Search

Cloud Computing and Big Data continue to be disruptive forces in computing and has introduced new threats and vulnerabilities to our networks. The paper seeks to demonstrate how an end-to-end network intrusion detection system can be built, trained, and deployed using Amazon Web Services (AWS), Microsoft Azure and Google Cloud Platform (GCP). We determined the performance of these tools by building a network intrusion detection system (NIDS) and evaluating the performance of each based on precision, accuracy, F1 Score, recall, user experience, cost and computation time for training and predicting the model. Overall, all three platforms performed greater than 90% accuracy with Google Vertex AI having the highest accuracy using the decision tree and Microsoft Azure performing the best based on accuracy, precision, and computation time. [ABSTRACT FROM AUTHOR]

Phishing and spam have been a cybersecurity threat with the majority of breaches resulting from these types of social engineering attacks. Therefore, detection has been a long‐standing challenge for both academic and industry researcher. New and innovative approaches are required to keep up with the growing sophistication of threat actors. One such illumination which has vast potential are large language models (LLM). LLM emerged and already demonstrated their potential to transform society and provide new and innovative approaches to solve well‐established challenges. Phishing and spam have caused financial hardships and lost time and resources to email users all over the world and frequently serve as an entry point for ransomware threat actors. While detection approaches exist, especially heuristic‐based approaches, LLMs offer the potential to venture into a new unexplored area for understanding and solving this challenge. LLMs have rapidly altered the landscape from business, consumers, and throughout academia and demonstrate transformational potential to profoundly impact the society. Based on this, applying these new and innovative approaches to email detection is a rational next step in academic research. In this work, we present IPSDM, an improved phishing spam detection model based on fine‐tuning the BERT family of models to specifically detect phishing and spam emails. We demonstrate our fine‐tuned version, IPSDM, is able to better classify emails in both unbalanced and balanced datasets. Moreover, IPSDM consistently outperforms the baseline models in terms of classification accuracy, precision, recall, and F1‐score, while concurrently mitigating overfitting concerns. [ABSTRACT FROM AUTHOR]

Generative Artificial Intelligence (AI) model is a revolutionary type of AI capable of producing high quality images based on textual inputs. These models utilize natural language processing (NLP) techniques and computer vision to understand and interpret the textual descriptions and then generate images that align with the given descriptions. This study evaluates four prominent text-to-image generative models- DALL-E, Google Imagen, Stable Diffusion, and GROK AI emphasizing on the text-to-image diffusion models. Using a comprehensive evaluation approach, we employ three mathematical formulas the Fréchet Inception Distance (FID), Structural Similarity Index (SSIM), and Peak Signal-to-Noise Ratio (PSNR) to assess image quality and realism across datasets collected from these AI platforms. Additionally, human evaluations are conducted to compare the perceptual impact of AI-generated images with mathematical metrics. Our findings highlight the varying degrees of perceived realism among different image generative With AI models, DALL-E and Imagen generally being perceived as more realistic than Stable Diffusion and GROK. As examined, human evaluation is the current gold standard in text-to-image evaluation; however, mathematical based metrics also have promise and value. Our findings contribute to the advancement of text-to-image synthesis and our results provide support for FID as the gold standard evaluation method as it most closely represented human evaluation. [ABSTRACT FROM AUTHOR]

Increases in digitalization have made it possible to track and measure every click, every payment, every message, and almost everyone's daily thoughts. Companies are extremely interested in the robustness of this data, specifically regarding understanding the sentiment of consumers. Yet the amount of information being produced and processed is quite staggering causing information overload. As such, companies tend to fall into analysis paralysis which can result in missing important insights that could help their business. The goal of this study is to analyze and categorize the top posts on multiple hacking subreddits to determine the most discussed topics and to examine the sentiment of these posts expressed by the users. We began by scraping data, specifically the title, ID, score, comments, and URL for each top post from multiple hacking subreddit communities. We then used the Natural Language Toolkit (NLTK) to perform the data preprocessing techniques for an effective analytic process and bias-free results. The results of the testing allowed us to filter through the posts and determine whether sentiment was positive, negative, or neutral. In the case of the hacking subreddits, many of the posts were of a neutral opinion. This study aims to provide a contribution by utilizing Natural Language Processing methods Topic Modeling such as Term Frequency Inverse Document Frequency, Latent Semantic Analysis (LSA) algorithm, and Sentiment Analysis to gather and synthesize cybersecurity data. [ABSTRACT FROM AUTHOR]

Facial Expression analysis (FEA) is a process that involves recognition and understanding of human emotions based on facial cues. While FEA has potential applications in various field, this can also be misused, leading to the spread of misinformation through deepfake technology. This research aims to evaluate the effectiveness of facial expressions in distinguishing between deepfake and genuine videos, addressing the gap in how well FEA can identify manipulated contents. To address this issue, a research experiment was conducted to gain an insight into how people react towards deepfake and authentic contents. Respondents were shown videos and an analysis was conducted on participant's facial expressions as well as assessing their knowledge of deepfake detection. A survey was designed to test their confidence with the level of deepfake and authentic video identification, trust, security, and attitude towards them. Facial expressions were analyzed using Noldus FaceReader 7 to detect and classify 7 facial expressions (such as happy, sad, neutral, angry, surprised, disgusted, and other). The study findings indicate that FaceReader analysis discerns a statistically significant difference in emotional responses between real and deepfake videos, while participants reported a higher percentage of neutrality (70% vs. 62.5%) in real videos compared to deepfakes. [ABSTRACT FROM AUTHOR]

Spreadsheets have long played a key role in business decisions and operations. The use and adoption of data analytics has substantially increased over the last few years and amplified this role. Spreadsheets are often a first tool for data analytics as such applications provide ease of calculation with basic statistics and chart development. For much of the last two decades universities have provided training in Microsoft Excel because that was what companies used and demanded. Since mid-2020, there has been an increase in use of Google Sheets causing some faculty to believe that MS Excel should be replaced. Faculty should always be aware of current and future employer demands and ensure programs meet the expectations of both employers and recent graduates. This study reviews business job postings seeking employees with two or fewer years of work experience between 2019 and 2021 and examines demand for spreadsheet application experience. Results overwhelmingly indicate that Microsoft Excel still is the most required spreadsheet application by employers. Faculty should pause before changing MS Excel training or removing certifications. [ABSTRACT FROM AUTHOR]

As a result of the health pandemic, the United States (U.S.) has experienced a labor shortage and a decrease in higher education retention and enrollment which has many educational institutions in a crisis mode as they rapidly search for guaranteed sustainable long-term student enrollment. Numerous research studies have explored students' viewpoints through surveys or focus groups to investigate their preferences regarding online or in-person courses. However, there has not been a research study that has examined the post pandemic registration process of student enrollment data over a traditional scheduling period to determine their unbiased preference in modality offerings. More specifically, when students are given a choice to enroll in an online or face-to-face course, which modality do they independently choose? Student enrollment data collected from a publicly accessible website starting from the first day for priority students and continuing throughout the entire registration period for business and technology courses. Results revealed a significant difference between student enrollment for online versus face-to-face courses. [ABSTRACT FROM AUTHOR]

Natural language Processing is a discipline under artificial intelligence that involves interaction between human language and computer systems. It involves analyzing and representation of natural language, the ability to comprehend both text and spoken words. Natural language processing has evolved to the extent of having the ability to give useful responses to human beings. Large language models have been making landmark advances with new more efficient algorithms and improved hardware and processing power. Models like Google's BERT power predictive text in search predictions. Recently, companies have been training models on billions of parameters, a task that was not feasible just a few short years ago. OpenAI is the market leader in this technology; however, competitors have emerged. This research project aims to investigate perceptions of OpenAI versus an emerging competitor, AI21 on the ability to answer questions and predict text. We developed two web applications that allowed users to key in any questions or text in the textbox, the web application will then answer the user query as a response. The web applications were embedded with Jurassic-1 language model API and the GPT-3 language model APIs. Subjects asked the AI systems questions and rated their perceptions of the results. Furthermore, we investigate perceived privacy of AI systems via a post survey. [ABSTRACT FROM AUTHOR]

Most people interact with websites expecting them to perform quick results and provide quick responses to their requests and many do not realize the performance is due to server side or backend programming. There are several types of backend web framework/scripting technologies. Programmers and developers often debate over which is the technologies is the better solution. Most debates are based on various dimensions such as performance, scalability, and architecture. The most common factor for settling the debate or choosing the most appropriate frameworks tends to be the performance dimension. This study assesses the performance of both Node.js and PHP by implementing well-known algorithms of binary, bubble, and quick sort along with Heap’s algorithm for permutations. These algorithms were selected for their increasing time complexities which allows us to observe the performance differences between the backend framework/scripting. By comparing the performance of these two backend scripting technologies, one can gain a better understanding of the circumstances when migrating from PHP to Node.js would be beneficial. Our results showed that a significant difference occurs in the performance of PHP and Node.js and specifically, Node.js outperformed PHP in terms of latency and other performance metrics. This study provides valuable information for software engineers, developers, and managers who are seeking the best framework for their web applications. [ABSTRACT FROM AUTHOR]

Advances in artificial intelligence (AI), technology integration, and cloud computing, has resulted in an increase of cybersecurity attacks by botnets over the last few years. Attackers use botnets to overwhelm and compromise networks with a goal of disrupting services or operations, stealing credentials, gaining unauthorized access to critical systems, or to obtain information for theft or ransom. The rise in this AI technology has made the job of protecting networks more challenging for network security analysis and professionals. The migration of companies and organizations into the chaotic cloud environment has really given new power to the botnets that is visualized best by scenes in any of matrix movies. One of the best methods of protection of any network or resource is early detection, which can prevent a network from being compromised or minimizing damage to the network. Two modern tools used in network security are Intrusion Detection Systems (IDS), and Security Incident and Event Management (SEIM) systems. This study proposes and tests a modern architecture to detect Botnet traffic through the implementation of modern security devices to defend against a configured local Botnet in a virtual cloud environment. Our model was successful in detecting and preventing botnet attacks. The model also allowed for the attack data to be stored and classified for report generation. [ABSTRACT FROM AUTHOR]