1,721,170 research outputs found
Using Machine Learning to Predict Stroke
Get PDFStroke is a leading cause of death in the United States and is a major cause of serious disability for adults. According to the World Health Organization (WHO), stroke is the 2nd leading cause of death globally, responsible for approximately 11% of total deaths. About 800,000 people in the US die yearly; about three in four are first-time strokes. Strokes are also the leading cause of long-term disability and the leading preventable cause of disability. The prediction of stroke in advance will help reduce the death rate. In recent years, predicting stroke in the real-life medical area has not been easy. A massive amount of healthcare data was collected for analysis. This paper proposes predicting stroke using five machine learning algorithms: Logistics Regression, Random Forest Classifier, Gradient Boosting Classifier, Decision Tree Classifier, and Support Vector Machine. The dataset we are using is from Kaggle. There are 5110 records and 12 columns. It includes one target and 11 features such as gender, age, hypertension, etc. Since the dataset is imbalanced, we applied SMOTETomek and SVM-SMOTE for oversampling. The study compares the results of five machine learning models before and after using SMOTETomek and SVM-SMOTE. It also compares the accuracy of these five different machine-learning techniques. In the research, we find that Logistic Regression and SVC perform better than others before oversampling and obtain the same recall and testing accuracy, which is 0.94 and 93.93%, respectively. After applying SMOTETomek, Logistic Regression and SVC do not perform as well as before. Gradient Boosting has the best performance. The testing accuracy is 88.85%. After applying SVMSMOTE, overall results for all models are improved compared to those using SMOTETomek. Gradient Boosting has the best performance. The testing accuracy is 91.10%, which is higher than the prior study on the Kaggle. Keywords Stroke, Machine learning, Logistics Regression, Random Forest Classifier, Gradient Boosting Classifier, Decision Tree Classifier, and Support Vector Machin
Using Machine Learning to Predict Stroke
Get PDFStroke is a leading cause of death in the United States and is a major cause of serious disability for adults. According to the World Health Organization (WHO), stroke is the 2nd leading cause of death globally, responsible for approximately 11% of total deaths. About 800,000 people in the US die yearly; about three in four are first-time strokes. Strokes are also the leading cause of long-term disability and the leading preventable cause of disability. The prediction of stroke in advance will help reduce the death rate. In recent years, predicting stroke in the real-life medical area has not been easy. A massive amount of healthcare data was collected for analysis. This paper proposes predicting stroke using five machine learning algorithms: Logistics Regression, Random Forest Classifier, Gradient Boosting Classifier, Decision Tree Classifier, and Support Vector Machine. The dataset we are using is from Kaggle. There are 5110 records and 12 columns. It includes one target and 11 features such as gender, age, hypertension, etc. Since the dataset is imbalanced, we applied SMOTETomek and SVM-SMOTE for oversampling. The study compares the results of five machine learning models before and after using SMOTETomek and SVM-SMOTE. It also compares the accuracy of these five different machine-learning techniques. In the research, we find that Logistic Regression and SVC perform better than others before oversampling and obtain the same recall and testing accuracy, which is 0.94 and 93.93%, respectively. After applying SMOTETomek, Logistic Regression and SVC do not perform as well as before. Gradient Boosting has the best performance. The testing accuracy is 88.85%. After applying SVMSMOTE, overall results for all models are improved compared to those using SMOTETomek. Gradient Boosting has the best performance. The testing accuracy is 91.10%, which is higher than the prior study on the Kaggle.Keywords Stroke, Machine learning, Logistics Regression, Random Forest Classifier, Gradient Boosting Classifier, Decision Tree Classifier, and Support Vector Machin
An Empirical Study to Comprehend the Capabilities of AI Chatbots in Detecting Security Code Smells
No full textAs software becomes increasingly complex, potential issues in code structure and design that may cause security vulnerabilities arise - known as security code smells. This research explores how viable AI chatbots are in identifying six diferent security code smells when fed problematic code snippets. A series of experiments were conducted comparing the performance of ChatGPT, Google Gemini, Meta LLaMa, Anthropic Claude, and Hangzhou DeepSeek when attempting to detect Bad Casts, Bufer Overfows, Hard-coded Secrets, Smelly Functions, Weak Cryptography, and Wrap-around Errors, as well as a check on which security code smells were found to have been the most difcult to detect. The fndings reveal that AI chatbots can efectively recognize these security code smells, with Meta LLaMa, Anthropic Claude, and DeepSeek performing at a 93% detection rate, ChatGPT at a 91% detection rate, and Google Gemini performing at a 73% detection rate. In addition, the "Smelly Function" security code smell yielded the lowest detection rating relative to other smells among the LLMs used, with DeepSeek performing well at a 94% detection rate, LLaMa and Claude having an 81% detection rate, ChatGPT having a 75% detection rate, and Gemini with a 37% detection rate. This paper discusses the implications of using AI chatbots for security code smell detection and how viable they may be as a tool in aiding software engineers
An Interview Study of Social Media Scams
Get PDFSocial media scams are escalating rapidly. The Federal Trade Commission (FTC) reports a staggering $8.8 billion loss in 2022 due to these scams, marking a 30% increase from the previous year. Our research delves into this escalating fraud issue particularly focusing on social media scams that are becoming increasingly sophisticated and widespread. These scams encompass various forms, such as cryptocurrency investment scams, fraudulent job offers, phishing attacks, and more, posing significant risks to users and platforms alike. They lead to financial losses, compromised personal information, and emotional distress for victims. To understand these scams better, we interviewed 30 individuals who had experienced social media scams. Our study methodically examines the stages of a scam's lifecycle: planning, launching, cashing out, ending, and making advice and recommendations. This thorough approach provides a deep insight into the strategies of scammers, their interactions with victims, the execution of fraudulent transactions, and the ensuing consequences. Our findings reveal that many victims were targeted through platforms like LinkedIn, Twitter, and other social networks, indicating that professional networking sites are increasingly becoming hotspots for scams. Scammers often initiate contact by presenting enticing opportunities or expressing interest in the victim's profile. Additionally, we evaluate the security behavior, security attitudes and privacy concerns of social media fraud victims and compare with US general population. The objective is to gain insights into how people navigate the realm of security in light of the growing prevalence of scams and threats on the internet. Our study underscores the necessity of raising awareness and educating the public on recognizing the signs of scams as preventative measures
Saudi Arabian Perspective of Privacy Concerns, Attitudes, and Behaviors on Facial Recognition Technology.
Get PDFFacial Recognition Technology (FRT) is a pioneering field of mass surveillance that sparks privacy concerns and is considered a growing threat in the modern world. FRT has been widely adopted in the Kingdom of Saudi Arabia to improve public services and surveillance. Accordingly, the following study aims to understand the privacy and security concerns, trust, and acceptance of FRT in Saudi Arabia. Validated Privacy Concerns (IUIPC-8), Security Attitudes (SA-6), and Security Behavior (SeBIS) scales are used along with replicate studies from Pew Research Center trust questions and government trust questions. In addition, we examine potential differences between Saudis and Americans. To gain insights into these concerns, we conducted an online survey involving 53 Saudi participants. Responses from closed-ended questions revealed that Saudis score higher than Americans when it comes to security attitudes, whereas they score lower when it comes to privacy concerns. We found no significant difference between Saudis' and Americans' acceptance of the use of FRT in different scenarios, but we found that Saudis trust advertisers more than Americans. Additionally, Saudis are more likely than Americans to agree that the government should strictly limit the use of FRT
Improved Glioma Grading using Deep Learning Techniques
Get PDFBACKGROUND AND PURPOSE Gliomas are a sort of primary brain tumor that originates from glial cells. These cells are accountable for supporting the central nervous system of the human body. These tumors vary widely based on their nature, they could be either malignant or aggressive. Glioma grading is an essential part of diagnosis and treatment planning, as it can give critical information regarding the tumor's characteristics and its behavior. The main objective of this study is to use MR images and develop an attention-based deep learning model to detect glioma and visually locate the tumor. Additionally, machine learning models are deployed on the radiomic features, which are extracted from the MRI scans. MATERIALS AND METHODS: The dataset includes a total of 285 patients with both high- and low-grade gliomas. The preprocessing steps applied on this dataset include interpolation to a standardized resolution of 1 cubic millimeter, alignment to a common anatomical template, and finally skull stripped. A pretrained VGG16 model with 2 attention modules is used for grade prediction. These attention modules extract intermediate features from the main architecture and predict the area of tumor by highlighting it. The proposed model's performance is evaluated with other pretrained models like ResNet, DenseNet, MobileNet, and EfficientNet. Along with the deep learning models, popular machine learning models are also used to evaluate the performance. RESULTS The proposed model was able to achieve an f1-score of 91.18%, demonstrating its robustness and capability to grade the tumor. Furthermore, the attention maps enabled detailed visualization of the tumor regions, enhancing the interpretability of the model's predictions. The proposed model produced almost the same results as the pretrained model ResNet50 but with additional visualization of the tumor. In comparison to the machine learning model, we can notice an improvement of about 6% in F1-score, from the top-performing machine learning model, i.e. random forest. CONCLUSION: To conclude, the proposed model, leveraging its fundamental capability to automatically learn features, has proven remarkable effectiveness in glioma tumor detection and its classification. The integration of two attention maps into the VGG16 pretrained model has enhanced its capability of precisely focusing and detecting the tumor region, a feature that was not available in earlier models. This holds a promising advancement in disease diagnosis and medical imaging
Commercial Anti-Smishing Tools and Their Comparative Effectiveness Against Modern Threats
Get PDFSmishing, also known as SMS phishing, is a type of fraudulent communication in which an attacker disguises SMS communications to deceive a target into providing their sensitive data. Smishing attacks use a variety of tactics; however, they have a similar goal of stealing money or personally identifying information (PII) from a victim. In response to these attacks, a wide variety of anti-smishing tools have been developed to block or filter these communications. Despite this, the number of phishing attacks continue to rise. In this paper, we developed a test bed for measuring the effectiveness of popular anti-smishing tools against fresh smishing attacks. To collect fresh smishing data, we introduce Smishtank.com, a collaborative online resource for reporting and collecting smishing data sets. The SMS messages were validated by a security expert and an in-depth qualitative analysis was performed on the collected messages to provide further insights. To compare tool effectiveness, we experimented with 20 smishing and benign messages across 3 key segments of the SMS messaging delivery ecosystem. Our results revealed significant room for improvement in all 3 areas against our smishing set. Most anti-phishing apps and bulk messaging services didn't filter smishing messages beyond the carrier blocking. The 2 apps that blocked the most smish also blocked 85- 100% of benign messages. Finally, while carriers did not block any benign messages, they were only able to reach a 25-35% blocking rate for smishing messages. Our work provides insights into the performance of anti-smishing tools and the roles they play in the message blocking process. This paper would enable the research community and industry to be better informed on the current state of anti-smishing technology on the SMS platform
Smishing Awareness from Brands: Insights from a Content Analysis Study.
Get PDFPhishing attacks through text, also known as smishing, are a type of social engineering tactic in which an attacker uses deception and impersonation to fool a victim into providing personal information and/or money. Attackers send a malicious link to trick victims into providing information, making victims think that the attacker represents an actual brand or person. A major cause of people falling prey to these attacks is the lack of smishing awareness and cyber education. Through this study, we aim to discover how different popular brands educate their customers about smishing and what smishing prevention and reporting advice the brands provide. After analyzing the websites of 149 brands, we identified major gaps in the smishing-related information the brands provide, including prevention and reporting guidelines. Our analysis reveals that only 46% of the 149 brands we explored mentioned the definition of smishing, and less than 1% had a video tutorial on smishing. Furthermore, we found that only 50% of the 149 selected brands provided instructions on how to report a smishing scam, with a few brands mentioning "ignoring the message" as one of the guidelines. In this paper, we provide recommendations for brands on how to offer streamlined education to their respective customers on smishing for better awareness and protection against increasing smishing attacks
The augmentation of heat transfer in a pipe flow using a swirling perforated twisted (SPT) tape insert
No full text
- …
