Daily Archives: April 6, 2025

Abstract-The development of new methodologies for identifying phishing attacks in the context of an increasing digital world is compromised due to lacking research and execution. This paper focuses on versatile approaches in Artificial Intelligence (AI) and Machine Learning (ML) to almost single-handedly eliminate phishing attempts. The work takes a holistic approach to problems of URL structure, content, and behaviors by XGBoost, LightGBM, Naïve Bayes, and CatBoost, as well as Graph Neural Network GNN. Multiple features are captured; for example, URL length, number of dots, slashes, numeric characters, and special characters will all be used for model training. Monitoring the system in real time and adapting it to new phishing paradigms makes it possible to tactically protect users and organizations from the continuous, unpredictable changes of cyber threats. This study covers the approach of employing diverse machine learning methods to combat phishing in a more direct and secure manner.

Abstract-College courses are haunted by participation and grade problems, particularly in large or online classes. This experiment investigates the impact of an adaptive learning (AL) system, CogBooks®, on student achievement in a blended statistics course. We employed a quasi-experimental experiment with two groups: an AL system (N=100) group and a group of students instructed using traditional techniques (N=100). Partic- ipation was measured with validated surveys, and performance was measured with standardized grades. Results indicated a statistically significant difference in the AL group, a 15 percent boost in average grades (p < 0.05) and a 20 percent boost in reported participation measures compared to the control group. The system’s real-time feedback and individually tailored learning paths efficiently addressed individual students’ needs. Surprisingly, our solution is offline-capable, offering accessibility in low-bandwidth settings—a major benefit compared to cloud- capable solutions. These findings present AL as a cost-effective, scalable solution to higher education, with potential applications to STEM and humanities courses. Long-term retention effects and compatibility with generative AI tools might be topics for future research.

Abstract-With the increasing reliance on technology for location-based services, finding an optimal and fair meetup spot for a group remains a challenge. This research proposes a machine learning-based approach to recommend the most suitable meetup location based on the geographic inputs of multiple users. The system utilizes clustering algorithms to identify equidistant locations while incorporating user preferences such as venue type (e.g., cafes, parks, malls). Google Maps API is leveraged for real-time location data and distance calculations, while various machine learning models, including K-Means and DBSCAN, are compared for efficiency and accuracy. The system enhances decision-making by offering optimized suggestions, ensuring fairness and accessibility for all participants. Future improvements include incorporating real-time traffic data and personalized recommendations based on user behavior.

Abstract-Agriculture is on the brink of a revolutionary change during the era of advanced technology meets age-old industries. This article give a Advanced tools for plant disease detection at an early stage is in demand for sustainable agriculture. Traditional methods, which rely on visual inspection, are slow and error- prone. This paper talks about the use of artificial intelligence and image processing for the accurate and effective diagnosis of plant diseases. Although quite promising, recent attempts suffer from small datasets as well as lack of generalization. As we are working with deep models, we have also significantly emphasized the diversity between the views in the dataset which not only makes our work more accurate but also can scale better than previous work. Hence, this research aids sustainability in agriculture with strong, automated solution for disease detection.

Abstract-Modern technology is changing the healthcare land- scape,and perhaps no greater impact is being made in the diagno- sis and prediction of heart disease. Inthis research work, machine learning (ML) models are applied to predict the probability of heart disease for the individual patient based on personal- related features such as age, blood pressure, cholesterol levels, and life experience information. Although several studies have implemented ML techniques, there are still challenges in limited datasets, accuracy, andinterpretability of the models used. The proposed system is intuitive as compared to otherswhere health information from the users is entered and risk is returned. If the model predicts a high risk, it recommends that the individ- ual seea health care provider. Night trip focuses on accessibility with this tool available to the generalpublic as well as medical doctors, uniting personal health with professional diagnostics. The system enhances prediction reliability responsively toheart disease prevention by harnessing the combined strengths of multiple models.

Abstract-Donate Management System Food and Clothes using AI In India, traditional donation systems lack transparency and often mismatch resources and poor allocation. India has millions of people who struggle to make ends meet with food and cloth donations to help feed them. Rural areas and people who are affected by natural disasters are also, in documents, in need of food and clothes. Using artificial intelligence, this system improves decision-making as it analyzes donation patterns and demand and uses real-time data to link donors and recipients. We are also leveraging AI algorithms to prioritize critical needs, reduce waste, and ensure equitable distribution to communities in need. The system has features such as registering the donor and recipient, automatic allocation. One cannot ignore the extreme disparity in wealth distribution in India where we have so many NGOs, government schemes and volunteer organizations and they are selflessly working to provide food and clothing to the needy when they need it, but still the effort is getting lost due to logistical inefficiencies, resource wastage and lack of real-time data.

Abstract-Electric vehicle information security improves through the implementation of AI & ML in automotive sector. A targeted study about data-driven implementations of AI and ML within electric vehicles represents a necessary research need. This paper describes the present situation regarding AI and ML implementations in EVs. Comprehensive analysis of pertinent studies and articles enabled our team to find important content after analyzing how different subjects tie together within these documents. This recent research has shown that AI along with ML technologies increasing appear as critical tools in protecting EV information security through improved authentication sys- tems and better attack detection methods. Table I shows different implementation applications of Machine learning techniques, encompassing deep learning and neural network models while blockchain technology shows increasing applications. The Study data shows the man in middle intrusion detection receives the highest attention at 75% while authentication covers 20% of the literature and prevention takes up only 5%. Deep learning consti- tutes 70% of analyzed works with neural networks representing 15% and remaining studies employing alternative methods..

Abstract-This paper adheres to web accessibility through a privacy-centric, AI-powered approach via an extension in Chrome. The extension implements a multi-model architecture that combines Google Chrome’s built-in AI capabilities using Gemini Nano with a JavaScript library transformer.js to process machine learning (ML) models directly in the browser that is web content run locally on users’ devices. Unlike existing solutions that rely on cloud processing or limited built-in browser features which might hinder the user’s privacy, our extension prioritizes user privacy by performing the computational/processing tasks on-device while providing comprehensive accessibility features. System also has voice commands for hands free navigation, generates summaries based on prompts and utilizes moondream(AI model) to provide detailed descriptions of images present in the web-content. Performance metrics indicate that the local processing approach maintains robust functionality while preserving user privacy.Our user testing shows remarkable improvements in web browsing for people with diverse accessibility needs. Users reported faster navigation, better understanding of content, and greater independence compared to traditional screen readers and similar tools. Our approach of processing information locally on users’ devices maintains strong performance while protecting privacy. This research advances accessible technology by showing how AI models can be integrated into browser extensions to make the web more inclusive without compromising privacy or requiring powerful computers.

Abstract-Brain tumour disease arises from the uncontrolled growth of cells. Detecting brain tumours early is crucial for successful treatment. Many current diagnostic methods are cumbersome, demand significant manual input, and yield less-than-ideal results. The EfficientNet-B0 architecture was utilized to diagnose brain tumours through magnetic resonance imaging (MRI). This refined architecture was applied to classify four distinct stages of brain tumours from MRI images. The fine-tuned model achieved 99% accuracy in identifying four different brain tumour classes: glioma, no tumour, meningioma, and pituitary. The proposed model excelled in detecting the pituitary class, with a precision of 0.95, recall of 0.98, and an F1 score of 0.96. It also performed exceptionally well in identifying the no-tumour class, with precision, recall, and F1 score values of 0.99, 0.90, and 0.94, respectively. The precision, recall, and F1 scores for the Glioma and Meningioma classes were also notably high. This proposed solution holds significant potential for improving clinical assessments of brain tumours.

Abstract-An advanced career guidance recommendation sys- tem is an innovative algorithm that differentiates itself from traditional recommendation algorithms, which work through a link to provide relevant recommendations and are heavily dependent on user behavior. These algorithms are trained to understand various aspects relating to an applicant—their skills, qualifications, work experience, and preferences—and utilize these data points to create customized recommendations based on their profile and career aspirations. Additionally, these intelligent talent recommendation systems continuously evolve, refining their suggestions by analyzing user feedback and performance insights. This study introduces an innovative approach to talent rec- ommendation specifically designed for college students who are preparing to enter the emerging job sector based on a Rank- Based Sequential Deep Learning (RBS-DL) Model. By studying students’ skills, qualifications, interests, and career aspirations, the algorithm intends to offer personalized recommendations. The effectiveness of the recommendation algorithm enhanced by RBS-DL is assessed through both simulated experiments and empirical verifications. Results show significant improvement in recommendation accuracy and relevance over traditional approaches. For the RBS- DL algorithm, students showed a higher job offer acceptance rate by 40% and around 30% increment in job satisfaction levels. This algorithm also learns from user interactions, adjusting its recommendations over time based on real-time user feedback.