Search

The field of artificial intelligence (AI) and machine learning (ML) has been expanding and is explored by researchers in various fields. In medical diagnosis, for instance, the field of AI/ML is being explored because if medical diagnostic devices are built and designed with a backend of AI/ML, then the benefits would be unprecedented. Automated diagnostic tools would result in reduced health care costs, diagnosis without human intervention, overcoming human errors, and providing adequate and affordable medical care to a wider portion of the population with portions of the actual cost. One domain where AI/ML can make an immediate impact is medical imaging diagnosis (MID), namely the classification of medical images, where researchers have applied optimization techniques aiming to improve image classification accuracy. In this paper, we provide the research community with a comprehensive review of the most relevant studies to date on the use of deep CNN architecture optimization techniques for MID. As a case study, the application of these techniques to COVID-19 medical images were made. The impacts of the related variables, including datasets and AI/ML techniques, were investigated in detail. Additionally, the significant shortcomings and challenges of the techniques were touched upon. We concluded our work by affirming that the application of AI/ML techniques for MID will continue for many years to come, and the performance of the AI/ML classification techniques will continue to increase.

Appendicitis is a condition wherein the appendix becomes inflamed, and it can be difficult to diagnose accurately. The type of appendicitis can also be hard to determine, leading to misdiagnosis and difficulty in managing the condition. To avoid complications and reduce mortality, early diagnosis and treatment are crucial. While Alvarado’s clinical scoring system is not sufficient, ultrasound and computed tomography (CT) imaging are effective but have downsides such as operator-dependency and radiation exposure. This study proposes the use of machine learning methods and a locally collected reliable dataset to enhance the identification of acute appendicitis while detecting the differences between complicated and non-complicated appendicitis. Machine learning can help reduce diagnostic errors and improve treatment decisions. This study conducted four different experiments using various ML algorithms, including K-nearest neighbors (KNN), DT, bagging, and stacking. The experimental results showed that the stacking model had the highest training accuracy, test set accuracy, precision, and F1 score, which were 97.51%, 92.63%, 95.29%, and 92.04%, respectively. Feature importance and explainable AI (XAI) identified neutrophils, WBC_Count, Total_LOS, P_O_LOS, and Symptoms_Days as the principal features that significantly affected the performance of the model. Based on the outcomes and feedback from medical health professionals, the scheme is promising in terms of its effectiveness in diagnosing of acute appendicitis.

Medical health records hold personal medical information and should only be accessed by authorized medical personnel or concerned patients. The importance of medical health records privacy is increasing as these records are shared in cloud environments. In this paper, we propose an enhanced system for securing patient data (Medical Health Records) embedded in multiple medical images in fragments for secure transmission and public sharing on the cloud or other environments. To protect the patient’s privacy, Medical Records are first encrypted, and then the ciphertext is broken into several fragments based on the number of multimodal medical images of a patient. A key generator randomly selects medical images from the multimodal image data to embed the encrypted patient health record segment using a modified least significant bit embedding process. The proposed technique enables an extra layer of security as even if files fall into the wrong hands and a fragment of the file is decrypted, it will not present any understandable information until all fragments from other medical images are extracted and combined in the correct order. The experiments are performed using multimodal 3255 MRI scans of 21 patients. The robustness of the proposed method was measured using different metrics such as PSNR, MSE, and SSIM. The results show that the proposed system is robust and that image quality is also maintained. To further study the stego image quality, a deep learning-based classification was applied to the images, and the results show that the diagnosis using stego medical images and performance remains unaffected even after embedding the encrypted data.

Digital Twins (DTs) have been gaining popularity in various applications, such as smart manufacturing, smart energy, smart mobility, and smart healthcare. In simple terms, DT is described as a virtual replica of a given physical product, system, or process. It consists of three major segments: the physical entity, its virtual counterpart, and the connections between them. While the data is collected from a physical entity, processed at the virtual layer, and accessed in the form of a DT at the application layer, it is exposed to several security risks. To ensure the applicability of a DT system, it is imperative to understand these security risks and their implications. However, there is a lack of a framework that can be used to assess the security of a DT. This paper presents a framework in which the security of a DT can be analyzed with the help of a formal verification technique. The framework captures the defense of the system at different layers and considers various attacks at each layer. The security of the DT system is represented as a state-transition system and the security properties are captured in temporal logic. Probabilistic model checking (PMC) is used to verify the systems against these properties. In particular, the framework is used to analyze the probability of success and the cost of various potential attacks that can occur at each layer in a DT system. The applicability of the proposed framework is demonstrated with the help of a detailed case study in the healthcare domain.

A dataset of fully labeled images of 20 different kinds of fruits is developed for research purposes in the area of detection, recognition, and classification of fruits. Applications can range from fruit recognition to calorie estimation, and other innovative applications. Using this dataset, researchers are given the opportunity to research and develop automatic systems for the detection and recognition of fruit images using deep learning algorithms, computer vision, and machine learning algorithms. The main contribution is a very large dataset of fully labeled images that are publicly accessible and available for all researchers free of charge. The dataset is called “DeepFruit”, which consists of 21,122 fruit images for 8 different fruit set combinations. Each image contains a different combination of four or five fruits. The fruit images were captured on different plate sizes, shapes, and colors with varying angles, brightness levels, and distances. The dataset images were captured with various angles and distances but could be cleared by utilizing the preprocessing techniques that allow for noise removal, centering of the image, and others. Preprocessing was done on the dataset such as image rotation & cropping, scale normalization, and others to make the images uniform. The dataset is randomly partitioned into an 80% training set (16,899 images) and a 20% testing set (4,223 images). The dataset along with the labels is publicly accessible at: https://data.mendeley.com/datasets/5prc54r4rt.

Abstract A critical requirement in Mobile Ad Hoc Networks (MANETs) is its ability to automatically discover existing services as well as their locations. Several solutions have been proposed in various communication domains which could be classified into two categories: (1) directory based, and (2) directory-less. The former is efficient but suffers from the amount of control messages being exchanged to maintain all directories in an agile environment. However, the latter approach attempts to reduce the amount of control messages to update directories, by simply sending broadcast messages to discover services; which is also a non-desirable approach in MANETs. This research work builds on top of our prior work (Nazeeruddin et al. in IFIP/IEEE international conference on management of multimedia networks and services, Springer, Berlin, 2006)) where we introduced a new efficient protocol for service discovery in MANETs (MSLD); a lightweight, robust, scalable, and flexible protocol which supports node heterogeneity and dynamically adapts to network changes while not flooding the network with extra protocol messages—a major challenge in today’s network environments, such as Internet of Things (IoT). Extensive simulations study was conducted on MSLD to: (1) initially evaluate its performance in terms of latency, service availability, and overhead messages, then (2) compare its performance to Dir-Based, Dir-less, and PDP protocols under various network conditions. For most performance metrics, simulation results show that MSLD outperforms Dir-Based, Dir-less, and PDP by either matching or achieving high service availability, low service discovery latency, and considerably less communication overhead.

Author verification of handwritten text is required in several application domains and has drawn a lot of attention within the research community due to its importance. Though, several approaches have been proposed for the text-independent writer verification of handwritten text, none of these have addressed the problem domain where author verification is sought based on partially-damaged handwritten documents (e.g., during forensic analysis). In this paper, we propose an approach for offline text-independent writer verification of handwritten Arabic text based on individual character shapes (within the Arabic alphabet). The proposed approach enables writer verification for partially damaged documents where certain handwritten characters can still be extracted from the damaged document. We also provide a mechanism to identify which Arabic characters are more effective during the writer verification process. We have collected a new dataset, Arabic Handwritten Alphabet, Words and Paragraphs Per User (AHAWP), for this purpose in a classroom setting with 82 different users. The dataset consists of 53,199 user-written isolated Arabic characters, 8,144 Arabic words, 10,780 characters extracted from these words. Convolutional neural network (CNN) based models are developed for verification of writers based on individual characters with an accuracy of 94% for isolated character shapes and 90% for extracted character shapes. Our proposed approach provided up to 95% writer verification accuracy for partially damaged documents.

The emerging 5G mobile network technology is envisioned to provide an efficient platform to interconnect machines, objects, and devices in addition to interconnecting people. Equipped with peak data rates, low latency, and massive capacity, 5G technology will empower new user experiences such as virtual reality and augmented reality and provide new service areas such as connecting massive IoT. Dual connectivity is an important feature where 5G systems are overlaid on the existing 4G core network. In this paper, we propose an MM (mobility management) algorithm to efficiently perform handovers between 4G and 5G RATs (radio access technologies). Our proposed MM algorithm utilizes the strength of DC (dual connectivity) for MM as DC inherently has lesser amount of handover interruption as compared to conventional hard handover. Our MM scheme suggests appropriate data split mechanism between 4G and 5G RATs based on application-specific strategy. We provide a framework based on probabilistic model checking that leverages DC and suggests strategy-based data split mechanism for a mobile user for a variety of market verticals. We model the system as MDP (Markov decision process) where a controller breaks all the nondeterminism in the MDP based on reward calculations. The proposed framework is implemented in a well known model checker and various scenarios are used to assess its applicability.

The increasing volume of unsolicited bulk emails has become a major threat to global security. While a significant amount of research has been carried out in terms of proposing new and better algorithms for email spam detection, relatively less attention has been given to evaluation metrics. Some widely used metrics include accuracy, recall, precision, and F-score. This paper proposes a new evaluation metric based on the concepts of fuzzy logic. The proposed metric, termed μO, combines accuracy, recall, and precision into a multi-criteria fuzzy function. Several possible evaluation rules are proposed. As proof of concept, a preliminary empirical analysis of the proposed scheme is carried out using two models, namely BERT (Bidirectional Encoder Representations from Transformers) and LSTM (Long short-term memory) from the domain of deep learning, while utilizing three benchmark datasets. Results indicate that for the Enron and PU datasets, LSTM produces better results of μO, with the values in the range of 0.88 to 0.96, whereas BERT generates better values of μO in the range of 0.94 to 0.96 for Lingspam dataset. Furthermore, extrinsic evaluation confirms the effectiveness of the proposed fuzzy logic metric.

Drones are increasingly adopted to serve a smart city through their ability to render quick and adaptive services. They are also known as unmanned aerial vehicles (UAVs) and are deployed to conduct area surveillance, monitor road networks for traffic, deliver goods and observe environmental phenomena. Cyber threats posed through compromised drones contribute to sabotage in a smart city’s airspace, can prove to be catastrophic to its operations, and can also cause fatalities. In this contribution, we propose a machine learning-based approach for detecting hijacking, GPS signal jamming and denial of service (DoS) attacks that can be carried out against a drone. A detailed machine learning-based classification of drone datasets for the DJI Phantom 4 model, compromising both normal and malicious signatures, is conducted, and results obtained yield advisory to foster futuristic opportunities to safeguard a drone system against such cyber threats.

The design and development of future smart city services will usher with unprecedented creativity, a plethora of functionalities and conveniences to the urban communities. The success of smart city services depends on the assurance that security is maintained and protected to meet community expectations. However, due to the variety of components involved in the design of smart city services and their complex interactions, it becomes challenging to properly assess the overall security of the system throughout the service life cycle. This paper presents a holistic approach to model the security of smart city service infrastructure at a higher level of abstraction. To match realistic security deployments, the proposed model is based on multiple tiers of defense that includes defense at component level, system level, and security operation center. For a given system configuration and component vulnerability vector, the model assesses key security parameters for a variety of attacks. The proposed model is generic and can be used for studying the security of various smart city infrastructures, deployment configurations, and attack vectors.

The concept of IoT-based Smart Cities has gained momentum in recent years. The research in this domain has focused on modeling key characteristics of future smart cities along with exploring their design and implementation aspects from multiple perspectives. There is, however, a lack of research effort to provide a holistic approach towards modeling smart city services. A comprehensive approach has to view the smart city as a dynamic, uncertain and complex environment where multiple events with varying severity occur in a continuous and non-deterministic manner. These events have to be handled efficiently with the available resources. In this paper, we present a holistic approach to model the probabilistic as well as non-deterministic aspects of the emergency management services of a smart city by using probabilistic model checking. Our proposed model captures the emergency events of varying severity occurring at several locations in a continuous and non-deterministic manner. These events are responded to by a smart emergency management unit (SEMU), which dispatches the required emergency response units (ERUs) to the event location. In addition to modeling a completely automated system, we introduce a human-assisted decision-making process in the model to reduce false alarms. The proposed model can be used to study and evaluate key parameters in designing a smart city emergency management system to meet given service level agreements. We have implemented our proposed model using the PRISM model checker and have conducted several case studies to highlight the effectiveness of our proposed approach for different scenarios of varying complexities.

Classification of brain tumor is one of the most vital tasks within medical image processing. Classification of images greatly depends on the features extracted from the image, and thus, feature extraction plays a great role in the correct classification of images. In this paper, an enhanced method is presented for glioma MR images classification using hybrid statistical and wavelet features. In the proposed method, 52 features are extracted using the first-order and second-order statistical features (based on the four MRI modalities: Flair, T1, T1c, and T2) in addition to the discrete wavelet transform producing a total of 152 features. The extracted features are applied to the multilayer perceptron (MLP) classifier. The results using the MLP were compared with various known classifiers. The method was tested on the dataset MICCAI BraTS 2015 which is a standard dataset used for research purposes. The proposed hybrid statistical and wavelet features produced 96.72% accuracy for high-grade glioma and 96.04% accuracy for low-grade glioma, which are relatively better compared to the existing studies.

It became apparent that mankind has to learn to live with and adapt to COVID-19, especially because the developed vaccines thus far do not prevent the infection but rather just reduce the severity of the symptoms. The manual classification and diagnosis of COVID-19 pneumonia requires specialized personnel and is time consuming and very costly. On the other hand, automatic diagnosis would allow for real-time diagnosis without human intervention resulting in reduced costs. Therefore, the objective of this research is to propose a novel optimized Deep Learning (DL) approach for the automatic classification and diagnosis of COVID-19 pneumonia using X-ray images. For this purpose, a publicly available dataset of chest X-rays on Kaggle was used in this study. The dataset was developed over three stages in a quest to have a unified COVID-19 entities dataset available for researchers. The dataset consists of 21,165 anterior-to-posterior and posterior-to-anterior chest X-ray images classified as: Normal (48%), COVID-19 (17%), Lung Opacity (28%) and Viral Pneumonia (6%). Data Augmentation was also applied to increase the dataset size to enhance the reliability of results by preventing overfitting. An optimized DL approach is implemented in which chest X-ray images go through a three-stage process. Image Enhancement is performed in the first stage, followed by Data Augmentation stage and in the final stage the results are fed to the Transfer Learning algorithms (AlexNet, GoogleNet, VGG16, VGG19, and DenseNet) where the images are classified and diagnosed. Extensive experiments were performed under various scenarios, which led to achieving the highest classification accuracy of 95.63% through the application of VGG16 transfer learning algorithm on the augmented enhanced dataset with freeze weights. This accuracy was found to be better as compared to the results reported by other methods in the recent literature. Thus, the proposed approach proved superior in performance as compared with that of other similar approaches in the extant literature, and it made a valuable contribution to the body of knowledge. Although the results achieved so far are promising, further work is planned to correlate the results of the proposed approach with clinical observations to further enhance the efficiency and accuracy of COVID-19 diagnosis.

A fully-labelled dataset of Arabic Sign Language (ArSL) images is developed for research related to sign language recognition. The dataset will provide researcher the opportunity to investigate and develop automated systems for the deaf and hard of hearing people using machine learning, computer vision and deep learning algorithms. The contribution is a large fully-labelled dataset for Arabic Sign Language (ArSL) which is made publically available and free for all researchers. The dataset which is named ArSL2018 consists of 54,049 images for the 32 Arabic sign language sign and alphabets collected from 40 participants in different age groups. Different dimensions and different variations were present in images which can be cleared using pre-processing techniques to remove noise, center the image, etc. The dataset is made available publicly at https://data.mendeley.com/datasets/y7pckrw6z2/1.

One of the key challenges in cyber-physical systems (CPS) is the dynamic fitting of data sources under multivariate or mixture distribution models to determine abnormalities. Equations of the models have been statistically characterized as nonlinear and non-Gaussian ones, where data have high variations between normal and suspicious data distributions. To address nonlinear equations of these distributions, a cuckoo search algorithm is employed. In this paper, the cuckoo search algorithm is effectively improved with a novel strategy, known as a convergence speed strategy, to accelerate the convergence speed in the direction of the optimal solution for achieving better outcomes in a small number of iterations when solving systems of nonlinear equations. The proposed algorithm is named an improved cuckoo search algorithm (ICSA), which accelerates the convergence speed by improving the fitness values of function evaluations compared to the existing algorithms. To assess the efficacy of ICSA, 34 common nonlinear equations that fit the nature of cybersecurity models are adopted to show if ICSA can reach better outcomes with high convergence speed or not. ICSA has been compared with several well-known, well-established optimization algorithms, such as the slime mould optimizer, salp swarm, cuckoo search, marine predators, bat, and flower pollination algorithms. Experimental outcomes have revealed that ICSA is superior to the other in terms of the convergence speed and final accuracy, and this makes a promising alternative to the existing algorithm.

The black phase of formamidinium lead iodide (FAPbI3) perovskite shows huge promise as an efficient photovoltaic, but it is not favoured energetically at room temperature, meaning that the undesirable yellow phases are always present alongside it during crystallization1-4. This problem has made it difficult to formulate the fast crystallization process of perovskite and develop guidelines governing the formation of black-phase FAPbI3 (refs. 5,6). Here we use in situ monitoring of the perovskite crystallization process to report an oriented nucleation mechanism that can help to avoid the presence of undesirable phases and improve the performance of photovoltaic devices in different film-processing scenarios. The resulting device has a demonstrated power-conversion efficiency of 25.4% (certified 25.0%) and the module, which has an area of 27.83 cm2, has achieved an impressive certified aperture efficiency of 21.4%.