This study investigates public attitudes towards the COVID-19 vaccine through Twitter data analysis. Using the Twitter API, tweets were collected, preprocessed, and labeled. Features were extracted using the Bag of Words representation, and sentiment analysis was conducted using Text Blob and Vader. Machine learning and deep learning models were trained and tested, revealing that deep learning models achieved the highest accuracy and F1 score. The research underscores the efficacy of machine learning and deep learning in analyzing COVID-19 vaccine-related tweets, shedding light on factors influencing vaccine resistance. These insights are crucial for pharmaceutical companies and public health officials, enabling them to address barriers to vaccine acceptance and enhance the societal benefits of widespread vaccination, contributing to the pandemic's resolution.
Due to the vast and fast growth of the internet, privacy is becoming a concern. Web browser data privacy is a part of the overall users' privacy. Since the attackers took social engineering to the next level, achieving the web browser history data secured from privacy breaches becomes difficult. In this work, we implemented three types of web fingerprinting attack using ML algorithms (DT_Attack, RF_Attack and AdaB_Attack) with the best performance results (Accuracy 99.9%, precision 100%, recall 100% and execution time 339 seconds) achieved using the Decision Tree classifier (DT_Attack). Then we proposed a defense mechanism using the forward feature selection algorithm named Reverse Forward Feature selection, which contributes to reducing the efficiency of the previously implemented web fingerprinting attacks. More specifically, accuracy, precision, and recall of the implemented attacks were reduced on all attack types between 44.8-54.3%, 24.3-38.6%, 53.6-70.2% respectively, without adding any overhead in terms of bandwidth and execution time.
Brain anatomy presents a unique pedagogical challenge in medical education due to its complex structures and the detailed intricacies inherent to various brain regions. Advanced imaging techniques such as computed tomography (CT), T1/T2 magnetic resonance imaging (MRI), and diffusion imaging provide 3D representations of these anatomical details, which require suitable tools and visualization setups to be taught efficiently to medical students. Machine learning and virtual reality have emerged as promising avenues for refining the visualization and reconstruction processes of these details. This paper delves into the integration of cutting-edge deep learning (DL) algorithms designed for the reconstruction of brain vascular trees and neural fiber tracts within an immersive virtual reality (VR) framework. Our proposed solution has the potential to improve brain anatomy education by serving as tool for illustrating subject-specific complex structures. The ensuing results underscore the viability and efficacy of our solution.
TIn an age dominated by internet usage, the threat of phishing attacks continues to plague users and organizations alike. Traditional cybersecurity mechanisms often fail to cope with the evolving tactics of cybercriminals, leading to a growing interest in machine learning-based solutions. This research evaluates the performance of four different machine learning classifiers-K-Nearest Neighbors (KNN), Naïve Bayes (NB), Decision Tree, and Artificial Neural Network (ANN)-in identifying phishing activities. A Python code framework is utilized for this evaluation, employing stratified k-fold cross-validation for reliable results. Multiple performance metrics indicate that machine learning provides an adaptable and effective tool against phishing. This research serves as a stepping stone for future development in this crucial area, urging continuous advancements to stay ahead of ever-evolving cybersecurity threats.
The detection of missing security operations is a complex task in software engineering, mainly due to the semantic and contextual understanding required. Prior research efforts have employed similar path differential analysis to detect missing security operations, but these approaches have been limited in their ability to simultaneously compare the similarity of intra- and inter-procedural paths. To address this limitation, this paper proposes a novel approach called SSD that can detect multiple missing security operation bugs both intra- and inter-procedurally. Our approach collects slices with similar semantics and contexts based on four program slicing criteria, providing more versatile construction of similar slices and more comprehensive detection than previous works. In our experiments, we have identified 65 real bugs in the Linux kernel, of which we have verified 27 as fixed bugs and submitted the remaining 38 for patching. The Linux maintainers have accepted 19 of these patches, confirming the effectiveness and availability of SSD.
To limit the spread of infectious diseases, ensuring vaccine safety and reliability is paramount. The introduction of new COVID-19 vaccines has been accompanied by a notable number of reported adverse reactions. Given the diversity of allergies among individuals, it is crucial to consider these factors before administering the vaccine. This study aims to enhance decision-making for alternative medical procedures in individuals allergic to COVID-19 vaccines by employing machine learning algorithms and evaluation techniques. Specifically, we utilize machine learning to categorize side effects of the COVID-19 vaccine in individuals with allergies related to food, animals, and weather.
Cloud data centers, comprising a diverse set of heterogeneous resources working collaboratively to achieve high-performance computing, face the challenge of resource dynamism, where performance fluctuates over time. This dynamism poses complexities in task scheduling, warranting further research on the resilience of existing static task scheduling algorithms when deployed in dynamic cloud environments. This study adapts three well-known task scheduling algorithms to the cloud computing context and conducts a comprehensive comparison to assess their resilience to dynamic conditions. The evaluation, employing simulation techniques, analyzes total energy consumption and total response time as key metrics. The results offer detailed insights into the effectiveness of the adapted algorithms, providing valuable guidance for optimizing task scheduling in dynamic cloud data centers.
Natural language processing (NLP) and sentiment analysis empower artificial intelligence systems to identify the perspectives and emotions conveyed in text. This capability has grown in importance and accuracy in recent times. ChatGPT, an OpenAI-developed language model, has garnered considerable attention due to its exceptional NLP abilities. In this paper, ChatGPT's effectiveness in emotion classification, using the Dair-ai/emotion dataset [1], is evaluated. ChatGPT has achieved an accuracy of 58%.
Graph clustering is a technique that used to uncover the structural patterns in the graph and reduce its dimension. Multiple algorithms have been developed to approach the task of graph clustering. Despite the progress that has been done, there are still many challenges in graph clustering such as clustering noisy graphs. In this paper, we introduce a graph clustering algorithm that approximates the graph's similarity matrix into a lower-rank one while conserving its fundamental structure and eliminating noise and redundancy. The approximated lower-rank matrix is then used to reveal the clusters through symmetric non-negative matrix factorization. The problem is solved through an alternating minimization framework. The conducted experiments on real-world graphs show that the proposed model enables efficient and accurate clustering results better than other existing algorithms.
Upgrading to the sixth generation (6G) network will bring new potential technologies for communication standards beyond the fifth generation (5G). Radio over free space optical (RoFSO) is a new version of optical communication that uses the atmospheric channel for high-data-rate transmission due to the large bandwidth it has. However, the signal attenuation caused by atmospheric conditions is the major factor that can affect the transmission of free space optics (FSO) and degrade ROFSO system performance. This paper investigates the performance of 60 GHz millimeter waves over FSO systems for ultradata rate delivery under different conditions, especially under the effect of oxygen (O 2 ) absorption, as it considers the primary attenuation at approximately 60 GHz. A validation of the previous 5G-RoFSO research work is efficiently recognized to prove the effectiveness of our work on the 6G band over FSO in the presence of atmospheric factors. Incorporating a Mach-Zehnder modulator (MZM) into the 60 GHz RoFSO link enables the delivery of different data rates up to 70 Gbps, thereby improving the spectral efficiency of the 6G-RoFSO system. The design is performed using Optisystem v.19 software. The analysis depicted that 60 GHz offers 20 Gbps up to a distance of 2.2 km and shows increases in the data rate up to 60 Gbps for a distance of up to 2 km and 70 Gbps up to 1.4 km with a bit error rate (BER) less than -9 dB. A comparison analysis is performed between the 26 GHz 5G and 60 GHz 6G bands at the same 20 Gbps data rate. The analysis depicted a lower distance performance for the 60 GHz RoFSO link, reaching up to 2.2 km with -11 dB, compared to the 26 GHz RoFSO link, reaching up to 4 km with -11 dB. The results show that the 60 GHz-6G band can achieve a higher data rate of up to 70 Gbps than the 26 GHz-5G band, which achieves up to 20 GHz.
To decrease time delay, energy consumption, and network utilization in the Internet of Things (IoT) is a key research area. The rapid advancement in IoT, leading to increase in size and cost makes traditional protocols and algorithms unfeasible to work with and requires redevelopment. Effective redevelopment may result in less energy consumption and resource utilization. Task scheduling algorithms on the Fog platform are one of the most time-consuming areas in IoT because it defines flow and queue for task processing. In time-critical applications like healthcare, an immediate response and access is required to the Fog platform. In that framework, this study evaluates a case study of healthcare applications to propose an efficient task scheduling algorithm known as Critical Task Indexing Scheduler (CTIS), that reduces the total time delay, energy consumption, and network utilization by indexing a task according to its criticality determined by its pre classified source. When compared with its predecessors such as First Come First Serve (FCFS), Shortest Job First (SJF), Critical Task First Scheduler (CTFS) and Cloud-only platforms, our proposed algorithm outperforms in all the three major parameters.
The rapid spread of TOR (The Onion Router) for anonymized communication and internet browsing raises relevant security concerns due to its frequent use in illicit activities. This research is centered around a comprehensive analysis of TOR artifacts on Windows 11 and the creation of a relevant dataset to aid researchers. The study commences with a thorough review and comparative assessment of the existing literature on TOR artifacts and traffic examination. Windows 11 TOR browser artifacts are identified and placed together with those in Windows 10 and preceding versions. The paper's core methodology involves establishing a virtual lab environment, facilitating integration, simulating TOR network traffic, and gathering network forensics artifacts. A comprehensive dataset, assimilated from traffic logs procured from tools such as Wireshark, snort, Suricata, and pfsense firewall, is designed and used to construct machine learning models for TOR traffic detection. Combining these traffic logs into a particular dataset enhances the accuracy and efficiency of the machine-learning models. The findings of this study provide important insights into the evolving landscape of TOR browser artifacts and traffic patterns on different Windows versions. The proposed models showcase impressive accuracy in identifying TOR traffic. This research is crucial in enhancing our knowledge of TOR traffic analysis. It lays the groundwork for further investigations to enhance the detection and prevention of malicious activities utilizing the TOR network. Moreover, it contributes to developing a valuable dataset that can aid ML researchers in future endeavors.
The most well-known and widely adopted digital currency, Bitcoin, was created in 2009. It is the most dominant cryptocurrency in trading volume, with around 0.47 of the total market volume. Most of the previous studies in the cryptocurrency field were done to predict Bitcoin prices, but not many of them focused on the altcoins. In this study, we focused on 2 of the most famous altcoins, Ethereum and Litecoin, to predict their prices using Long-short term memory. In this study, we explore the impact of the technical indicators and the correlation of Bitcoin with the altcoins. We found that there was a significant improvement when using technical indicators along with correlation with Bitcoin. The error decreased noticeably, especially in Lite coin, where MAPE decreased by as much as 0.41.
Increasing technology led to increasing collaboration between the countries in general, especially the Arabic world. In recent years, many impacts affect the growth of research work in different Arabic countries. In this research, we show the growth of research work and its quality between Arabs and the rest of the world. We collected over 2500 publications and used it to create a new table that contains the intersection between each Arab country and each co-author's continent to show the relationship between each Arabic country with the rest of the world. We divided the data into two periods, before and after 2015, then we used Hue to apply our analysis on the data. We focused on the number of publications and the number of citation for each Arabic country. We noticed the rapid growth in the last period in some Arabic countries such as Jordan, Qatar, United Arab Emirates, and Saudi Arabia. In contrast, the rest of the countries had slow or decreasing growth.
In this paper, a genetic algorithm is used to optimize of a Direct Current to Direct Current boost converter with inverted-gamma filter. The objective is to reduce the electro-magnetic interference (EMI) produced from the non-linear elements by minimizing the control-to-output transfer function. The optimization was done using MATLAB/Global optimization toolbox for 3 different initial populations, each population was optimized using 5 different genetic algorithm operations, for a total of 16 options and 48 runs. Experimental results show that the proposed approach achieved an average fitness function value of (-234.87), which is 5.41% better fitness function value compared to a previous related work.
Money laundering has significantly increased in recent years as a result of the quick growth of financial systems and the emergence of cryptocurrencies. These crimes threaten the stability of both economic and social systems and impact the security of public and private financial institutions. In response, detecting and preventing money laundering has become a top priority for financial systems, which rely on deep analysis of assets and the nature of funds to uncover illegitimate sources. With the rise of cryptocurrencies like Bitcoin, criminals have adopted new technologies and channels to cover up their illegal activities. This research aims to develop a new deep learning architecture for detecting illegitimate Bitcoin transactions in order to prevent Bitcoin laundering crimes. In this study, a deep investigation of the previous works will be conducted for better understanding of the problem statement. Besides, a new framework will be designed to exceed the performance of current best practices with regard to accuracy. Ultimately, the findings of this research will aid the financial sector in combating Bitcoin laundering activities and strengthening the security of digital financial transactions. An elliptic dataset used for bitcoin transactions that belong to real entities, where it consists of 203,769 nodes and 234,355 edges for illicit and licit transactions. The licit transaction families are licit services, exchanges, miners, and wallet providers. While the illicit transaction families are malware, frauds, Ponzi (fraud) schemes, ransomware, and terrorist organizations. The proposed model semi-supervised generative adversarial network (SGAN) utilized to complete the labelling process of the unknown entities. Based on the experimental results, the proposed model significantly improved upon previous methods, achieving a 98% success rate in accurately predicting illicit transactions. Thus, we consider it as a competitive stand in terms of anti-laundering for the Bitcoin cryptocurrency.
Although considerable improvements have been made in online signature verification (OSV) over the last decade, none of them take both temporal and spatial information into consideration, and thus there is still a room for boosting the performance. In this paper, we propose a novel ensemble based deep learning framework, which consists of a convolutional neural network model and our recently designed convolutional gated recurrent network (CGRN) for extracting spatial feature and temporal feature, respectively. However, it is not easy to combine these two types of features since temporal feature is two-dimensional with various length while the other is a fixed-length vector. In order to incorporate both types of representation, we firstly introduce cosine similarity for spatial feature to calculate the shape similarity and use dynamic time warping (DTW) for temporal feature alignment. Thereafter, the distance between reference signature and given signature is obtained by multiplying DTW distance and similarity score. In addition, we design a novel approach for DTW distance normalization, which significantly enhances the verification accuracy. Our method achieves new state-of-the-art result on DeepSignDB, and outperforms other existing OSV methods with at least 16.2
In recent years, the spread of fake news in social networks has become a serious threat to network security. To address this problem, various fake news detection methods have been proposed. However, most of the existing methods cannot jointly capture the intra-modal and inter-modal correlation relationships between image regions and text fragments, resulting in the model not making full use of multimodal information, thus limiting their ability to detect fake news accurately. To solve this limitation, we propose a novel fake news detection method based on a multimodal cooperative attention network (MCAND). Firstly, we use BERT and VGG19 to learn text and image representations, respectively. Secondly, the multimodal cooperative attention network is used to generate the high-order fusion features that fuse the image and text representations by calculating the similarity between the information segments in the modalities and the inter-modal similarity. Finally, the multimodal fusion features are input into the fake news detector to identify fake news. The experimental results show that the proposed MCAND has outperformed the state-of-the-art (SOTA) method in terms of performance, demonstrating its effectiveness.
Discrimination in decision-making systems is of growing concern as machine learning techniques (especially deep learning) are increasingly applied in systems with societal impact. Multiple recent works have proposed to identify/generate discriminative samples through fairness testing. State-of-the-art fairness testing methods can efficiently generate many discriminative samples, which can be subsequently used to improve the fairness of the model. Unfortunately, the applicability of these approaches is limited in practice as they require the availability of both the model and the training data, i.e., a white-box setting. In a black-box setting (e.g., testing online services), existing approaches are impractical for multiple reasons, e.g., they require huge testing budgets. In this work, we propose a black-box fairness testing approach for neural networks, namely BREAM, which addresses two challenges, i.e., how to generate many discriminative samples without querying many times and how to guide the searching without the original model. Our overall idea is to obtain approximate gradients by training shadow models to effectively guide the discriminative sample generation for black-box DNNs. We also observe the density diversity of the distribution of discrimination, which enables incremental maintenance of shadow models and rational allocation of search resources by dividing multiple subspaces. We evaluated BREAM on three widely adopted datasets for fairness research. The results show that BREAM achieves a 9X higher performance than existing black-box methods, comparable to the state-of-the-art white-box fairness method.
The satellite Internet of Things (satellite IoT) has the characteristics of large space-time span and highly open communication links. While effectively expanding the spatial capability of the traditional Internet of Things, it will face security threats such as impersonation, replay, tampering and eavesdropping of the traditional Internet of Things and satellite communication. In this paper, an SM2-based certificateless integrated signature and encryption scheme (SM2-CL-ISE) is proposed for satellite IoT with key optimization and conditional anonymity. Then incorporating Geostationary Earth Orbit (GEO) satellite, a Low Earth Orbit (LEO) satellite authentication protocol and a static terminal device authentication protocol are designed. In addition, we prove the security of SM2-CL-ISE under the formal security model, and further discuss how the proposed authentication schemes can satisfy those essential security requirements. To evaluate the effectiveness of our proposed protocols, we conducted several experiments and compared their performance with that of existing protocols. The experimental results show that our scheme achieves more efficient performance with a slightly increased communication overhead on authentication.
