Multi-cloud strategy allows distribution of workloads across multiple cloud platforms. These cloud platforms that are served include, public, private, as well as hybrid clouds. This gives organizations more flexibility than they’d have, if they depended on a single cloud vendor and in turn, it helps them optimise costs, dodge vendor lock in and reinforce operational resilience. As a result, while multi-cloud deployments have their inherent complexity, this increases risk of attack surface and it presents new and evolving cloud security challenges. A holistic security approach is needed for multi-cloud security that addresses diverse security vulnerabilities in multi-cloud enterprise world and enforces security across heterogeneous platforms in a consistent way. In this work we present a Hybrid Metaheuristic Federated Learning Approach Based Attack Detection System (HMFLA-ADS) for multi-Cloud scenarios. The use of Federated Learning (FL) in the HMFLA-ADS model makes it possible to collaborative train across various cloud platforms while retaining data privacy and safe. The HMFLA-ADS technique achieves better attack detection by aggregating knowledge from locally trained models at a Cloud Exchange Point (a federated node deployed in each cloud region) without requiring data sharing in the centralized manner. A deep Stacked Sparse Autoencoder (SSAE) for attack detection and a hybride Fish School Search (FSS) based strategy for hyperparameter tuning are used in the study. Experimental validation of the HMFLA ADS framework is carried out on NSL-KDD, CICIDS-2017 and BoT-IoT Datasets and results are measured through various evaluation metrics. The findings show that HMFLA-ADS framework has better performance than existing IDS models, obtaining maximum detection accuracy up to 99.52

Background Organizations achieve agility, scalability, and enhanced resource utilization in multi-cloud environments, but face challenges in ensuring uniform and robust security across diverse cloud platforms. Variations in configuration and security mechanisms among providers hinder consistent policy enforcement and expose systems to data breaches and evasive threats. Additionally, the dynamic and distributed nature of multi-cloud operations broadens the attack surface, making real-time threat mitigation more complex. Methods To address these challenges, we introduced a groundbreaking software-defined networking (SDN)-enabled framework that incorporates deep learning for attack detection and adaptive security policy management. The framework consists of two primary components: the Software-Defined Multicloud Defense Controller (SDMDC), which enables centralized, real-time security policy enforcement (control plane), and the Multicloud Intrusion Detection System Gateways (MCIDS-G), which facilitate distributed threat detection across cloud platforms (data plane). The SDMDC’s integrated IDS is built using the Cross-Cloud Threat Transformer (CCTT) model, while the MCIDS-G’s regional IDS is based on the Long Short-Term Memory (LSTM) model. Additionally, the Lemurs Optimizer (LO) is employed in the SDMDC for cost-efficient policy management. SDMDC enforces security standards across all cloud environments where applications operate. Results The proposed solution addresses long-standing cloud security issues by combining coordinated global security strategies with automated threat detection and centralized policy management. SDMDC ensures consistent policy enforcement across cloud environments and manages ingress/egress and east–west traffic between cloud domains, including Amazon Web Service (AWS) and other service providers. The architecture supports dynamic resource orchestration and horizontal scalability, enabling adaptive performance under varying load conditions. The system also enables automatic policy implementation across platforms and facilitates real-time threat response to maintain consistent security. Conclusion The presented framework represents a significant advancement over existing multi-cloud protection solutions. It introduces new research directions, particularly in traffic management, firewall integration, and fully qualified domain name (FQDN) policy enforcement with proxy management.

As the prevalence of Distributed Denial of Service (DDoS) attacks continues to escalate, safeguarding cloud environments against these threats becomes paramount. This paper introduces a Deep Neural Network (DNN) model designed to enhance the accuracy and efficiency of DDoS attack detection in cloud environments. Leveraging the inherent capabilities of deep learning, the proposed model exhibits improved performance on the widely recognized NSL-KDD dataset. The research findings demonstrate a substantial increase in accuracy, underscoring the efficacy of the DNN model in fortifying the security posture of cloud infrastructures. The escalating frequency and sophistication of Distributed Denial of Service (DDoS) attacks pose a substantial threat to the security of cloud environments. In response to this pressing concern, this paper introduces a Deep Neural Network (DNN) model engineered to significantly enhance the accuracy and efficiency of DDoS attack detection in cloud infrastructures. By harnessing the inherent capabilities of deep learning, the proposed model represents a breakthrough in fortifying the security posture of cloud systems. This research employs the widely recognized NSL-KDD dataset, a comprehensive resource for Intrusion Detection System (IDS) evaluation, to evaluate the model's performance. The proposed DNN model transcends conventional methods by autonomously learning intricate patterns within network traffic data, adapting to the evolving landscape of DDoS attacks. The literature survey delves into the vulnerabilities associated with DDoS attacks in cloud environments, emphasizing the need for innovative detection mechanisms. Previous research has underscored the effectiveness of deep learning, particularly DNNs, in addressing complex cybersecurity challenges, positioning them as ideal candidates for enhancing threat detection capabilities.

With the increasing reliance on cloud computing services, ensuring the security of cloud-based infrastructures has become paramount. This paper proposes a novel approach utilizing Convolutional Neural Networks (CNNs) to detect Distributed Denial of Service (DDoS) attacks in a multi-cloud environment. The utilization of CNNs, known for their proficiency in feature extraction from structured and unstructured data, offers an innovative solution to the dynamic and evolving nature of DDoS attacks. The datasets utilized in this study include NSL-KDD. As the global frequency of cyberattacks rises, the digital landscape faces significant threats impacting both individual online presence and corporate entities. This paper employs deep learning techniques to enhance security against DDoS attacks, utilizing the inherent ability of deep learning to extract intricate patterns from vast datasets. This makes it a potent tool for constructing effective detection and mitigation systems for the DDoS threat. The research presents a comprehensive approach for detecting DDoS attacks by leveraging CNNs (Convolutional Neural Networks) and advanced data preprocessing methods, focusing on the widely recognized NSL-KDD dataset. The research findings reveal that the proposed CNN-based approach consistently outperforms, achieving an impressive accuracy score of 97.46%. These results underscore the promising potential of the proposed methodology in significantly improving the accuracy and effectiveness of intrusion detection systems. In a parallel exploration, another research paper introduces an alternative approach by offering a CNN model on the same dataset, achieving an accuracy outcome of 96.61%. While this alternative approach demonstrates commendable performance, the findings highlight the superior accuracy achieved by the proposed CNN methodology in the context of DDoS attack detection.

Mostly, Weeds are the responsible for agricultural losses in recent years. Removing weeds is a challenging task as there are much similarity between weed and crop in terms of texture, color and shape. To deal with this challenge, a farmer needs to spray herbicides uniformly throughout the field. In addition to requiring a lot of pesticides, this method has an adverse effect on the environment and people’s health. To overcome this, precision agriculture is used. Unmanned aerial vehicles (UAVs) have been shown great prospective for weed detection, as they can cover large areas of farmland quickly and efficiently. For this experiment, phantom p4 drone was used to take the images of rice field. Therefore, in this work, we propound a weed recognition system using UAVs and a combination of RCNN model and modified RCNN-LSTM and RCNN-GRU. The performance was compared using accuracy, precision, recall, and f1-score as evaluation criteria. Among all RCNN with GRU outperformed with 97.88%.

Garbage management is exceptionally critical and poses enormous environmental challenges. It has always been a vital issue in municipal corporations. However, municipal agencies have developed and used garbage management systems. Garbage forecasting still plays a crucial role in the management system and helps improve or create a garbage management system. This research examines the information from 212 cities to suggest a helpful regression model for garbage forecasting and control. To establish a connection between the variables, the descriptive study employs statistical techniques to learn about the composition of data collected from municipal corporations and conduct correlation analysis. Population and garbage depend highly on one another, as evidenced by their correlation coefficient of 0.922,144. The primary research is used to build an alternate hypothesis that shows the chosen variables are highly dependent on one another. The dataset is scaled and divided into a training and testing 80:20 ratio during the pre-processing data phase. This research aims to do a regression analysis with daily garbage production, urban area, and population as independent variables. This research initiates a variety of regression models, including multiple linear regression (MLR), artificial neural network (ANN), decision tree regression (DTR), and random forest regression (RFR). The MLR model's R2 value of 0.85 indicates that it has the potential to accurately forecast daily garbage production based on just two independent variables and a single dependent variable. Random Forest Regression (RFR) with (MSE: 100,078.749 & MAE: 182.212) shows that it has the lowest MSE among all the models, which provides the most accurate predictions on average and the fit values of 8.85 and 316.23 obtained from the error distribution with a bin value 25. The estimated results from each model are compared to the test data values on line graphs and Taylor plots. The mean square error and the mean absolute error in the analysis and the Taylor plot show that the RFR model is best suited for predicting daily garbage production in a city. This research, therefore, provides a Random Forest model that is optimal for such challenges and is recommended for this class of problem.

Vehicle detection is still challenging for intelligent transportation systems (ITS) to achieve satisfactory performance. The existing methods based on one stage and two-stage have intrinsic weakness in obtaining high vehicle detection performance. Due to advancements in detection technology, deep learning-based methods for vehicle detection have become more pop-ular because of their higher detection accuracy and speed than the existing algorithms. This paper presents a robust vehicle detection technique based on Improved You Look Only Once (RVD-YOLOv5) to enhance vehicle detection accuracy. The proposed method works in three phases; in the first phase, the K-means algorithm performs data clustering on datasets to generate the classes of the objects. Subsequently, in the second phase, the YOLOv5 is applied to create the bounding box, and the Non-Maximum Suppression (NMS) technique is used to eliminate the overlapping of the bounding boxes of the vehicle. Then, the loss function CIoU is employed to obtain the accurate regression bounding box of the vehicle in the third phase. The simulation results show that the proposed method achieves better results when compared with other state-of-art techniques, namely Lightweight Dilated Convolutional Neural Network (LD-CNN), Single Shot Detector (SSD), YOLOv3 and YOLOv4 on the performance metric like precision, recall, mAP and F1-Score. The simulation and analysis are carried out on PASCAL VOC 2007, 2012 and MS COCO 2017 datasets to obtain better performance for vehicle detection. Finally, the RVD-YOLOv5 obtains the results with an mAP of 98.6% and Precision, Recall, and F1-Score are 98%, 96.2% and 97.09%, respectively.

In recent years, there has been an increased emphasis placed on the identification of face traits in studies. The human face is the most significant characteristic that may be used in the process of identifying a person. Even the most genetically identical twins may be distinguished from one another by a few key facial characteristics. Therefore, to discern one from the other, a human face identification and detection system that is based on facial traits is necessary. This study suggests a technique for automated human face identification and recognition based on facial characteristics that are achieved via the use of deep learning. It would seem that deep learning, with its high rate of accuracy, would be an appropriate method to use while carrying out face recognition. Face detection and identification may be accomplished using a process known as deep learning. According to the results of the study, it is possible to conclude that the approach that was suggested is superior to other ways in terms of accuracy, precision, recall, and f1-score.

Today, detecting waste, collecting it, processing it, and getting rid of it are among the most significant environmental issues in developing and undeveloped counties. It has been observed that a large amount of garbage remains strewn on the roadside. This study presented a garbage detection technology such as machine learning and gadgets connected to the Internet of Things (IoT), such as an IP-enabled CCTV camera, to take pictures and send them to the city's main server. The input images are transformed into a two-dimension array of integers using Python modules and divided into the garbage and no garbage classes. There is an 80:20 split between the training and testing datasets from the input dataset. Preprocessed images are then utilised as inputs for a wide range of machine learning and neural network models for classification; these include K-Nearest Neighbour (KNN), Logistic Regression (LR), Naive Bayes (NB), and Support Vector Machine (SVM). The test data sets are applied, and a confusion matrix is formed for all models to analyse the efficiency and performance of the trained models. Results from the confusion matrix are contrasted with those from the area under the Receiver characteristics operating curve (AUC). As a result, the ConvNet model is best suited for classifying garbage or no garbage present in open space, and the LR model proposed best suits the garbage detection problem. The proposed models are best suitable for improving the efficiency of existing garbage identification systems and developing a new system for smart cities.

Blockchain, as the name suggest is a long chain of blocks which are connected using the hashed of the data which is stored in the previous blocks. In present scenario schools, colleges and universities are being forced towards the creation, maintenance, retrieval, safety and privacy of educational records. Challenge is not only to store and maintain the data but also ensures fast access of data to all the stakeholders. Blockchain based transcripts save the time and money and smart contracts help in the authentication via codes. But there are several limitations of blockchain database, When sharing blockchain data from a server to multiple clients, it is highly likely that multiple clients will need the same data. They are working on the same case and may be investigating the same wallet. Blockchain and IPFS based transcripts save the time and money and smart contracts help in the authentication. Transcripts are safeguarded by blockchain in the event of a server outage or natural disaster. This work analyses the performance of IPFS database that uses content based addressing and cryptographic hashes with other database. In this paper IPFS is compared with blockchain database. IPFS is hypermedia protocol that is both safe and easy to use. It uses the content based addressing with the merger of cryptographic hashes to work upon the distributed file system. This paper compares IPFS with Blockchain database upon the following factors like Size of files shared by nodes, Degree of linking through number of peers linked, Bandwidth used.

Weeds are the major source of concern for farmers, who anticipate that weeds may lower crop productivity. Thus, it is essential and vital to detect weeds. Traditional weed classification methods such as hand cultivation with hoes have many hindrances such as labour cost and time consumption. Currently, weed reduction farmers are using herbicides, but they have a negative impact on farmer health as well as on the environment. So, farmers want to lower the use of herbicides. Precise spraying is one of the methods in present-day agriculture to lower the usage of herbicides and to destroy the weeds with the assistance of new technologies. Deep learning approaches are already being employed in a variety of agricultural and farming applications and gave better results. This chapter uses convolution neural networks to provide a short overview of some significant agricultural research endeavours. Different architectures of CNN for classification and detection were used. In the sector of agriculture, the authors have outlined the notion of CNNs.

Effective face recognition is accomplished using the extraction of features and classification. Though there are multiple techniques for face image recognition, full face recognition in real-time is quite difficult. One of the emerging and promising methods to address this challenge in face recognition is deep learning networks. The inevitable network tool associated with the face recognition method with deep learning systems is convolutional neural networks (CNNs). This research intends to develop a new method for face recognition using adaptive intelligent methods. The main phases of the proposed method are (a) data collection, (b) image pre-processing, (c) normalization, (d) pattern extraction, and (e) recognition. Initially, the images for face recognition are gathered from CPFW, Yale datasets, and the MIT-CBCL dataset. The image pre-processing is performed by the Gaussian filtering method. Further, the normalization of the image will be done, which is a process that alters the range of pixel intensities and can handle the poor contrast due to glare. Then a new descriptor called adaptive local tri Weber pattern (ALTrWP) acts as a pattern extractor. In the recognition phase, the VGG16 architecture with new chick updated-chicken swarm optimization (NSU-CSO) is used. As the modification, VGG16 architecture will be enhanced by this optimization technique. The performance of the developed method is analyzed on two standards face database. Experimental results are compared with different machine learning approaches concerned with noteworthy measures, which demonstrate the efficiency of the considered classifier.

Lane detection is the most common application for detecting lane boundaries in autonomous vehicles intelligent driving systems. Lane detection performance has a significant impact on autonomous vehicle driving systems. Lane detection helps with both vehicle positioning and lane departure. However, lane detection is still a problem that needs to be solved entirely for self-driving cars. These strategies must be not only effective but also be efficient. Due to improvements in computer vision and deep learning-based technologies, lane detection currently achieves better results in the accuracy of lane detection. However, accurate lane detection is still the most challenging task in difficult conditions such as poor light, unclear lanes, and occlusions. The results of recent research on lane detection systems are presented in this review paper. First, we discussed the history of traditional and deep learning-based lane detection methods. Then we have discussed the importance of loss function in lane detection. Second, we have compared the experimental results of each technique for deep learning and state-of-the-art methods. Third, the summarized list of existing datasets for lane detection, performance evaluation criteria, and lane detection based on deep learning methods are discussed. Finally, we looked into some of the current issues of deep learning algorithms.

A major challenge in computer vision is detecting and tracking vehicles in real-time.However, existing algorithms fail to detect vehicles at high speed and accuracy.Therefore, an algorithm that detects vehicles with higher accuracy is required for surveillance in traffic scenarios.This paper proposed an improved algorithm for vehicle detection based on YOLO (You Only Look Once) Version 4 through Convolution Neural Network (CNN) and Hard Negative Example Mining (HNEM) dataset in the training process to improve the accuracy of the vehicle detection.In the end, videos are used to detect vehicles using a deep learning technique called You Only Look Once (YOLO).The test results indicate good real-time performance and high detection accuracy of the proposed algorithm.Several parameters such as accuracy, precision, recognition recall, F1, and mAP have been used to measure the proposed algorithm's performance.The experiments have proved that the proposed algorithm achieved satisfactory performance in real-time due to occlusion and change in viewpoint.Finally, our proposed algorithm achieves improved precision, recall and mAP compared to the existing algorithms for occluded vehicle detection.

Satoshi Nakamoto created the first decentralizedblock chain in 2008. Block chain is a relatively new and widely accepted kind of decentralized, distributed computing. The advantages of block chain technology are improved security and privacy, decreased expenses, enhanced speed, and consistency. Block chain has applications in a variety of areas, including IT, healthcare, finance, and educational institutions. For data storage, many organizations now adopt centralised cloud systems. Educational Institutes stored and shared their digitally stored data with other institutes or Vendors for various purposes like issuing certificate, so protecting the privacy while sharing the data is a problem of researchers. People create a large number of files every day and wish to save them somewhere. However, they usually rely on unsecure centralized systems. Block chain is a decentralized system that keeps files and data safe. The suggested model will eliminate all cloud storage limits. It uses a block chain and an interplanetary file system to operate. It divides educational data files into blocks and then encrypts each block with hash keysOur goal is to come up with a privacy-preserving method for sharing data via block chain. A model along with Blockchain principles and features to serve the requester and data owner authentication has been developed in a two-way system

From past few years cloud computing has become an IT slang. It is the grassland of computing that is spreading briskly day-by-day in scholastic and business in form to accomplish conditions of final users. Cloud computing in real sense is achieving any duty by building use of assistance that are granted by cloud providers. It empowers an ample dimension of customer to access Shared, Ascendable, and Virtualized resources on top of the World Wide Web. It is a chunk of shared computing. It is currently arriving grassland by dint of its act, towering opportunity, and small price. With the brisk rise of internet computing machinery there is a big requirement for data storage security on cloud. There endure many cloud data storage security techniques now a days. In this paper, we are declaring some of these current techniques, literature review, and comparison between them. From the given comparison table of various current cloud computing data security approaches, it is concluded that the best approach for cloud data storage security must be taken and considered. We believe that there exist many more approaches for cloud data storage security and among them the best approach must be chosen.

Abstract We are introducing and implementing a novel method for cloud data security in this paper. In this novel method, five arithmetic operations which are basic (i.e. Addition, Subtraction, Multiplication, Division and Modulo) are performed on ASCII digits of every cover text character and these values which are newly generated are used for encryption of plain text parallelly. These each new generated values and two plain text characters in ASCII code will be encrypted in parallel way from beginning and from end of plain text. One cover text character hides at most ten plain text characters in our proposed approach. Since only one cover text character is used therefore less memory is required for storing cover text on cloud storage. Since we are using basic arithmetic operations therefore execution time is reduced and due to our parallelism approach, performance of overall system is enhanced.

The innovation of cloud computing technologies gained huge interest wherein more individuals started to outsource their data to cloud-based servers. However, open networks and untrusted cloud platforms face privacy and security issues while outsourcing data to the cloud. The proposed Blockchain-based data sharing mechanism effectively solves failure issues of the cloud storage system and provides several benefits by maximizing throughput. This framework embeds the interplanetary file system and blockchain for effective data sharing between cloud owners and the requester. Here, the system comprises five different entities and the proposed method involves six different phases. This model acquires interplanetary file system (IPFS) for effective data sharing between cloud owners and the requester. The proposed method provides superior performance in contrast to other techniques with the highest genuine detection rate of 0.880, minimal communication overhead of 0.642, and maximal private rate of 0.564.