important aim of the CFAL framework is to ensure the improved robustness against various types of attacks with minimal latency and maximum dependability. The CFAL framework has been suggested with the aim of integrating the Federated Learning technique with the Generative Adversarial Learning technique. Moreover, a cross-domain learning technique was utilized. The proposed CFAL framework was evaluated based on accuracy, zero-day detection rate, false alarm rate, latency, reliability, and availability. As presented in the experimental results, CFAL significantly improves the detection of zero-day attacks while ensuring a high accuracy rate, low false alarm rate, near-real-time system latency, and high system reliability and availability. This clearly shows that CFAL is a promising security solution for next-generation Cyber-Physical Systems.

Accurate software effort estimation is critical for effective project planning, resource allocation, and cost control. Yet, reliable prediction remains challenging due to the heterogeneous, noisy, and nonlinear characteristics of software project data, which often lead to schedule delays and cost overruns. This study presents a systematic empirical comparison of machine learning and ensemble-based models for software effort estimation, focusing on performance consistency, robustness across datasets, and the practical value of ensemble complexity under both tuned and untuned settings. An extensive experimental evaluation is conducted on five widely used benchmark datasets - DESHARNAIS, CHINA, ISBSG, COCOMO81, and MAXWELL - covering traditional single learners, strong tree-based ensembles, and stacking approaches. Models are evaluated using multiple accuracy and robustness metrics, including Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Median MRE, PRED (0.25), and Standardized Accuracy, with nonparametric statistical tests and Friedman's rank test applied to ensure a rigorous comparative analysis. The findings indicate that ensemble-based models consistently outperform traditional single learners across all datasets; however, model rankings remain largely stable between tuned and untuned configurations, suggesting that performance gains are not primarily driven by hyperparameter optimization. Among all evaluated methods, Extra Trees demonstrates the most robust and consistent performance with the best overall Friedman rank and minimal sensitivity to tuning, while stacking ensembles fail to provide statistically significant or consistent improvements despite higher computational cost. Overall, the results provide strong empirical evidence that well-designed tree-based ensemble models offer the best balance of accuracy, robustness, and efficiency, challenging the presumed advantages of increased ensemble complexity in practical software effort estimation.

Detecting and understanding visual semantics within natural scenes constitutes a pivotal research domain in pattern recognition and text analysis. Visual semantics involves understanding the pattern and layout of text in scene images. Text detection presents considerable challenges due to variations in size, color, font style, complex backgrounds, and low brightness levels. Although deep learning frameworks have significantly better performance over conventional techniques, the issue of text presented in arbitrary orientations amidst intricate backgrounds persists. This paper introduces a hybrid text segmentation method that tackles these issues by employing deep learning techniques. We propose a segmentation method that extracts features at multiple resolutions using VGG19 encoder unified with Attention Guided Adaptive Feature Pyramid Network (AG-AFPN) that captures richer multiscale representation. Furthermore, a Type-2 fuzzy logic approach serves to refine the edge map, thereby improving the accuracy of text boundary segmentation. For post-processing, Differential Binarization (DB) is applied to generate a precise binary mask from the network's output, thus enhancing segmentation performance by effectively managing variations in arbitrary text and cluttered backgrounds. The proposed method is assessed using multiple benchmark datasets, such as ICDAR 2013, ICDAR 2015, and Total Text.

Mission critical systems must present good performance during failures and dynamic changes. Classic stochastic performability models have in depth analysis but use static data for failure and repair rate which in turn decreases their accuracy. In this paper we present a new hybrid AI stochastic framework for performability evaluation in which we use machine learning models to dynamically put forth failure and repair rates that are put into a continuous time Markov chain (CTMC) based reward model. We present that this approach also has in-depth analysis as before but also is adaptive to different workloads and system aging. Also, we present simulation based on synthetic operation data which show our put forth framework does better than the fixed rate CTMC models. Our numbers show a 3.3% improvement in steady state performance, which we see greater changes in high workload stress and aging issues. We note that by combining data driven parameter estimates with stochastic modeling we present a very good solution for the performability analysis of mission critical systems.

Social media produces a huge amount of dirty data. This allows harmful posts and comments to spread easily and create social violence. It is hard to automatically detect and stop these harmful messages. Therefore, we propose a model for identifying expressions that harm the health of mental youth. We explore the potential of understanding and identifying multimodal posts on X (formally Twitter) depicting the encouragement of violence in tweets. Moreover, the proposed model is used to refine the parameters of the Convolutional Neural Network (CNN) to maximize its strength. We have evaluated the performance of multiple models like CNN, BERT, Multilayer Perceptron, Random Forest, Logistic regression, BERT+CNN, BERT+MLP and BERT+LSTM. Based on our evaluation, the accuracy of models is as follows: Logistic regression at 94.08%, Random forest at 93.84%, Decision tree at 91.70%, Naïve Bayes at 43.53%, BERT+CNN at 61.67%, BERT+MLP at 91% and LSTM+MLP at 95.44%. The results reveal that the BERT+LSTM model performs very well in comparison to other models showing excellent results as 95.44% accuracy and 95.57% F1-Score. LSTM gives better results than all other models in the detection of the abusive data.

Software testing has always been an essential pillar in ensuring software reliability and satisfaction of user requirements. Software systems are complex and require thorough testing to improve reliability and quality. However, manual test case design has a notorious history of being time-consuming and is subject to human error. Even most available automated methods are inflexible and require significant time, effort, and financial resources. Recently, search-based test data generation has become a significant, effective, and practical approach to overcoming these obstacles, and many meta-heuristic algorithms have been proposed to generate test cases to achieve branch coverage. Even though these strategies have shown good performance, researchers can further optimize these approaches. This paper proposes an automated test-case generation and optimization model that integrates activity diagram modelling with a Hybrid Adaptive Bat Particle Swarm Algorithm (ABPSA). Activity diagrams are used to represent the system's dynamic behavior, while the ABPSA is a synergistic combination of the exploratory nature of the Bat algorithm and the adaptive optimization of Particle Swarm Optimization. The algorithm is aimed at dynamically tracking the development of the activity-diagram model and thus effectively producing optimized test data. The effectiveness of the given framework is empirically demonstrated through a case study of the ATM withdrawal process.