Computer vision plays an important role in smart home applications, particularly in reducing food waste through automated identification of fruits and vegetables in smart refrigerators. Convolutional Neural Network (CNN) models have shown promising results for such tasks; however, their performance depends on multiple factors such as dataset size, number of classes, training time, and classification accuracy. This paper proposes a systematic evaluation of CNN performance using the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). The novelty of this work lies in the use of a multi-criteria decision-making framework to jointly optimize CNN parameter combinations instead of relying solely on accuracy, as commonly reported in existing studies. InceptionV3 and MobileNetV3 models are trained and evaluated using different configurations of the FRUITS360 and FIDS30 datasets. Based on TOPSIS ranking, the complete FRUITS360 dataset with 90,483 images and 131 classes is identified as the optimal configuration for both models. Experimental results show that InceptionV3 achieves an accuracy of 94.9 % with a training time of 415 minutes, while MobileNetV3 achieves 99.9 % accuracy with a training time of $\mathbf{3 6 1}$ minutes. The proposed approach supports informed CNN model selection for real-time fruit and vegetable identification applications.

Vehicular Ad Hoc Networks (VANETs) are vital additives of ITS (Intelligent Transport Systems), as they drastically improve avenue safety by supplying up-to-date visitor records, such as traffic congestion and accidents the records supplied. Unscrupulous those who breach the community can manipulate visitors data, developing a risky state of affairs. Consequently, securing VANETs by enforcing strong algorithms to hit upon and prevent undesirable malicious visitors is essential. A privacy-keeping authentication approach is proposed to overcome this hassle. The approach prioritises verifying the motors earlier than connecting to the car community, detecting and disposing of malicious content material accordingly. The instance is administered on a Docker field that internally clones the network on a Linux gadget using Ubuntu 20.04. This framework affords a regulated and steady test environment to evaluate the performance of the proposed solution. Pseudo-IDs are assigned to compromised vehicles to ensure confidentiality. This technique ensures the identification of the cars even during a communications failure. The implementation results of the model display high performance, speedy verification time, and occasional computational cost compared to different methods. These homes are essential inside the vehicular community surroundings, wherein speedy and green conversation is critical to retaining traffic flowing and ensuring protection and privateness-shielding authenticated users. Its implementation’s usefulness and robustness are validated with its usage in a Docker box that mimics Linux surroundings. This technique shows a top-notch development in smart transportation systems and offers a viable technique for vehicular network protection.

Advancement in IOT technology over years has led to emergence of many new concepts, smart agriculture is one of them. Smart agriculture systems include technologies such as IOT and wireless networks to reduce human intervention by monitoring environmental conditions and then take appropriate action based on user input. In India agriculture sector contributes more than 20

Road safety and traffic efficiency are expected to be significantly enhanced by 5G-enabled Vehicle-to-Everything (V2X) communications. Security and low-latency authentication within such systems remain challenges, however. For 5G-enabled Internet of Vehicles (IoV), we propose a Cluster-Oriented Authentication and Key Update Protocol. This model reduces computation time and enhances vehicle anonymity by leveraging dynamic vehicle clustering and group authentication. As a result of the system, spectrum efficiency is improved and congestion is reduced through the utilisation of infrastructure-based networks. To ensure secure communication, it also employs a hybrid model that combines trust management mechanisms with blockchain technology. It utilises a lightweight, scalable authentication framework to achieve low latency, a crucial requirement for applications such as autonomous driving.

The process of breaking up a digital image into many parts is called segmentation. These sections in scanned papers, are those that have backgrounds, texts, and images. In applications linked to document analysis, text segmentation is a significant issue. The separation of a complex document into non-text and text components is a major challenge in document image analysis. Localization of text in printed document images is a critical processing step for page layout analysis (PLA), specifically for obtaining textual data. Several algorithms have been developed for this subject. However, many algorithms only produce accurate answers for a limited type of documents due to their reliance on specific attributes or assumptions. There is a need to develop effective techniques for many sorts of documents, such as newspapers, magazines, documents, and articles, with arbitrary layouts and nonhomogeneous backgrounds. This study examines the various ways that distinguish between non-text and text in document images.

Outlier detection is essential for identifying unusual patterns or observations that significantly deviate from the normal behavior of a dataset. With the rapid growth of data science, the prevalence of anomalies and outliers has increased, which can disrupt system modeling and parameter estimation, leading to inaccurate results. Recently, deep learning-based outlier detection methods have gained significant attention, but their performance is often limited by challenges in parameter selection and the nearest neighbor search. To overcome these limitations, we propose a three-stage Efficient Outlier Detection Approach (named EODA), that not only detects outliers with high accuracy but also emphasizes dataset characteristics. In the first stage, we apply a feature selection algorithm based on the Boruta method and Random Forest to reduce the data size by selecting the most relevant attributes and calculating the highest Z-score of shadow features. In the second stage, we improve the K-nearest neighbors algorithm to enhance the accuracy of nearest neighbor identification in the clustering phase. Finally, the third stage efficiently identifies the most significant outliers within clustered datasets. We evaluate the proposed EODA algorithm across eight UCI machine-learning repository datasets. The results demonstrate the effectiveness of our EODA approach, achieving a Precision of 63.07%, Recall of 82.49%, and an F1-Score of 64.53%, outperforming the existing techniques in the field.

There are too many flowers in the world. It is hard to recognize one specific flower type from millions of flowers. Offering automatic flower species reticle recognition method have many advantages for the farmers, other interested parties. flowers from one another by virtue of their similar forms and colours. Flower classification is one of the challenging problem with high shape variance, color diversity, different illumination situations and exposure deformation. For some images it is hard to classify flowers that look very alike, color- and shape-wise, with the human eye. It requires extraordinary training for people to correctly identify species/types and generally, microscopic features that are really network specific and two closely related flower types/species. Such CNN models have been employed by researchers to dismiss the closely related species in various classification problems recently. Recently CNN architectures have been adopted by researchers in a wide range of classification tasks, so as to avoid hand-crafted features. In this work, we experiment traditional based (CNNs) and ensemble model (VGG16, ResNet and InceptionV3) along with Transfer Learning models for comparison. The findings point out the advantages and limitations of each strategy with suggestions for choosing the best one based on certain standards such as considering the size of dataset, processing power on given dataset, and how much accuracy required for that approach.

Two-wheeled motor cycles riders are the more sensitive users of the road, the mishaps have the tendency to lead to grievous injuries or even death. The loss of lives due to delayed emergency services, especially remote or during certain times that the number of individuals using the means are low, is a critical factor with regards to high mortality. The accident detection using sensors has become a possible solution to this problem. These systems are using the accelerometers, gyroscopes as well as GPS modules as sensors that capture the sudden impact, unusual motion and/(or) imbalance that can be caused by accidents. Upon detection the system automatically passes emergency alerts and exact location information to predesignated contacts or emergency personnel. This review demonstrates the use of microcontrollers, inertial measurement units (IMUs) and real-time data analysis that increases the accuracy of detection. It also talks about ongoing research works, prototypes, and commercial usages which show that these systems can be capable of shortening the response time and saving lives. Limits such as false alarms, sensor tuning, and network connectivity are overviewed, and future developments with regard to machine learning, cloud, and better user feedback described.

Non-Fungible Tokens (NFTs) have arisen as a major term in the twenty-first century, signifying unique digital assets such as photographs, music, and movies. These tokens receive their value and legitimacy from blockchain technology, which is mostly used on the Ethereum platform. NFTs, which use smart contracts, enable the production, ownership, and trading of one-of-a-kind digital assets, revolutionizing sectors including as art, music, and gaming. While blockchain’s uses are broad, there are worries about the environmental effect of blockchain, notably Ethereum’s energy-intensive proof-of-work consensus. Despite their revolutionary potential, NFTs pose dangers connected to market instability, legal concerns, and technological weaknesses. NFT trading is enabled by a variety of third-party programmers and markets, with notable participants being OpenSea and MetaMask. Looking ahead, the future of NFTs in India is dependent on legislative changes, as the country sees an increase in interest in these digital assets among artists and producers, necessitating a balanced approach to encourage innovation while resolving possible difficulties. NFTs are a game changer in the digital economy, opening up new opportunities for producers and investors. However, environmental problems and regulatory uncertainties must be resolved in order for the NFT ecosystem to expand sustainably. Technological improvements, legislative reforms, and the continuous growth of blockchain technology will most likely influence the future of NFTs in India and throughout the world.

Due to advancements in the deep learning technology, object detection has become significantly important for lane detection and vehicle detection. In recent times, lane detection has become more popular as it plays a significant role in traffic surveillance compared to other object detection technology. However, these strategies have several intrinsic flaws which need to be addressed. Traditional-based techniques still suffer from the challenges of the effectiveness and accuracy, whereas a complex convolutional layer is a challenge for deep learning-based strategies. A parameter selection issue affects the majority of the available lane detection algorithms, which further contributes to their unsatisfactory detection performance. In this study, we provide an effective lane detection method based on semantic segmentation to identify lane lines in a high-dimensional dataset by adding vertical spatial properties and contextual driving information. This paper employs two created frames—feature merging block and information exchange block—to identify unclear and obstructed lane lines more effectively. The simulations have been carried out for the proposed model on TUSimple and CULane datasets which resulted with 94.42

Blockchain technology is transforming business models and processes across multidisciplinary industries through increased transparency and decentralization. As blockchain adoption accelerates, it attracts cybersecurity threats so it is important to necessitate robust security strategies. The chapter examines the important role of cybersecurity in enabling organisations and businesses to harness blockchain technology strategic value. How cybersecurity comprehensively builds confidence and trust in blockchain solutions is discussed and allows companies to maximize benefits like cost savings, enhanced security, and also new revenue opportunities. Blockchain components include smart contract auditing, education, and network security. It also emphasizes how to proactively address vulnerabilities and protect blockchain systems which gives organisations a competitive advantage in this digital economy. Lastly, it provides some practical insights in implementing a blockchain cybersecurity strategy that unlocks the strategic potential of this transformative innovative technology.

Cryptography is used to protect data from adversaries who are unauthorized to access it, Cryptograph is used to changes the form of the data using a specific algorithm which can be reversed using a particular key. We already have many Cryptographic algorithms but there is always a need for new algorithms because old techniques can be cracked with new technology and the high processing power of computers, hence we need new techniques that are much more robust and harder to crack. In this paper, we propose a new encryption algorithm that uses the Basic principle of the Caesar cipher and random generation technique to encrypt the input text. This algorithm is much more secure than the normal Caesar cipher as it generates a random number for each letter and converts it to encrypted text, the use of random numbers makes it very hard to crack. It is also important to understand the concept of key management, which involves generating and storing cryptographic keys securely. Our algorithm we are using 2 lists of random numbers which will together create a large list of random numbers which will be used to encrypt the text. The 2 lists used previously will be our 2 keys which can be used to decrypt the encrypted text.

X (formerly Twitter) has become a vital source of information on various variety of social, political, and economic concerns, as a consequence of its growth and popularity which has resulted in an enormous number of people sharing their opinions on a wide range of areas.To determine people's emotions about the Russia-Ukraine war (RUW), this study examines trends in English-language tweets.In this work, we have engaged 34 countries to tweet opinions that produce a strong perception of the people about the war and message to the world what people famine from the countries and that affects their lives.To analyze positive and negative emotions in tweets, which are represented by hope and fear, the LSTM-CNN model is based on deep learning.A time series is calculated that correlates with the rate of recurrence of negative and positive tweets in different nations.Additionally, an approach based on the average of the neighborhood has been used for modelling and grouping the time series of various countries.The clustering method gives results as significant information, how people feel about this dispute and share their opinions about RUW is approached.When compare on different models on overall data that the 96% accuracy is Achieved by the LSTM-CNN model.97.09% accuracy, is achieved when the comparing the tweets from the cluster 1 countries.When comparing the tweets from the cluster 2 countries the 99% accuracy, is achieved.97% accuracy is achieved by comparing the tweets from cluster 3 countries.97% accuracy, is achieved when the comparing the tweets from cluster 4 countries.96% accuracy is achieved by the LSTM-CNN model when the comparing the tweets from cluster 5 countries by the different models.This research study helps the uninfluenced press members to have an impartial source of information for their reports and articles.

Defending blockchain technology from cyber dangers is essential in today's age of advances in technology and also increase in usage of Blockchain Technology. Due to the ever-changing nature of these dangers, people as well as businesses now seriously worry about protecting sensitive data and digital assets. Blockchain structure and cybersecurity are two essential elements of modern digital security that work together to increase the safety of Blockchain Technology. By utilising cybersecurity techniques, Blockchain Technology can be secured. However, cybersecurity refers to a variety of techniques and procedures used to protect systems, networks, and data from risks and attacks that arise online. Therefore, the primary inquiry that is the aim is to answer is: How can cybersecurity techniques be used to improve Blockchain Technology? This chapter will provide significant insights into this vital collaboration between technology and security as we thoroughly examine how blockchain technology may be improved and secured against the cyberattacks.

Vehicle AdHoc networks have an important role in intelligent transport systems that enhance safety in road usage by transmitting real traffic updates in terms of congestion and road accidents. The dynamic nature of the vehicular AdHoc networks make them susceptible to attacks because once malicious users gain access to the network they can transform traffic data. It is essential to protect the vehicular ad hoc network because any attack can cause unwanted harm, to solve this it is important to have an approach that detects malicious vehicles and not give them access to the network. The proposed approach is a privacy preserving authentication approach that authenticates vehicles before they have access to the vehicular network thereby identifying malicious vehicles. The model was executed in docker container that simulates the network in a Linux environment running Ubuntu 20.04. The model enhances privacy by assigning Pseudo IDs to authenticated vehicles and the results demonstrate effectiveness of the solution in that unlike other models it boasts faster authentication and lower computational overhead which is necessary in a vehicular network scenario.