Authors - Keerthi AJ, Kalyanasundaram V, Srinivasa Perumal R Abstract - Individuals who are visually impaired or have dual sensory impairments, such as both hearing and vision loss, face significant challenges in navigating public spaces. These challenges often limit their independence and pose risks of unintentional harm to themselves and others. While traditional mobility aids like canes or guide dogs provide some assistance, they lack the ability to deliver real-time, comprehensive awareness of the user's surroundings. To address these limitations, The Intelligent system-powered Smart Device designed to enhance mobility and safety for visually impaired individuals. This device leverages advanced object detection technology to enable users to navigate public spaces more effectively and confidently. The solution employs a SSD MobileNetV3 Convolutional Neural Network (CNN) model for real-time, efficient, and accurate object detection. Integrated with the Ov7670 for computer vision tasks and an Arduino microcontroller for hardware coordination, the system captures live video through a mounted camera to detect and classify obstacles. Users receive instant alerts via auditory or haptic feedback, promoting safer navigation. To ensure robust performance, Azure Custom Vision is used to evaluate and visualize the precision, recall, and average precision (AP) using the COCO dataset. By offering enhanced mobility and reducing risks, this innovative device fosters independence and inclusivity for visually impaired individuals in public environments.
Authors - K.L.Sailaja, Gollapudi Vanditha, Goriparthi Krishna Swapnika, Mohammad Sania Sultana, Madala Pavani Abstract - The aim of this project is to create an effective machine learning model for the detection and forecast of Urban Heat Island (urban heat islands) phenomenon in the mid region of Andhra Pradesh state specifically Vijayawada. High temperatures in urban areas relative to their rural areas are called Urban Heat Islands. The negative impacts include increasing energy use, health risks, and environmental destruction. The satellite imagery and Random Forest models, in particular, have a long-standing reputation of being inaccurate when it comes to geolocalization and even when time-based forecasts are provided, they are mostly misleading. Thus, this gives rise to inaccuracies and inconsistencies in hotspot identification and forecasting’s metrics. This Project suggests an improved Recurrent Neural Network (RNN) model that incorporates Long Short-Term Memory (LSTM) algorithms, driven by the need for more precise and reliable predictions. The proposed LSTM-based model targets the traditional approaches shortcomings of being spatially and temporarily inaccurate in the detection of hotspots. The patterns of temperature, humidity, and soil moisture in city regions can be explained better by this model. It increases the model's predictive capability and explains urban island’s patterns. The project uses data obtained through NASA/POWER CERES/MERRA2 Native Resolution Daily Data, which provides an extensive collection of temperature, humidity, and soil moisture records. These factors will be used to develop forecasting and predictive models of the Urban Heat Islands hotspots. Normalization is one of the methods employed even during advanced data preprocessing.
Authors - Payal Khode, Shailesh Gahane, Arya Kapse, Pankajkumar Anawade, Deepak Sharma Abstract - The technology behind cryptocurrency is secure and transparent. Currently, numerous investors are attracted to cryptocurrencies because of their transparent and safe technology. Additionally, investors find cryptocurrencies fascinating due to their high return potential and innovative possibilities. To optimizes trading and predict prices for investment strategies, some artificial intelligences are required. As of 2024, the global cryptocurrency market capitalization has exceeded 2.5 trillion dollars. Since then, cryptocurrency has established itself in the financial arena, with daily transaction volume reaching enormous heights. Navigating investment strategies are one of the main challenges for investors. There for, leveraging artificial intelligence for optimal investment decisions stands as effective solutions. The study examines recent developments in the field of artificial intelligence methods for cryptocurrencies investment, focusing on trading digital currencies such Bitcoin, Altcoin, Meme coin, and others. Even though price prediction for investing strategies has been the subject of extensive research, notable gaps remain in enhancing cryptocurrency trading through AI for successful investment outcomes. This paper reviews these gaps by examining the role of AI in accurately predicting cryptocurrency prices to enhance optimal investment. The study's findings demonstrate the critical role that precise price forecasting plays in developing adaptable and cautious trading and investing methods, which are essential in the erratic cryptocurrency market. Additionally, the study highlights current issues and suggests future research possibilities, highlighting the importance of ethical issues and multidisciplinary methods in the investment. By filling the knowledge vacuum and providing direction for future study, this synthesis hopes to promote more advanced and successful investment techniques in the crypto space.
Authors - Grishma Bobhate, Pawan Bhaladhare Abstract - Internet of behavior is a primitive approach to study behavior analysis and predictive learning system to understand user experience and interpret their psychological patterns for betterment of the society. Basically, Internet of Things and Internet of Behavior are closely related to each other and can offer different techniques in various areas for developing technology and significant applications. Major psychological disorder, or depression, is a common but fatal neurological condition that has a disruptive impact on feelings, actions, and ways of seeing actuality. Various Machine learning algorithm have been implemented to detect depression through fusion modalities applied on different parameters such as visual, textual and gaze movement. To ensure preventive measure and provide ethical frameworks, this study aims to identify Internet of behavior technology that can have crucial importance in the comprehensive study of depression detection in healthcare sectors. With the assessment in health monitoring systems, the main objective is to explore and analyze the strategies in the Internet of behavior technology for understanding patient behavior and mental health to detect depression and mood. Various challenges towards Internet of behaviors has discussed. In order to ensure the reliability of the system, it also explores the different machine learning and deep learning approaches to determine depression with the performance validation. This will help to assist medical personnel in acquiring details and evaluating the actions of patients for an effective regimen of treatments. This study outlines the strategies to adopt the behavioral analysis for effective learning of depression detection model.
Authors - Balasubbareddy Mallala, Azka Ihtesham Uddin Ahmed, P. Kowstubha, T. Murali Krishna Abstract - The world is now transiting towards Renewable Energy sources (RES) at a rapid pace to overcome the limitation of fossil fuel and generate Green Energy. But due the irregular generation of power in RES (like Solar PV Plant) throughout the day is making it less reliable. This paper integrates RES with an Energy Storage System (ESS) and Fuel Cell to overcome this disadvantage. With the help of this system the dependence on conventional energy sources can be reduced, the cost of generation of power can be brought down to 1/4th compared to an existing traditional system and also increases energy independence. During the morning hours, Combined with the fuel cell, the solar photovoltaic plant will supply power. Any excess power generated will be stored in the energy storage system (ESS). This way, when sunlight is unavailable, the ESS can meet the load demand, ensuring continuity and making the system more efficient and reliable.
Authors - Pradeepkumar G, Priya Devi T, S A Suje, Gobinath S, M Dhanapal Abstract - Nowadays self-driving cars are gaining attraction globally but their implementation in India faces significant hurdles due to the inadequacies of existing approaches reliant on GPS and sensor technologies. The erratic nature of Indian roads, characterized by variable road conditions and inaccuracies in mapping, renders conventional methods unreliable. To address these challenges, propose a novel approach utilizing pattern matching techniques for autonomous navigation. The solution involves deploying specialized patterns on the road surface, facilitating accurate detection and identification of pathways suitable for autonomous driving. By utilizing a modelled car equipped with a Raspberry Pi for image processing, the system captures road imagery via onboard cameras. These images are then transmitted to a remote computer for analysis and subsequent navigation instructions. Additionally, an array of sensors is deployed to detect and avoid obstacles in the vehicle's vicinity. The key innovation lies in the hybrid approach, which combines traditional sensor-based navigation with the novel pattern matching methodology. By leveraging these complementary technologies, the prototype aims to provide robust autonomous navigation tailored to the unique challenges of Indian roads.
Authors - Dhanalakshmi R, Prashaanth S, Hari Prasath S, Dhanaselvam J, Harish R Abstract - India is mainly an agricultural country, where almost three-fourths of the country's population works on farms. Several crops are grown according to regional situations. High-quality production of these crops can be achieved only with new techniques. The appropriate management of crops and identification of diseases and their respective treatments are very significant to prevent losses after harvesting as it usually happens. Diseases in crops deviate from their normal functions and show symptoms that hinder growth. Pests and insects always devastate major crops like rice, wheat, maize, and soyabeans. Consequently, productivity becomes low. With the adoption of deep learning technologies, pest infestation detection and management in agriculture have accuracy and efficiency. A solution is proposed in this paper that integrates image processing techniques with the MATLAB platform for the classification of pests and the proper fertilizers and pesticides to be applied. An autonomous robotic sprayer is used by this system to remotely traverse crop fields, ensuring pinpoint treatment applications. On the other hand, the infrastructure cost is reduced by the proposed solution. The camera setup density in an agricultural IoT monitoring system is minimized by it. Thus, the advanced technology is integrated with agricultural practice by this approach to promote sustainable farming. A validation accuracy of 99.80% is achieved by it to maximize crop production while minimizing losses due to pests and diseases.
Authors - M.Kavitha, N.Revathy Abstract - Forest cover prediction has applications in environmental monitoring, forest management, and land-use planning. Governments and conservation organizations can use it to assess forest cover types and predict land changes. The research article depicts the use of the XGBoost algorithm applicable in forest cover prediction, focusing on evaluating model performance through key metrics like Mean Squared Error (MSE), Logarithmic Loss (Log Loss), and confusion matrices. The XGBoost model, optimized through hyperparameter tuning, demonstrates robust performance with a relatively low MSE, indicating accurate predictions. The Log Loss value of 0.5786 suggests that while the model's classifications are reasonably confident, there is room for refinement. The confusion matrix reveals strong performance for certain classes, such as class 1, but highlights significant errors in others, particularly class 5, which shows a high error rate of 60.93%. The proposed model effectively captures underlying data patterns and performs well across most classes. However, further enhancements, such as addressing class imbalances and refining hyperparameters, are needed to improve accuracy in challenging cases. The model's high hit ratios, where the correct class is often among the top predictions, indicate its reliability in multi-class classification tasks, making it a valuable tool for forest management and environmental monitoring.
Authors - Shailesh Gahane, Payal Khode, Arya Kapse, Deepak Sharma, Pankajkumar Anawade Abstract - Nowadays, many universities and information technology (IT) institutes throughout the world provide online courses, tests, and certificates. In order to administer the tests from any location, delivery technologies have been developed. Putting this into practice will result in time and travel cost savings. Due to the COVID-19 epidemic, there is currently a significant demand for online courses and exams. The majority of universities presently use a variety of assessment methods to evaluate their pupils. These include of oral, paper-based, electronic, and electronic-paper. To help identify the most secure and acceptable assessment method, a survey was carried out. Participants were selected from One Universities in India and Uganda. Population Sample Participants were selected from One Universities in India and Uganda. Using the Krejcie and Morgan formula, a sample of 98 participants was drawn from the 110 research participants. Data was gathered using a questionnaire instrument, and descriptive statistics were generated through data analysis using SPSS software.
Authors - Pradeepkumar G, Pavithramathi R, Jahina J, Tamilselvan K, Arulanantham D Abstract - This work presents an innovative system designed for remote monitoring of landslides using IoT technology. The solution uses a wireless underground sensor network (WUSN), a cloud computing platform and a dedicated mobile application to provide real-time monitoring capabilities. In this system, a sensor network uses Arduino components connected via Wi-Fi modules to collect data on soil moisture levels. This collected data is then transferred to a cloud computing environment for secure and permanent storage. In addition, the cloud platform hosts the model that can trigger alarms when potential landslides are detected. In addition, the system includes a user-friendly mobile application that facilitates real-time data visualization and alerts on potential landslides. This end-to-end solution, from humidity sensor data collection to citizen-facing data visualization, is particularly suitable for smart cities and IoT environments. The effectiveness of the system was evaluated with both real-life tests and simulated scenarios. The results show that the network of sensors accurately measures soil moisture, while the landslide monitoring model continuously sends alerts when necessary.
Authors - Aafiya Anjum Abdul Rafique, Martin H Mollay, Shailesh Gahane, Deepak S. Sharma, Pankajkumar Anawade Abstract - Blockchain technology is a complete makeover for digital identity management, which solves the major problems seen with the previous centralized systems, like inefficiency, lack of control in front of the users, and susceptibility towards data breaches. The paper draws the modes of changing the digital identification system using blockchain technology by providing insights into how it is decentralized, transparent, and safe. Blockchain could improve privacy and trust by advising people to take control of their data and reduce dependency on other parties. The literature analysis shows that blockchain technologies, particularly those based on cryptographic security, give solid answers to privacy problems, which allow users to share data but keep sensitive information safe. Analysis of scalability issues, limitations in storage, and massive computational overhead of consensus protocols like Proof-of-Work. It proposes alternative solutions such as proof-of-stake, sharding, and sidechains that may circumvent their weaknesses. Interoperability between different blockchain systems remains one of the most significant areas of development toward support at a broader scale of adoption. Despite the challenges above, blockchain has immense possibilities in many sectors, including government identity systems, healthcare, and finance, for safe and independent management of identity. It underlines the revolutionary importance of blockchain-based identification solutions within the digital economy. It calls for further research, pilot projects, and regulatory modifications to overcome these issues and realize the full potential of these solutions.
Authors - Sanila S, S Sathyalakshmi, D. Venkata Subrahmanyan Abstract - Electroencephalography (EEG) remains the leading technique for identifying and diagnosing epileptic seizures due to its effectiveness in monitoring brain activity. However, the nonstationary nature, large volume, and rapid accumulation of EEG data present significant challenges for traditional analysis methods. To address these issues, a transition from basic data mining techniques to advanced machine learning and deep learning approaches is essential. This study focuses on developing algorithms to enhance the accuracy of seizure predictions while minimizing the volume of EEG data processed. The proposed method involves dividing EEG signals into fixed-sized windows to reduce data complexity, followed by extracting key features such as the top_k amplitude values from each window. These extracted features, combined with statistical measures of the EEG data, are then used to train classification algorithms to determine whether a seizure is occurring or not. This approach aims to balance efficiency and predictive accuracy, addressing both the computational and diagnostic challenges associated with EEG analysis. The entire raw dataset is experimented with Deep Neural Network Algorithms like Bidirectional Long Short-term memory with additional functionalities like Attention Mechanism and Spatial Weight matrix addition. Finally, both 2 D CNN and BiLSTM are applied in parallel with the additional functionalities of FFT applied EEG signal. The results are promising and found that ANN predicted with an average accuracy of 98.5%, 2DCNN with 94.3% BiLSTM with 99.6% and bi model architecture of top_k 2D CNN with Bi LSTM on FFT applied EEG signal including attention mechanism and Spatial weight matrix predicts better than all previous models with 99.8 % accuracy. .
Authors - Dhairya Goel, Chakshu Gupta, Saarthak Bansal, Chaitali Bhowmik Abstract - Mushrooms are a type of fungi, which have unique traits and health advantages. They also helps to fight against the cancer cells. Our main initiative of this research is to classify the mushrooms into two categories one is poisonous and other is non-poisonous. Mushrooms are of of different categories some of them can be used for daily needs but some of them has toxic ingredients in them which is harmful for consumption. So classifying the mushrooms correctly becomes very important as if someone uses a toxic mushrooms it can lead to serious health effects. Classification algorithms helps to solve this issue. Here we are using various ml algorithms to classify the mushrooms some of them are random forest and Decision tree Algorithm. Our main goal is to categorize the mushrooms correctly. We have achieved the highest accuracy with random forest. It is able to differentiate the mushrooms most effieciently and effectively. These results shows that classification algorithms can prove to be very important in categorizing mushrooms.
Authors - Pradnya Apte, Dipti Durgesh Patil Abstract - Agriculture forms the backbone of many developing countries, including India. Accurate crop yield estimation can give societies a better handle on food security and resource management. Existing studies on crop yield and harvest prediction apply deep learning as well as machine learning models. Various crop parameters like soil type, climate, water content, and so on have been used to predict crop yield. More advanced techniques include the use of satellite imagery and optical and SAR data along with plant indices like NDVI. Machine Learning algorithms, including KNN, SVM and Random Forest regression have been widely used for yield estimation. Deep learning approaches work by extracting salient and relevant features from images or non-visual data to estimate crop production. Networks like 3D-CNNs, LSTMs and Auto encoders have achieved significant improvement in accuracy in estimating crop yields from satellite images. This paper aims to summarize the techniques and models being used for the purpose of yield estimation along with limitations, and possible areas of further study.
Authors - Kasif Qamar, Supriya Narad Abstract - Emotional intelligence in Artificial Intelligence is an important and exciting growing field with much potential for assisting human-computer interactions in different domains. Self and others’ emotional awareness are referred to as emotional intelligence and is steadily being deemed paramount to develop technologies in artificial intelligence that will in the end be effective when handling needs and human states. In recent researches, it has been observed that although in real way, AI cannot be so expected to feel like human beings simulation does exhibit mimic emotions that add to the enhancement of the user experience and acceptance more so in service-oriented applications Emotional Simulation of Artificial Intelligence and The awareness of the EI is really rising in various organizations in today’s workplaces especially with integration and Automation of work using artificial intelligence. Since the roles and responsibilities of human beings working in industries will change due to AI and automation, EI skills will be are important to be applied by the employee’s at all organizational levels. However, machines are good at things that involve rules of logic and therefore getting them to understand and among the limitations of URLs of expressing human feelings, answering to the feelings still remain a problem to be solved in AI. Hence, improving EI becomes something of enormous value to be competent in the new era of jobs.
Authors - Anup Vinod Pachghare, Smita Deshmukh, Satish Salunkhe Abstract - Human Resources Management plays a key role in the company’s growth by recruiting high-quality employees and evaluating their performance by using the Machine Learning (ML) technique. Despite these rigorous efforts some employees still resign before their contracts expire, which negatively impacts business. Existing methods have considered various factors influencing employee turnover across different employee groups. This paper proposes an Ensemble Learning approach which integrates AdaBoost, K-Nearest Neighbour (KNN), Random Forest (RF), stacking and Voting to enhance churn prediction accuracy. The ensemble learning mitigates the risk of overfitting by combining predictions from multiple models, making it less sensitive to irrelevant features. This approach efficiently captures diverse patterns in the employee churn data, achieving better accuracy. AdaBoost captures complex patterns, while KNN extracts valuable data from employee churn data. By stacking these methods, their combined strengths lead to enhanced accuracy in predicting churn data. Initially, the data collected from employee churn records and pre-processing phases handle unwanted noise and min-max normalization, which standardizes the feature vector to ensuring uniformity across the dataset. The proposed ensemble model obtained 91.06% accuracy, and 0.8853 of recall on the employee churn dataset compared with conventional techniques like Artificial Neural Networks (ANN).
Authors - Pranav Jawale, Srushti Bonde, Dhruv Gidwani, Omkar Aher, Bhavana Kanawade Abstract - The GPS-based toll tracking and speed monitoring system uses GPS technology to revolutionise highway toll collection and road safety. The system calculates payables based on distance travelled by the vehicles on highways and monitors vehicle speeds to ensure fair charges while discouraging speeding. The system delivers real-time notifications for toll deductions and penalties, supporting transparency and eliminating the need for traditional toll booths. Users can access a dashboard to track their journey, view toll routes, and review their vehicle and transaction history, offering a comprehensive and user-friendly experience.
Authors - Akshay Kumar, Sudhir Agarmore, Edidiong Akpabio, Kumar Gaurav, Nitesh Kumar, Aditya Mandal Abstract - Intrusion Detection Systems (IDS) has been seen to be an integral aspect of network security, where an extra layer of protection mechanisms may contribute to protection against different kinds of cyberattacks. With fast-evolving cyber threats, from simple malware to sophisticated zero-day attacks, continuous developments are required in IDS technologies. Traditional IDS models, such as signature-based detection, work better when applied to known threats but are pretty weak against emerging and unseen types of attacks. On the other hand, anomaly-based IDS models give pretty good results in finding unknown attacks, but most of them report a high false-positive rate. Finally, this work provides an overview of current state-of-the-art IDS methods and focuses on machine learning, deep learning, and hybrid models for intrusion detection. We further discuss the benefits and limitations of both the supervised and unsupervised learning algorithms and their applications in anomaly detection and pattern identification within network traffic. It investigates the applications of deep learning methodologies, such as CNNs and RNNs, within the IDS framework. These models give a high chance to work with complex data and detect various kinds of sophisticated attacks efficiently. Hybrid systems that combine traditional detection methods with machine learning demonstrate better accuracy and fewer false positives. Among the future directions of research that will be discussed are development challenges for the scalability of IDSs, real-time detection, and handling zero-day vulnerabilities for improving the efficiency of IDS in securing modern network infrastructures.
Authors - Koyana Jadhav, Aditya Shinkar, Mayank Sohani Abstract - In the light of dynamic pricing policy, flying is becoming too expensive, and it's really hard to book tickets at proper prices. In response to this, researchers began discussions on how machine learning models could be used to predict an approximate fare for a flight so that the passenger could purchase at the most ideal time to get low fares. These models consider travel dates, destination, airlines, stopovers, timing of booking, holidays, and demand. Techniques used are Decision Trees, Random Forest, Gradient Boosting, and ANNs. These have different strengths-some, such as ensemble methods, Random Forest and Gradient Boosting, with high robustness in terms of predictions because they average multiple decision paths; ANNs represent complex, non-linear relationships but at the cost of significant computation. Model performance was evaluated using Mean Absolute Error, Root Mean Square Error, and R-squared. This research informs passengers on price trends, and better booking decisions will be achieved. Real-time data integration and more advanced algorithms comprise future improvement prospects. The research work bridges the gap between revenue strategies employed by airlines and the need of the travellers to travel affordably, thereby optimizing passengers travel costs.
Authors - Ketan J P, Amulya A Shetty, Ashwini Bhat Abstract - In an era defined by using heightened virtual connectivity, the safety and privateness of voice communication are important. This research paper introduces an innovative system designed to establish secure voice communication between two systems. Including a combination of cryptographic algorithms, including RSA, 3DES (Triple Data Encryption Standard), modified RSA, and modified 3DES, the proposed solution gives confidentiality and integrity of voice data during transmission. The need for such a system is underscored by the growing demand for secure communication in sectors where sensitive information is exchanged, such as military and intelligence operations, healthcare, and business. This study explains a client-server architecture where the client system employs RSA and 3DES for encryption, while the server system utilizes corresponding decryption mechanisms. Socket programming serves as the connectivity bridge, with the server's IP address acting because the transmission key.
Authors - Karthika M, Raghu Nandan KS, Salanke Anni Rao, Ramkumar S Abstract - Fire detection and prevention are essential to preventing fire spread and substantial loss or damage, especially in remote regions like lakes where traditional approaches are almost useless. This review covers fire alarm system advances, focusing on machine learning (ML) to improve detection. The paper also examines how these innovations improve remote fire alarm systems functioning and how they integrate with emerging IoT protocols, sensor networks, and radio technologies like LoRaWAN. It focuses on ML models like CNNs and deep learning to analyse sensor data and detect fires accurately and quickly. This paper discusses fire detection innovation and how ML can improve future systems' coverage and accuracy.
Authors - Vishal Karpe, Meetu Kandpal Abstract - In order to better understand how Amazon Web Services (AWS) charges for its services and how users can efficiently manage expenses, this research examines the price choices offered by AWS. It illustrates how various pricing strategies such as pay- as-you-go, volume discounts, and cost savings affect users' AWS spending by explaining them and providing examples of how they work. To emphasize the advantages and disadvantages of each, the research also contrasts the costs of Microsoft Azure and Amazon Web Services. In order to track and optimize user spending, it also comes with features like AWS Budgets, AWS Pricing Calculator, and AWS Trusted Advisor. The goal is to provide clients with an extensive manual on how to maximize their financial and material investments when utilizing cloud services.
Authors - Ritveek Rana, Manisha Manoj, Anitha Dhanasekaran Abstract - The escalating threat of climate change has made it imperative to understand and mitigate the environmental impact of human activities, particularly by reducing carbon footprints. This research ventures on predicting carbon emissions for India using autoregressive integrated moving average (ARIMA) models. The findings may signal appreciable implications for decisions in governmental policies and energy sector. This study highlights a potential situation for India in the coming years due to increased expenditure of carbon-based fuel sources to meet the need for increased manufacturing and demand. The ARIMA models developed in this research can serve as a valuable tool for forecasting carbon emissions and guiding future energy policies.
Authors - Amoggha C H, Padmapriya R, Adithya Narayana Holla, Manoj C Aradhya Abstract - This paper presents the development of a deep learning model for recognizing handwritten Kannada characters. Kannada character recognition presents unique challenges due to the complexity of the script and the variety of symbols. To address these, we utilize a hybrid model combining ResNet50 and VGG16 architectures. ResNet50 is leveraged for its ability to train deep networks on complex patterns, while VGG16 excels in capturing detailed feature representations. The model is trained on carefully pre-processed datasets, optimized through iterative parameter tuning to ensure high accuracy and robustness. The backend infrastructure uses Flask and TensorFlow, with the frontend built using java script, HTML, and CSS. The system features a sketchpad where users can draw Kannada characters, which are then processed by the deep learning model for recognition. An interactive tool further supports language learning. Through extensive testing, the system has proven to be reliable and effective. This project represents a significant advancement in automated Kannada language processing, offering a powerful tool for character recognition. By enabling accurate, efficient recognition, it contributes to promoting linguistic diversity and inclusivity, making it an invaluable resource for Kannada language processing applications.
Authors - Edidiong Akpabio, Sudhir Agarmore, Akshay Kumar Abstract - As digital technologies become increasingly ingrained in critical energy infrastructure, a looming threat is cyberattacked as the sector has absorbed all the data acquisition and supervisory control systems, smart grids, and industrial control systems, with associated operational efficiencies, but at the cost of an expanded attack surface in terms of cyber threats. This paper aims at identifying the unique cybersecurity issues in CEI that pose threat scenarios that include, for instance, their vulnerability to legacy system vulnerabilities, insider threats, and more complex attack vectors such as advanced persistent threats and ransomware. Finally, it points out the need for proactive risk assessment, network segmentation, advanced defence mechanisms such as intrusion prevention and detection systems, and zero trust architectures. Newer technologies like machine learning, blockchain, artificial intelligence, and quantum cryptography offer new opportunities for better cybersecurity. It can foresee the occurrence of a particular attack through AI-based threat detection systems. Blockchain provides security in energy transactions while making unbreakable encryption of critical communications. This paper insists on better, much more comprehensive disaster recovery and incident response plans to minimize the impacts caused by cyberattacks and it concludes by advocating a multi-layered cybersecurity strategy with the intent of integrating advanced detection systems and risk management practices into a solid collaboration between the government and private sectors aimed at enhancing the stability of global energy supplies.
Authors - Archana L. Rane, Sanskruti R. Talele, Rashika A. Ghavate, AditiS.Khairnar, Harisha A. Chothani Abstract - Nowadays, the world is increasingly focused on health care, with hair care emerging as a key aspect of personal well-being. Many people face confusion when selecting the best shampoo based on their scalp and hair health. The purpose of this study is to provide a natural alternative to conventional shampoos by incorporating eggshell powder, a readily available, eco-friendly resource, into future hair care formulations. A comprehensive study was conducted to evaluate various shampoos currently available and to identify the benefits of eggshell powder. This study highlights the potential of eggshell powder in enhancing shampoo production. Machine learning algorithms such as Naive Bayes, K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Random Forest were employed to analyze manufacturing parameters and optimize the absorption of eggshell powder. The results of the analysis revealed varying accuracies for each model: Naive Bayes (52%), KNN (71%), SVM (72%), and Random Forest (82%). These techniques allowed for precise adjustments to ingredient concentrations and interactions, improving the overall efficacy of the shampoo. The results demonstrate that shampoos formulated with eggshell powder offer several advantages, including stronger hair, better moisture retention, and enhanced scalp health. Additionally, eggshell powder proved to be a sustainable material, aligning with growing consumer demand for environmentally friendly products. This study highlights the potential of using natural resources and machine learning to drive data-driven improvements in hair care formulations, offering a promising alternative to conventional products while meeting the increasing preference for sustainability.
Authors - Vidhi Aakash Pandya, Meetu Joshi Abstract - The present research looks at how artificial intelligence (AI) is affecting various interpersonal sectors and offers opportunities, problems, and potential solutions. It investigates how artificial intelligence (AI) has developed into a crucial instrument for tackling social problems and providing answers in a variety of fields, including healthcare, education, the environment, and agriculture. The history of AI's development from historical turning points to modern deep learning applications opens the study. It then dives into a thorough review of the literature, highlighting important research and the condition of AI application in many industries at the moment. The study examines AI's potential applications in healthcare, with a focus on tailored treatment methods, diagnostics, and disease prediction. It also addresses ethical issues. AI is being used in education to investigate how diversity and specific instruction might be achieved using voice assistants and virtual mentors, among other technologies.
Authors - Yash Dargude, Jui Ambekar, Yash Gadakh, S.T Gandhe Abstract - Mental health disorders, such as depression, anxiety, and stress, are global challenges that significantly affect individuals’ well-being and productivity. Early detection and diagnosis are crucial for effective intervention, yet traditional methods often rely on subjective assessments, leading to potential delays. Electroencephalography (EEG) has emerged as a promising non-invasive tool for objectively monitoring brain activity, offering valuable insights into mental health conditions. This survey paper explores the current state-of-the-art in mental health detection using EEG signals. We provide an overview of EEG-based systems, highlighting key signal processing techniques such as filtering, artifact removal, and noise reduction. Feature extraction methods, including time-domain, frequency-domain, and time-frequency domain techniques, are reviewed to emphasize how patterns in brainwave activity correlate with mental health states. Additionally, we examine various machine learning and deep learning algorithms, such as Support Vector Machines (SVM), Random Forest, and Convolutional Neural Networks (CNNs), which have been applied to classify mental health conditions based on EEG data. The paper also presents a comprehensive analysis of the effectiveness of these models in detecting specific mental health conditions like depression, anxiety, and stress. We discuss the challenges faced in using EEG for mental health detection, such as signal variability and the need for large datasets, and propose future directions for enhancing the accuracy and generalizability of these models. This survey aims to contribute to the development of more reliable, EEG-based diagnostic tools for mental health assessment.
Authors - Krishan Pal Singh, Emmanuel S. Pilli, Vijay laxmi Abstract - Tor network provides anonymity and privacy to online users. Hence, analyzing Tor traffic to identify applications and services, especially when encrypted tunnels and pluggable transports are used, remains a significant challenge. This paper presents a novel framework for identifying obfuscation techniques by analyzing their unique traffic characteristics, such as packet sizes, inter-arrival times, byte sizes, and byte frequencies. A custom-built network traffic collection environment is established to evaluate the proposed framework. A large Tor traffic dataset is created that contains Obfs4 and Snowflake Plugin traffic, ensuring realistic user behavior simulation utilizing modified Tor browser configurations. The framework leverages a combination of statistical analysis of encrypted payloads, examines timing sequences during authentication, and packet length filtering. The Traffic data is evaluated on diverse deep learning models, such as Neural Networks, Adaboost, and XGBoost, achieving high accuracy rates (95% to 98%) across different Tor plugins. The proposed framework demonstrates robustness with low false positive rates. It is also adaptable to new Tor obfuscation techniques such as Obfs4 and Snowflake. The research findings highlight the importance of using up-to-date and diverse datasets to train effective Tor plugin identification models, with potential applications for improving Tor network security.
Authors - Vinayak Suresh Bhajantri, Aishwarya B Kalatippi, Rahul B Sajjan, Babusingh Ramsingh Rajput, Kiran M R, Suneeta Budhihal Abstract - In recent years, natural disasters like earthquakes, tsunamis, floods, and storms have happened frequently, causing severe damage. These disasters have shown how crucial it is to have reliable communication for rescue operations. Often, disasters damage communication network. The heavy demand for data transfer on the Internet is pushing its infrastructure to the limit, making it difficult to respond quickly to emergencies and disasters. To solve this problem, Internet networks need to prioritize certain types of data traffic: Security, Health, and Emergency (SHE) data traffic. These specialized networks work in private domains to support specific tasks for particular groups of users. We proposed network flow priority management system based on Software-Defined Networking (SDN) to give SHE data traffic the highest priority. Using the Mininet simulator, we tested our system extensively. The results show significant improvements in handling SHE data traffic, ensuring that during network congestion, SHE data is transmitted quickly, improving the effectiveness of emergency response efforts.
