Authors - Tushar Kulkarni, Pradyumna Khadilkar, Behara Roshan Kumar, Rupesh Jaiswal Abstract - In the modern era of rapid Internet access, it is essential to safeguard electronic devices and the sensitive data they contain from cybercriminals, who constantly find new ways to exploit users by seeking holes in the system and manipulating them. Globally, the average cost of a data breach in 2024 is expected to exceed $4 million, an increase of ten percent from the previous year. A better intrusion detection system capable of handling both legacy and zero-day threats is needed. To address this, we reviewed several relevant publications, most of which were published in the last five years. This has enabled us to compile the latest techniques and breakthroughs. Although NSL-KDD, CICIDS-2017, and UNSW-NB15 received the most attention, other datasets were included. The performance of various intrusion detection systems has been compared in the literature that has been cited. We have proposed a novel methodology known as hybrid intrusion detection system (Hybrid-IDS), which combines both signature- and anomaly-based IDS.
Authors - Vaishali Rajput, Swapnil Patil, Diksha Shingne, Uzair Tajmat, Yashodip Undre, Urja Wagh Abstract - With the rapid advancements in artificial intelligence (AI) and machine learning (ML) within the Ed-Tech sector, our project explores the development of an Adaptive Learning Platform aimed at enhancing student comprehension and engagement. The platform is specifically designed to cater to diverse learning styles by delivering personalized learning experiences that adapt to each individual’s pace, needs, and preferences. Leveraging AI-driven features, the platform ensures that students achieve mastery of core concepts. By identifying patterns in user interactions, such as repeated engagement with specific video segments or prolonged pauses, the system recognizes areas where learners may struggle and responds with customized explanations powered by GPT technology to clarify complex concepts in text form. Upon successful course completion, the system generates certificates as formal recognition of the user's mastery, providing tangible proof of their progress and achievement. This innovative approach not only addresses the inherent limitations in traditional e-learning platforms by fostering a dynamic and responsive learning environment but also sets a new benchmark in personalized education technology. By incorporating cutting-edge AI and ML, this platform promises to significantly enhance student learning efficiency, knowledge retention, and overall academic performance, offering a transformative shift in how education is delivered and experienced.
Authors - Sara Umalkar, Aditya Toshniwal, Varad Vanga, Siddhesh Upasani, Vedant Joshi, Sanchit Joshi Abstract - This paper surveys the integration of an Online Judge System (OJS) within a Learning Management System (LMS) for engineering education, focusing on its role in enhancing programming skills and student engagement. The OJS automates the evaluation of coding assignments, providing real-time feedback, scalability and secure code execution through containerization and sandboxing techniques. The LMS, equipped with interactive modules and gamification, uses the OJS to assess algorithmic solutions against predefined test cases, supporting both learning and competitive programming. The study explores key features such as automated grading, performance evaluation and robust security measures to ensure fairness. By analyzing existing OJS platforms and their applications, this paper highlights the effectiveness of such systems in fostering problem-solving skills and preparing students for industry challenges. The survey also identifies emerging trends and opportunities for improving the OJS in educational settings.
Authors - Ruchita Borikar, Sakshi Thorat, A.S. Ingole, U.A. Kandare Abstract - Credit card fraud remains a significant issue in the financial industry, with increasing numbers of transactions being processed online and through digital platforms. Traditional fraud detection systems, relying on predefined rules and manual analysis, are no longer adequate to combat the growing complexity and scale of fraudulent activities. In response, machine learning (ML) has emerged as an effective solution for detecting fraud in real time, offering the ability to analyze large datasets and recognize patterns that may indicate suspicious behavior. This project aims to build a system for fraud using machine learning methods, based on a dataset sourced from Kaggle and guided by an IEEE paper. The project involves several key stages, starting from raw data preprocessing and feature selection, followed by training and evaluating machine learning models with algorithms like Decision Trees, Logistic Regression and Random Forest and Support Vector Machines (SVM). Additionally, the model is evaluated through 5-fold cross validation to ensure robustness. This system not only enhances fraud detection accuracy but also minimizes false positives, thereby improving overall efficiency.
Authors - Md Imran Alam, Swagatika Sahoo, Angshuman Jana Abstract - Wildlife is a crucial part of our environment that must be protected to preserve various animal species and their habitats. Effective wildlife conservation faces numerous challenges, including securing funding, preventing illegal wildlife trade, protecting animals from unnecessary harm, ensuring proper treatment, and managing shelters for domesticated animals. Current solutions to these challenges are often centralized, making them vulnerable to corruption and single points of failure. The adoption of blockchain technology to wildlife protection provides secure tracking of animal medical records, transparent management of funding transactions, and real-time alerts to authorities regarding potential threats, all automated through smart contracts. This paper explores the theoretical and practical benefits of integrating blockchain into animal welfare systems. By addressing key issues like illegal trade and medical record management, blockchain technology can significantly improve the security and effectiveness of wildlife conservation efforts.
Authors - Vaishali S. Pawar, Mukesh A. Zaveri, Radhika P. Chandwadkar, Varsha H. Patil Abstract - Pattern recognition involves identifying specific patterns or features within the provided data. Social Network Analysis, Fraud Detection, Biological and Medical Networks, Recommendation Systems, Telecommunication Networks, Traffic and Transportation Networks, Computer Vision and Image Processing, Natural Language Processing (NLP) are among the constantly expanding applications of pattern recognition. Graphs are a potent model utilized in several fields of computer science and technology. This study presents a technique based on graph databases for graphical symbol identification. The suggested method employs graph-based clustering of the graph database, which markedly decreases the computational complexity of graph matching. The suggested algorithm is assessed with a substantial quantity of input hand drawn images, and the output results indicate that it surpasses previous algorithms.
Authors - Aatish Aher, Chinmay Sanjay Lonkar, Ganak Bangard, Sanjay T. Gandhe Abstract - Bluetooth Low Energy (BLE) technology is increasingly recognized as an effective indoor navigation solution, particularly in areas where GPS cannot function. This review focuses on BLE-based systems tailored for visually impaired users, examining their advantages and limitations. It compares various indoor localization methods such as Wi-Fi, BLE, and image processing in terms of accuracy, efficiency, and deployment simplicity. BLE beacons are highlighted for their low power consumption, adaptability, and cost-effectiveness, making them suitable for large-scale, real-time navigation. To enhance accessibility, the review discusses integrating assistive features like audio navigation using BLE signals paired with voice assistants, ensuring hands-free operation. Additionally, it covers technologies such as haptic feedback and obstacle detection, which provide non-verbal cues to alert users about obstacles or landmarks. The system architecture is explored, focusing on app integration, user-friendly interfaces for visually impaired users, and cloud management for delivering real-time updates. By identifying research gaps, this review suggests directions for future development of BLE navigation systems aimed at enhancing the independence and mobility of visually impaired users in indoor settings.
Authors - Nivedita Shimbre, Prema sahane, Shrutika Amzire, Arya Ganorkar, Pavan Kulkarni, Pawan Bondre Abstract - As INDIA is on the verge of getting digitalised India, farmers also need to become smart. They need a system that recommends them crops and its varieties and proper scheduling methods for increasing their crop yield. To obtain the best crop varieties for a given area factors such as soil moisture, fertility and climatic condition are required. And our system will suggest crops and its varieties on the basis of previous History of crop production and time required for the crop. This system also focuses on recommending suitable fertilizers for crops. It also provides the schedule and quantity of fertilizer to be used. This system will use various Machine learning Algorithms like Support Vector Machine, Random Forest Regressor, Gradient Boosting, Linear Regression etc to recommend best crops, its varieties and fertilizers. It also provides Automatic Irrigation Management by using IoT.
Authors - Jay Bodra, Anshuman Prajapati, Priyanka Patel Abstract - India is one of the leading agricultural countries in the world, and the nation's economy depends heavily on agriculture. For good crop yield, prediction of precipitation is necessary to increase agricultural output and ensure a supply of food and water to maintain public health. To reduce the issue of drought and floods occurring in the nation, wise use of rainfall water should be planned for and implemented. Numerous studies have been carried out utilizing data mining and machine learning approaches on environmental datasets from various nations in order to forecast rainfall. This study's primary goal is to pinpoint the amount of rainfall in several regions of India in the past hundred years and apply machine learning techniques to forecast the amount of rain that will fall in a particular month and year in a given region. The dataset was collected from the government site of the rainfall database for performing machine learning techniques. The Random Forest model's ensemble approach, robustness to noise, ability to handle nonlinear relationships, feature importance analysis, scalability, and tuning flexibility make it a particularly effective choice for rainfall prediction in this project. Its versatility and performance make it a valuable asset for providing accurate and reliable rainfall forecasts to support decision-making in various sectors, such as agriculture, water resource management, and disaster preparedness.
Authors - Mangesh Salunke, Tilak Shah, Vishal Bhokre, R. Sreemathy Abstract - Chatbot also known as conversational agents is an interactive software that responds to users’ queries using artificial intelligence. While traditional machine-learning chatbots have shown promise, LLM-powered chatbots offer more natural and relevant conversations, enhancing the user experience. The rise of OpenAI’s ChatGPT, Google’s Gemini, LangChain, etc has widened the horizon of applications of chatbots to almost every sector including education, healthcare, banking, entertainment, e-commerce and telecommunications. The main objective of this comprehensive study is to explore and analyze the current advancements in chatbot development using different artificial techniques. This survey paper examines key trends in the development of chatbots, the components and techniques used, and the evaluation metrics employed to measure performance. We discuss different metrics used to evaluate chatbots' performance like accuracy, BLEU, ROUGE and relevancy. The results suggest that LLM-powered chatbots facilitate more natural and contextually appropriate conversations than traditional machine-learning models, leading to a marked enhancement in user experience. By synthesizing insights from existing research, we aim to provide a comprehensive understanding of RAG-based chatbot technology.
