Diabetic Retinopathy (DR) is a predominant cause of visual impairment and loss. Approximately 285 million worldwide population is affected with diabetes, and one-third of these patients have symptoms of DR. Specifically, it tends to affect the patients with 20 years or more with diabetes, but it can be reduced by early detection and proper treatment. Diagnosis of DR by using manual methods is a time-consuming and expensive task which involves trained ophthalmologists to observe and evaluate DR using digital fundus images of the retina. This study aims to systematically find and analyze high-quality research work for the diagnosis of DR using deep learning approaches. This research comprehends the DR grading, staging protocols and also presents the DR taxonomy. Furthermore, identifies, compares, and investigates the deep learning-based algorithms, techniques, and, methods for classifying DR stages. Various publicly available dataset used for deep learning have also been analyzed and dispensed for descriptive and empirical understanding for real-time DR applications. Our in-depth study shows that in the last few years there has been an increasing inclination towards deep learning approaches. 35% of the studies have used Convolutional Neural Networks (CNNs), 26% implemented the Ensemble CNN (ECNN) and, 13% Deep Neural Networks (DNN) are amongst the most used algorithms for the DR classification. Thus using the deep learning algorithms for DR diagnostics have future research potential for DR early detection and prevention based solution.

Aiming to increase the shelf life of food, researchers are moving toward new methodologies to maintain the quality of food as food grains are susceptible to spoilage due to precipitation, humidity, temperature, and a variety of other influences. As a result, efficient food spoilage tracking schemes are required to sustain food quality levels. We have designed a prototype to track food quality and to manage storage systems at home. Initially, we have employed a Convolutional Neural Network (CNN) model to detect the type of fruit and veggies. Then the proposed system monitors the gas emission level, humidity level, and temperature of fruits and veggies by using sensors and actuators to check the food spoilage level. This would additionally control the environment and avoid food spoilage wherever possible. Additionally, the food spoilage level is informed to the customer by an alert message sent to their registered mobile numbers based on the freshness and condition of the food. The model employed proved to have an accuracy rate of 95%. Finally, the experiment is successful in increasing the shelf life of some categories of food by 2 days.

Distributed Denial of Service (DDoS) attack is known to be one of the most lethal attacks in traditional network architecture. In this attack, the attacker uses botnets to overwhelm network resources. Botnets can be randomly compromised computers or IoT devices that are used to generate excessive traffic towards the victim, and as a result, legitimate users cannot access the services. In this research, software-defined networking (SDN) has been suggested as a solution to fight DDoS attacks. SDN uses the idea of centralized control and segregation of the data plane from the control plane. SDN is more flexible, and policy implementation on the centralized controller is easy. SDN is now being widely used in modern network paradigms because it has enhanced security. In this work, an entropy-based statistical approach has been suggested to detect and mitigate TCP SYN flood DDoS attacks. The proposed algorithm uses a three-phased detection scheme to minimize the false-positive rate. Entropy, standard deviation, and weighted moving average have been used for intrusion detection. Multiple experiments were performed, and the results show that the suggested approach is more reliable and lightweight and has a minimal false-positive rate.

Advancement in technology has led to innovation in equipment, and the number of devices is increasing every day. Industries are introducing new devices every day and predicting 50 billion connected devices by 2022. These devices are deployed through the Internet, called the Internet of Things (IoT). Applications of IoT devices are weather prediction, monitoring surgery in hospitals, identification of animals using biochips, providing tracking connectivity in automobiles, smart home appliances, etc. IoT devices have limitations related to security at both the software and hardware ends. Secure user interfaces can overcome software-level limitations like front-end-user interfaces are accessed easily through public and private networks. The front-end interfaces are connected to the localized storage to contain data produced by the IoT devices. Localized storage deployed in a closed environment connected to IoT devices is more efficient than online servers from a security perspective. Blockchain has emerged as a technology or technique with capabilities to achieve secure administrational authentication and accessibility to IoT devices and their computationally produced data in a decentralized way with high reliability, interrogation, and resilience. In this paper, we propose device, end-user, and transactional authentication techniques using blockchain-embedded algorithms. The localized server interacts with the user interface to authenticate IoT devices, end-users, and their access to IoT devices. The localized server provides efficiency by reducing the load on the IoT devices by carrying out end-user heavy computational data, including end-user, IoT device authentication, and communicational transactions. Authentication data are placed on the public ledger in block form, distributed over the system nodes through blockchain algorithms.

Designing an efficient, reliable, and stable algorithm for underwater acoustic wireless sensor networks (UA-WSNs) needs immense attention. It is due to their notable and distinctive challenges. To address the difficulties and challenges, the article introduces two algorithms: the multilayer sink (MuLSi) algorithm and its reliable version MuLSi-Co using the cooperation technique. The first algorithm proposes a multilayered network structure instead of a solid single structure and sinks placement at the optimal position, which reduces multiple hops communication. Moreover, the best forwarder selection amongst the nodes based on nodes’ closeness to the sink is a good choice. As a result, it makes the network perform better. Unlike the traditional algorithms, the proposed scheme does not need location information about nodes. However, the MuLSi algorithm does not fulfill the requirement of reliable operation due to a single link. Therefore, the MuLSi-Co algorithm utilizes nodes’collaborative behavior for reliable information. In cooperation, the receiver has multiple copies of the same data. Then, it combines these packets for the purpose of correct data reception. The data forwarding by the relay without any latency eliminates the synchronization problem. Moreover, the overhearing of the data gets rid of duplicate transmissions. The proposed schemes are superior in energy cost and reliable exchanging of data and have more alive and less dead nodes.

For a better detection in Network information security monitoring system, the author proposes a method based on adaptive depth detection. A deep belief network (DBN) was designed and implemented, and the intrusion detection system model was combined with a support vector machine (SVM). The data set adopts the NSL-KDD network communication data set, and this data set is authoritative in the security field. Redundant cleaning, data type conversion, normalization, and other processing operations are performed on the data set. Using the data conversion method based on the probability mass function probability mass function coding, a standard data set with low redundancy and low dimensionality can be obtained. Research indicates that when the batch size reaches 64, the accuracy of the test set reaches its maximum value. As the batch size increases, the accuracy first increases and then decreases. When the batch size continues to increase, the model will inevitably fall into the local optimal state, resulting in the degradation of the detection performance of the system. In terms of the false alarm rate, the DBN-SVM model is also the highest; however, it is only 10.73%. Under the premise of increasing the detection rate, the false alarm rate is improved; for the overall detection performance of the model, it is within an acceptable range. In terms of accuracy, the DBN-SVM model also scored the highest. The accuracy rate is the ratio of normal and correct classification for intrusion detection. It can explain the detection ability of the model. In summary, the overall detection ability of the DBN-SVM model is the best. The good classification ability to use SVM is proved, and the classification of low-dimensional features is expected to increase the detection rate of the system.

Underwater wireless sensor networks (UWSNs) have emerged as the most widely used wireless network infrastructure in many applications. Sensing nodes are frequently deployed in hostile aquatic environments in order to collect data on resources that are severely limited in terms of transmission time and bandwidth. Since underwater information is very sensitive and unique, the authentication of users is very important to access the data and information. UWSNs have unique communication and computation needs that are not met by the existing digital signature techniques. As a result, a lightweight signature scheme is required to meet the communication and computation requirements. In this research, we present a Certificateless Online/Offline Signature (COOS) mechanism for UWSNs. The proposed scheme is based on the concept of a hyperelliptic curves cryptosystem, which offers the same degree of security as RSA, bilinear pairing, and elliptic curve cryptosystems (ECC) but with a smaller key size. In addition, the proposed scheme was proven secure in the random oracle model under the hyperelliptic curve discrete logarithm problem. A security analysis was also carried out, as well as comparisons with appropriate current online/offline signature schemes. The comparison demonstrated that the proposed scheme is superior to the existing schemes in terms of both security and efficiency. Additionally, we also employed the fuzzy-based Evaluation-based Distance from Average Solutions (EDAS) technique to demonstrate the effectiveness of the proposed scheme.

Fruits significantly impact everyday living, i.e., Citrus fruits. Numerous fruits have a solid nutritious value and are packed with multivitamins and trace components. Citrus fruits are delicate and susceptible to many diseases and infections. Many researchers have suggested deep and machine learning-based fruit disease detection and classification models. This research presents a precise fruit disease identification model based on context data fusion with Faster-CNN in an edge computing environment. The goal is to develop an accurate, efficient, and trustable fruit disease detection model, a critical component of autonomous food production in a robotic edge platform. This research examines and explores four different diseases of Citrus fruits using CNN deep learning models to be adopted as edge computing solutions. Identification of citrus diseases such as cankers black spot, greening, scab, melanosis, and healthy citrus fruits are implemented using the proposed sequential model without pruning, with pruning having different sparsity levels followed by post quantization. Through the transfer learning method, this model is optimized for the assignment of fruit disease detection employing visuals from two patterns: Near-infrared (NIFR) and RGB. Early and late data fusion techniques for integrating multi-model (NIFR and RGB) facts are evaluated. The accuracy obtained from the proposed model for the canker disease is 97%, scab 95%, melanosis 99%, Greening 97%, Black spot 97% and healthy 97%. In this paper, the results of the proposed model are compared and evaluated with the sparsity levels of 50–80%, 60–90%, 70–90%, and 80–90% pruning and also obtained the results of post-quantization on each level. The results show that the model size with 60–90% pruning can be counteracted to the 47.64 of the baseline model without significant loss of accuracy. Moreover, post-quantization can reduce the 60–90% pruning from 28.16 to 8.72. In addition to enhanced precision, the above initiative is much faster to implement for new fruit diseases because it needs bounding box annotation instead of pixel-level annotation.