The swift development of Internet of Things (IoT) technology has led to the expanded application of sensor networks in smart cities. Streetlights, as a vital aspect of smart city infrastructure, serve as a reflection of a town’s development. While streetlights are crucial for ensuring optimal visibility, road security, and public space utilization at night, they contribute significantly to electricity consumption. Governments across the globe are committed to reducing residential and industrial electricity usage. Regarding electricity usage, the Intelligent Street Lights System based on LED lights is a crucial concept today. This system functions by switching on the LED lights when a vehicle approaches and adjusting them to a dimmed or off state when no vehicle is present. The primary focus of this research paper, implementing an Intelligent Street Lights System based on LED lights, has emerged as a critical approach to achieving energy efficiency and cost reduction. Experimental findings demonstrate the potential for up to 80% energy savings compared to traditional streetlamp systems. Additionally, the system enables remote monitoring and intelligent management of urban street light conditions through terminal devices.

DeepFake technology is rapidly advancing, fundamentally changing how digital media is created and consumed. This evolution brings both exciting opportunities, such as enhanced entertainment and communication, and serious challenges, including misinformation, fraud, and threats to privacy and security. Using advanced AI methods such as deep learning and generative models, DeepFakes can create highly realistic but fake images, text, audio, and videos that are hard to identify. These synthetic media are often hard to distinguish from real content, which raises concerns about their misuse. This research presents a comprehensive review of the latest DeepFake detection methods across different types of media, including images, audio, video, and text. We investigate the application of deep learning models for detecting fraudulent information, including transformers, hybrid frameworks, convolutional neural networks (CNNs), and generative adversarial networks (GANs). By analyzing a wide range of studies, we compare these methods based on their accuracy, speed, and ability to perform well on various datasets such as FaceForensics++, DFDC, and DeepFakeTIMIT. One significant issue we identify is the lack of diversity in current datasets, which often leads to biased detection results and weak performance when models encounter new or real‐world DeepFakes. The present research emphasizes the need to develop better dataset standards that respect ethical and privacy concerns. Improving these areas will improve the reliability and generalization of detection systems. Our findings emphasize the need to develop adaptive detection models, privacy‐preserving techniques, and ethical frameworks to address emerging threats. The overall goal of this study is to give practitioners and scholars a comprehensive grasp of the state of DeepFake detection today. By identifying key challenges and suggesting promising future directions, we hope to support efforts to build safer digital environments where trust and authenticity are preserved.

The rapid digitization of financial service companies and organizations has heightened concerns over data security, particularly regarding unauthorized access to personal, financial, and password information. Among the various cyber threats, phishing attacks remain a pervasive and evolving challenge, exploiting deceptive emails, URLs, and SMS messages to illicitly harvest sensitive user data. In response, this research investigates a machine learning and deep learning–based framework for comprehensive phishing detection across multiple phishing datasets. We integrated diverse datasets, including phishing emails from Kaggle, malicious SMS messages from Mendeley, and both phishing and legitimate URLs from Kaggle and OpenPhish, to train the proposed models. Leveraging this carefully preprocessed and heterogeneous dataset, we systematically evaluate and compare the performance of nine state‐of‐the‐art machine learning and deep learning models. Furthermore, we propose an ensemble approach that synergizes the strengths of these models to enhance detection accuracy. Our experimental results demonstrate exceptional performance, with support vector machine (SVM) and bidirectional gated recurrent unit (BiGRU) models achieving the highest detection accuracies of 99.91% and 99.75%, respectively. The findings indicate that deep learning architectures generally outperform traditional machine learning techniques in identifying phishing attempts. This work not only provides a robust, multimodal phishing detection solution but also offers valuable insights for strengthening cybersecurity defenses in an increasingly digital world.

The advancements in technology have brought about significant changes in the automobile industry. A system that combines the control of a physical process with computing technology and communication networks is called a cyber–physical system (CPS). The enhancement of network communication has transitioned vehicles from purely mechanical to software-controlled technologies. The controller area network (CAN) bus protocol controls the communication network of autonomous vehicles. The convergence of technologies in autonomous vehicles (AVs) and connected vehicles (CVs) within Connected and Autonomous Vehicles (CAVs) leads to improved traffic flow, enhanced safety, and increased reliability. CAVs development and deployment have gained momentum, and many companies and research organizations have announced their initiatives and begun road trials. Governments worldwide have also implemented policies to facilitate and expedite the deployment of CAVs. Nevertheless, the issue of CAV cyber security has become a prevalent concern, representing a significant challenge in deploying CAVs. This study presents an intelligent cyber threat detection system (ICTDS) for CAV that utilizes transfer learning to detect cyberattacks on physical components of autonomous vehicles through their network infrastructure. The proposed security system was tested using an autonomous vehicle network dataset. The dataset was preprocessed and used to train and evaluate various pre-trained convolutional neural networks (CNNs), such as ResNet-50, MobileNetV2, AlexNet, GoogLeNet and YOLOV8. The proposed security system demonstrated exceptional performance, as demonstrated by its results in precision, recall, F1-score, and accuracy metrics. The system achieved an accuracy rate of 99.90%, indicating its high level of performance.

Deep fake technology has significantly advanced the creation of synthetic images and videos, sparking widespread concerns about its potential misuse in spreading misinformation, violating privacy, and enabling identity theft. As these manipulations be-come increasingly sophisticated, the development of reliable detection methods has become a pressing necessity. This research tackles this challenge by proposing a robust deep fake detection pipeline, leveraging a custom dataset created using Roboflow. The dataset is divided into two primary classes: real and fake, with the fake class further categorized into three subtypes based on complexity: easy fake, mid fake, and hard fake. Easy fake images involve basic manipulations that are easily identifiable by the human eye, while mid fake images combine AI-generated and human-generated elements, and hard fake images are entirely AI-generated, posing significant challenges for detection. To ensure authenticity and diversity, real images were collected from personal networks and online repositories. We trained and evaluated four YOLO-based models YOLOv8, YOLOv9, YOLOv10, and YOLOv11 for the detection task. YOLOv8 emerged as the top-performing model, achieving an accuracy of 96.2% in distinguishing between real and fake images. Finally, addressing ethical considerations and developing countermeasures to mitigate the societal impact of deep fakes should remain a priority for future research.