The telecommunications industry has a significant challenge in customer churn, especially for companies aiming for sustainable growth in a competitive market. To overcome this issue, we suggested a method utilizing Long Short-Term Memory (LSTM) networks, a deep learning strategy that has revolutionized predictive modeling. We used 4250 customer profiles, each containing 19 features including consumption data and plan options, to train and test the LSTM model. Data preprocessing included encoding categorical variables, normalizing numerical features, and eliminating irrelevant variables to maintain data integrity and authenticity. The LSTM model was fine-tuned to recognize significant patterns rather than simply memorizing data. We also incorporated an additional layer to assess diverse information types, which improved the model's ability to process sequential data effectively. The proposed model achieved a prediction accuracy of 95.3% in identifying potential customer churn, demonstrating its effectiveness in predicting when customers might leave their telecommunications provider. These results demonstrate how LSTM networks can reduce customer attrition in the telecom industry, making them a valuable tool for companies aiming to retain customers and maintain growth in a competitive market. The study's findings pave the way for innovative applications of deep learning techniques in customer retention strategies across diverse industries. This breakthrough has the potential to transform business approaches to customer relationship management and loyalty programs.

This study proposes a practical technique for detecting false information in both textual and video formats utilizing the eXtreme gradient boost (XGBoost) and Convolutional Neural Networks (CNNs) classification techniques. We increase a robust and flexible system that integrates human insights with the computational energy of XGBoost and CNNs leveraging these strategies. Initially specializing in textual analysis, linguistic skills are employed to enhance the system’s knowledge of written content, drawing insights from social media posts and information memories. Inspired by the adaptable nature of XGBoost and CNN algorithms, these insights facilitate the creation of a capable fake news detection system. Furthermore, spotting the increasing significance of video content material in the propagation of incorrect information, the proposed scheme is adapted to assess video material thoroughly. We analyze video speech content to detect fixed and misleading content using CNN and XGBoost together. Through rigorous testing with different datasets and simulated scenarios, our system demonstrates an impressive accuracy of around 96 %. Structured to deal with various types of misinformation in text and video content, our approach provides a practical way of identifying misinformation, ensuring the credibility of the content in the digital age. Compared with other existing methods, it reveals high accuracy, highlighting the effectiveness of our collaboration method in truthful detection.

Skin cancer is one of the most common types of cancer in the world, and it poses major health risks due to its ability to spread quickly and metastasize. Early and accurate identification is crucial for treatment success and improved patient outcomes. This proposed work combines the MobileNetV2 and Vision Transformer (ViT) architectures to create a hybrid automated skin cancer classification technique. This technique aims to increase the accuracy and efficiency of dermatological diagnosis tools by combining MobileNetV2’s effective feature extraction capabilities with ViT’s self-attention mechanism. After testing on the HAM10000 dataset, this hybrid model outperformed individual models with a remarkable $96.3 \%$ classification accuracy. Not only did the integration of ViT and MobileNetV2 improve the classification accuracy but it also demonstrated how various deep-learning architectures work together to tackle complex image analysis tasks. Classifying skin cancers is critical to the medical industry because it allows for the early diagnosis of various dermatological disorders, allowing prompt intervention and treatment. These benefits can eventually improve patient outcomes and save lives in huge numbers. The findings of this research highlight the potential of deep learning to transform dermatological diagnostics and open the door to creating systems that will significantly impact clinical practice by detecting skin cancer with greater accuracy and efficiency.

The automated recognition and identification of license plates is an essential element of intelligent transportation systems that enable effective traffic management, security measures, and the development of efficient e-payment algorithms in toll booths and parking facilities. Even though license plate recognition (LPR) technology is fairly advanced, significant efforts are being made to enhance its accuracy and resilience The growing adoption of deep learning technologies across multiple industry domains, particularly computer vision, presents a novel idea implemented in this paper. To develop an approach to improve detection accuracy using YOLOv9 and transfer learning. Through combining the power of YOLOv9 with transfer learning, it is feasible to tackle the traditional LPR system's inherent flaws. Character recognition, picture pre-processing, license plate extraction, and character segmentation are the four main parts of the proposed system. Each step involves unique methodologies, primarily focusing on character recognition methodologies using YOLOv9 which has not been adequately tackled in existing research. By thorough testing and analysis, this study seeks to further the current developments in LPR systems, ensuring discernibly better-established intelligent transportation solutions. We have achieved a mAP50 Score of 99.5% and a mAP50-95 score of 71 %

Excellent automotive inspection is one of the key application areas for getting enhanced quality at a low cost, and it may be obtained using advanced computer vision technology. Automated car component assembly and quality control depend on autonomous automotive part recognition. ReXNet’s efficient and lightweight architecture enables the automation of auto component assembly and the improvement of quality control processes with an advanced solution. With the help of ReXNet, autonomous cars can effectively recognize and inspect a wide range of parts, including electrical modules, actuators, connectors, and sensors. The dataset comprises 16,600 RGB photos depicting 50 distinct types of auto parts. It encompasses a train set, test set, and validation set. Each of the 50 classes in both the validation and test sets has five images. In the identification of individual parts, the maximum testing accuracy of 98.6% and the maximum training accuracy of 98.4% were reached. Efficient car inspection systems, model verification, and vehicle manufacturing are just a few of the areas that stand to benefit significantly from such precise car part classification. Classifying car parts is helpful for several purposes, including model verification, automotive inspecting systems, and automobile manufacturing.

A revolution for ultra-security in image transmission, due to the innovation and upgrade of cryptographic algorithms occurred during recent years. Proposed image encryption algorithm with ChaCha20 Our matrix-based shuffling not only improves image security but also adds to the complexity of data stealth. The scheme is made up of four main operations: Rearranging pixels in the image with a transform matrix, encryption using ChaCha20, decryption to get back the original image and finally descrambling it by applying inverse transformation changes. In this way, integrity is maintained and image privacy becomes secure. The provided transformation matrix makes scrambling with reversible operations and speed possible against all of the known standard attacks to which ChaCha20 encryption is subjected. The suggested method, led to a Peak Signal Noise Ratio of 361.20 and zero Mean Squared Error indicating a top-quality reconstructed image concerning the original image. The study is raising the security alerts provided by ChaCha20, a relatively new but efficient and high-fidelity encryption solution that makes it an attractive option in today’s increasingly digitalized environment where data security has become everyone’s top priority.

The phenomenon of distraction is very common, and its adverse effects are seen among people. The major cause underlying this issue is the ease with which adversarial web sites and web pages can be accessed. It is of utmost importance to locate, evaluate, and actively block such web pages in an effort to comprehensively and globally solve this societal issue. Thus, the given paper proposes an extension or plug-in to detect, analyze, and block websites smartly. The proposed approach takes keywords entered by the user as input into consideration, which eventually leads to the generation of web pages and web links. The filtering of web links is done by the proposed extension, followed by the extraction of features, the utilization of support vector machines (SVM) for binary classification, and the summarization of textual data using natural language processing (NLP). Lastly, the precise results corresponding to relevant web links are presented to the users, which will increase their productivity, thereby reducing their distraction levels under all working circumstances. The performance of the proposed approach is validated against existing recent studies based on evaluation metrics such as adversarial website detection time (ms) and accuracy (

The growth of e-commerce has altered how consumers shop, providing a digital space where convenience, vast product offerings, and competitive pricing converge. In today’s world, e-commerce websites are transitioning from traditional search-driven methods to customized and intuitive approaches via product suggestions. Product recommendation systems are vital in e-commerce, from bringing new business to retaining existing ones. Our three-part recommendation system is designed so that new users have a great and engaging experience as the Product Popularity -Based System shows them carefully chosen products that are in demand. Collaborative Filtering Recommendations are highly personalized recommendations given to people who have already made their first purchases based on their prior actions and preferences. The K-Means Clustering-Based Recommendation System uses textual clustering analysis to deliver contextually relevant recommendations. We use a variety of evaluation metrics, such as click-through rates, user engagement, and the Silhouette Score, to assess the effectiveness and accuracy of our recommendation systems. Our findings show significant increases in user engagement, conversion rates, and relevant recommendations. Our findings demonstrate the transformative power of well-designed recommendation systems, which improve user experiences and retention and provide invaluable solutions for businesses entering the e-commerce space. This paper provides an in-depth examination of the multifaceted landscape of e-commerce recommendations, shedding light on their far-reaching implications for customer acquisition and retention in this dynamic digital era.

An authenticated mobile key exchange is necessary for post quantum secure communication over a public channel. It enables two parties to establish a secure fresh session key for each of the communication. In the last few years, many post quantum authentication key exchange protocols have been designed. This article contains a discussion how the Dabra et al. protocol is vulnerable to insider attack. There is an error in the password change phase, and it suffers from signal leakage attack. The protocol has incorrect login phase or they uses incorrect values during reconciliation. We have discussed the signal leakage attack in detail, and how the protocol vulnerable to this attack. The proposed protocol achieves both security and efficiency against quantum attacks.

YOLO has developed into a primary real-time object identification platform for applications such as video surveillance systems, autonomous vehicles, and robots. This research proposes an improved real-time object recognition approach based on the YOLOv9 architecture, trained on a football-specific custom dataset. The 255 annotated images in the dataset are separated into training, validation, and test sets with an 80-15-5 split ratio; there are 204 images in the training set, 38 in the validation set, and 13 in the test set. Our gelan-c model achieves remarkable detection performance with 467 layers, 25,414,044 parameters, and 102.5 GFLOPs. Upon validation, Gelan-c shows an aggregate mAP50 of 75.9%, with mAPs ranging from 32.2% (ball) to 93.7% (goalkeeper) depending on the specified class. The impressive aspect of the Gelan-C is its 31.6 ms picture processing rate, which assures real-time performance. The methodology covers environmental setup, preparation of data, YOLOv9 architecture introduction, and experimental evaluation to demonstrate the model's football-related recognition capabilities. The suggested approach has potential applications in sports event monitoring and evaluation due to its reliability and efficacy in detecting football players, referees, and the ball.

The use of multi-class Support Vector Machines (SVM) for snowfall prediction is an advanced meteorological application. The study emphasizes how valuable snowfall forecasts are in various contexts, including agriculture, aviation, transportation management, and climate reporting. This approach presents a novel methodology for classifying snowfall types by analyzing various meteorological factors such as temperature range, wind dynamics, and precipitation. A 400-entries of datasets were pulled from GitHub to demonstrate the methodology. The sample is selected to reduce the complexity and to increase the performance. The dataset is split into a testing set (30

Basketball is a very popular sport, and artificial intelligence (AI) has attracted a lot of attention because it can help with player training, help coaches come up with winning strategies, lessen injuries related to sports, and make the game more enjoyable overall. To precisely track and record player scores in real-time, this article presents an inventive AI-based automated basketball scoring system that makes use of computer vision and deep learning techniques. Using a variety of libraries, including TensorFlow, Fast RCNN, OpenCV, Yolov4, Yolov7, and the mmpose tool, the system uses pose recognition, basketball and hoop detection, and tracking techniques. This technology increases the effectiveness of player training by updating scoreboards and game data on its own. The suggested approach has an average accuracy of 78.74% for identifying objects, 83.75% for detecting shots, and 80% for identifying the shooter. The accuracy of the scoring system in allocating scores to shots is 66.67%. This technology is a significant development in the field of sports analytics, improving the experience of analyzing basketball games and maybe opening up new applications for real-time monitoring situations and other sports. Future research will concentrate on broadcast content analysis and real-time deployment.

In the digital age, the integrity and privacy of information transmitted, communicated over computer networks or stored on computer systems are crucial. This paper suggests a novel and robust symmetric encryption scheme based on Fermat's theorem to facilitate seamless and secure communication while preserving the confidentiality and integrity of data sent over a public channel. The knowledge of the public key does not reveal any valuable information to an intruder for computing the encryption/decryption keys used here. An essential feature of this scheme is that no additional information is passed again through the public channel or a private courier in the computation of the encryption/decryption keys. Also, the keys are dynamic for each data transmission, and the Plaintext and Ciphertext are not the same lengths. Consequently, the proposed encryption-decryption technique is safer against all possible standard attacks and more efficient than the existing encryption algorithms.

A three-party authenticated key exchange protocol enables two entities to agree on a session key with the help of a dedicated server through an insecure channel. Lattice based cryptography plays a very important role in authentication and key exchanges that protects against the threat of quantum attacks. However, it is not easy to design quantum resistant password based three-party protocol due to the high demand for security requirements and the limited resources nature of mobile devices. In this article, we have proposed a new post quantum three party key exchange based on a variant of lattice assumption, the ring learning errors. The protocol ensures security against impersonation attack, stolen smartcard attack, password guessing attack, and other existing attacks. In authentication phase, the protocol have used lattice based cryptography that plays a very important role in authentication and key exchanges that protects against the threat of quantum attacks. The proposed protocol ensures both securities against quantum attacks and efficiency due to simple algebraic operations that are polynomial addition and multiplications.

Abstract This study has clarified the bipolar alternate mark inversion encoding scheme for performance signature of optical transceiver systems under thermal effects. The light power through optical fiber channel variations with spectral wavelength at various temperature variations are clarified. The total electrical power through PIN receiver at various temperature variations are demonstrated. The electrical power amplitude, max Q Factor and min BER through PIN receiver at different temperature variations are clarified in details. Electrical and optical signal per noise ratio with laser bias current at various ambient temperatures are demonstrated. The total optical/electrical power and maximum quality signal with laser bias current at various temperature variations.

The Internet of Vehicles (IoV), with advanced technology in 5G communication, is considered the backbone of a smart city’s intelligent transport system. There are mainly two types of vehicular communication: (1) Vehicle to Vehicle (V2V) and (2) Vehicle to Infrastructure (V2I). In IoV, there is a dynamic change in the network topology according to the controllers, destination, vehicle movement, and road structure. Intelligent vehicles are assumed to work with the capacity of data processing, data storage devices, and communication devices to communicate with vehicles or Roadside InfraStructure (RSI). This article presents a post-quantum secure ring learning with error-based key exchange. This assumption ensures security against quantum attacks. The proposed design is an ant colony optimization-based ad hoc ordered distance vector routing algorithm that avoids suspicious vehicles in IoV broadcasting. The proposed framework consists of three parts: (i) certificate authority, (ii) suspicious vehicle detecting algorithm, and (iii) optimal path selection algorithm. The performance analysis Section contains an analysis of the proposed design with related ones, and the proposed has better results.

The most important component that can express a person's mental condition is facial expressions. A human can communicate around 55% of information non-verbally and the remaining 45% audibly. Automatic facial expression recognition (FER) has now become a challenging task in the surveying of computers. Applications of FER include understanding the behavior of humans and monitoring moods and psychological states. It even penetrates other domains-namely, robotics, criminology, smart healthcare systems, entertainment, security systems, holographic images, stress detection, and education. This study introduces a novel Robust Facial Expression Recognition using an Evolutionary Algorithm with Deep Learning (RFER-EADL) model. RFER-EADL aims to determine various kinds of emotions using computer vision and DL models. Primarily, RFER-EADL performs histogram equalization to normalize the intensity and contrast levels of the images of identical persons and expressions. Next, the deep convolutional neural network-based densely connected network (DenseNet-169) model is exploited with the chimp optimization algorithm (COA) as a hyperparameter-tuning approach. Finally, teaching and learning-based optimization (TLBO) with a long short-term memory (LSTM) model is employed for expression recognition and classification. The designs of COA and TLBO algorithms aided in the optimal parameter selection of the DenseNet and LSTM models, respectively. A brief simulation analysis of the benchmark dataset portrays the greater performance of the RFER-EADL model compared to other approaches.

Early detection identified by dermoscopy images significantly decreases the mortality rate from skin cancer. However, the accuracy of the system diagnosis is impacted by multiple factors. A significant issue in this procedure arises during the process of acquiring images. The image quality in medical photography is frequently affected by unanticipated circumstances such as noises and brightness variations, initial digitalization, and sampling. In this work, we suggest a method to reduce the possibility of erroneous diagnosis. The proposed approach begins with preprocessing the data set through CNN and deep learning techniques for data visualization, pre-processing, and augmentation. This preprocessing aids in accelerating the rate of reorganization by eliminating all irrelevant textures. The outcomes identified a 96% accuracy rate with a 4% margin of error. This scheme demonstrated a high degree of accuracy in determining the images, as evidenced by its Precision and F1 Scores of 85% and 87%, respectively.

The Internet of Vehicle enables million of vehicles to transfer sensitive information through an Internet of Things assisted networking system. It is a very challenging idea wherein different types of data authentication protocols have been designed number theory based assumptions (RSA/DLP). The quantum computers are big threats to number theoretic assumptions. In this article, we have designed a post quantum secure aggregate signature assisted data authentication framework for internet of vehicles. Further, a secure aggregate signature assisted data authentication framework have been proposed that supports batch data verification. We have proposed an efficient aggregate signature to verify authenticity of multiple vehicles in one step. The security of protocol relies upon two assumptions, (1) Module Learning With Errors and (2) Module Short Integer Solution. The proposed design ensures security against quantum attacks. The design is simulated in ns3 to capture latency and throughput, respectively.

Recently, Internet of the things (IoTs) devices have been used highly in surveillance applications in smart environments, cities, and buildings. Most of the surveillance data recorded by the IoT devices is in the form of video and needs to be transmitted to the admin or server by video streaming technology. Fractal video compression uses a self-similarity concept for video compression, meaning the fractal image contains self-similarity of itself that is explained and denoted by change. Computational cost is more in fractal video coding; different methods have been developed to reduce this computational cost. The proposed work's main objective is to implement a video streaming approach by combining the Diamond-search-pattern- block-matching motion estimation algorithm and hash-based fractal video compression algorithm to reduce the encoding time. It improves both the delay and security of the block matching system and the developed approach measures based on evaluation parameters: Peak signal-to-noise ratio (PSNR), encoding time, decoding time, Mean squared error (MSE), and Compression ratio for Smart City Applications. Compared to other algorithms, the IoTs video data is streamed efficiently in a smart city to analyse the abnormalities.