Proceedings of the 7th International Conference on Communications and Cyber Physical Engineering

Table of Contents

1. Advancements in Air Quality Forecasting

   Suparna Das, D. Harshitha, B. Navouya, K. Bhavya

   Abstract

   Advanced machine learning approaches, especially Long Short-Term Memory models, are used in the field of air quality forecasting. The dataset is organized to capture the concentrations of several pollutants gasses as factors used to determine the Air Quality Index. After a thorough comparison analysis of the dataset with other models such as K-nearest neighbor, gradient booster, linear regression, decision tree, Support Vector Machine and Long Short-Term Memory, it is shown that Long Short-Term Memory is the most effective in providing increased predicted accuracy. The project also has a strong security mechanism that includes login and user registration capabilities, strengthened by the generation of OTP using email authentication. By allowing users to enter important parameters such as the name of the city, the country, and the goal date for AQI prognostication, the user interface simplifies the procedure. Then, with ease, the interface smoothly displays the predicted AQI value for the given date, improving usage and accessibility.

2. IOT-Based Patient Health Monitoring System

   Swathi Gowroju, R. Vaishnavi, Shaik Noorus Sabiha, P. Rakesh Anand, P. Nithish Kumar

   Abstract

   In order to offer practical healthcare alternatives, we present and put into practice an Internet of Things-enabled health monitoring system in this publication. Because technology is developing so quickly, experts are always looking for new electronic tools that can quickly identify abnormalities in the body. The development of contemporary, non-invasive medical support systems is made possible by the Internet of Things (IoT). Our healthcare monitoring system is made to provide people with diseases like COVID-19, high BP, and hypertension with prompt access to medical care, particularly in rural areas of developing nations like Bangladesh. The expense of buying particular medical equipment or making frequent hospital trips might be prohibitive for the typical person, and many of these patients do not have instant access to healthcare services or emergency clinics. Our designed system tracks vital health indicators such as blood oxygen levels, heart rate, and body temperature. Making healthcare more accessible and cheap for regular people is our main objective. In addition to being financially viable, this method makes it simple for patients to receive individualized medical treatment in the convenience of their own homes. This study presents an Internet of Things (IoT)-based solution that makes it easier and less expensive to provide care from home using often complicated medical gear.

3. Deep Neural Network for Respiratory Disease Prediction Using Respiratory Sound

   Nagaratna P. Hegde, Sireesha Vikkurty, Sriperambuduri Vinay Kumar, Tirupati Naredla, Sai Pranay

   Abstract

   Objective of the paper is to explain the predictive model developed for identifying diseases related to respiratory system like pulmonary fibrosis, asthma, lung cancer chronic obstructive pulmonary disease (COPD), pneumonia, URTI, Bronchiectasis, and Bronchiolitis using Deep Neural Networks (DNNs) or deep learning techniques. Utilizing respiratory sound data as input, the constructed DNN model is designed to accurately predict the status of a person’s respiratory system. Notably, the technique used is not only capable of distinguishing between various respiratory diseases but also accurately determines if an individual’s respiratory system is healthy. Key components of the project include the utilization of GRU (Gated Recurrent Unit) networks for temporal data analysis, data augmentation techniques to enhance model robustness, and feature extraction methods to extract prominent features from the data corresponding to respiratory sound. The ultimate aim is to gain higher accuracy and precision in disease classification, thereby facilitating early diagnosis and intervention for respiratory conditions.

4. Predicting and Addressing Mental States in Real-Time Using Wearable Vital Data

   Nagaratna P. Hegde, Sireesha Vikkurty, Guru Sai Shreesh, Tarun Krishna Dodda

   Abstract

   In an age characterized by the widespread adoption of wearable technology and growing awareness of mental health issues, this presentation introduces an innovative method for predicting and addressing mental states in realtime. Our study utilizes data from wearable sensors like heart rate monitors and oxygen saturation sensors to continuously track individuals’ physiological reactions. We suggest a comprehensive system that merges AutoRegressive Integrated Moving Average (ARIMA) modeling with a Random Forest-Based Ensemble Approach to offer instantaneous understanding of users’ mental conditions.

5. Integration of Solar Charging System in Electric Cycles for Sustainable Mobility

   V. Suma Deepthi, P. Swathi, B. Sri Vaishnavi, G. Namratha, M. Aparna, A. Kavya

   Abstract

   The integration of a solar charging system involves mounting photovoltaic panels onto the frame or surface of electric cycles. These panels convert sunlight into electrical energy, which is then used to charge the cycle's battery either while it is in use or parked. The efficiency and practicality of this system depend on factors such as the size and quality of the solar panels, the cycle's energy consumption and the availability of sunlight. This approach offers a sustainable solution for urban mobility by reducing reliance on grid electricity and minimizing environmental impact.

6. Advancing Reversible Data Hiding Using GAN-Enhanced Steganography and Cryptography Synergy

   P. Vishnupriya, Manoj Rajanala, S. P. Maniraj

   Abstract

   In today’s age, data security and privacy are paramount concerns. Reversible data hiding is a technique that’s used to embed data into a cover media in a manner that allows both embedded data and original media to be recovered. Unlike traditional data hiding techniques, which permanently alter the cover media, RDH ensures that the original content can be restored exactly as it was before embedding. The concept of RDH relies on the fact that the embedded data does not cause irreversible distortion or permanent damage to the cover media. Along with this feature it is necessary for us to increase the data security of hidden information. This paper introduces a methodology that combines Generative Adversarial Networks (GANs) with steganography and cryptography techniques to advance reversible data hiding and enhance data security. By integrating Generative AI-enhanced steganography, the paper aims to embed data within these cover images in a reversible manner, ensuring that the original media can be recovered with enhancement in image quality in comparison with the original image. Furthermore, the integration of cryptography techniques ensures the integrity and confidentiality of the concealed data. This makes the data resistant to unauthorized access and detection. The synergy between Generative AI-driven steganography and cryptography in RDH helps to sustain key elements of reversible data hiding that includes reversibility, embedding capacity and data security. The enhancement in the performance can be used in various applications such as medical imaging, forensics, intellectual property protection, secure communications, and military applications.

7. Linguistic Sensitivity Upgrade: AI-Enabled Normalization of Offensive Text

   Swathi Gowroju, Danniel Dominic Savio Kennedy, D. Vinodin, D. Pavithra Jahnavi, A. Sriram Goud

   Abstract

   In recent years, the surge of harmful content on internet has impacted a vast majority of users, necessitating advanced measures to curb this trend. In this paper, we delve into the application of linguistic sensitivity upgrades using some techniques to address and mitigate offensive text. With the rampant growth of social media, there has been alarming increase in harmful and offensive content that negatively impacts users. Such content not only creates an unsafe online environment but also propagates hate, misinformation and discord. While there are mechanisms to detect and remove offensive content, few focus on contextual understanding and transformation of this content into a non-offensive format. Thus, there is an imminent need for an advanced solution that doesn’t just detect but also convert offensive text into contextually appropriate and respectful language.

8. Diabetic Retinopathy Detection Using Deep Learning Techniques

   Vaddhiraju Swathi, K. Sathish Kumar

   Abstract

   High blood sugar levels are what lead to diabetes. Numerous illnesses, including heart conditions, kidney problems, nerve damage, and eye impairment, can be brought on by diabetes. Diabetic retinopathy is one such complication brought on by diabetes that, if not treated or detected in a timely manner, may also result in vision loss. By training algorithms on retinal images to recognize specific features, categorize the presence or absence of the condition, or divide the image into discrete parts, machine learning can be used to recognize and diagnose diabetic retinopathy. Support Vector Machine, logistic regression, convolutional neural network, K-Nearest-Neighbor, and random forest are the current techniques utilized to identify diabetic retinopathy. The most often used deep learning methods for image detection are convolutional neural networks. To perform image classification tasks, a Convolutional Neural Network (CNN) architecture known as VGG16 was trained on a sizable dataset of images. For image classification problems, a well-known deep learning architecture is VGG16. The images are classified using the retrieved features using a variety of ML algorithms, such as KNN, SVM, Logistic Regression, Boost, AdaBoost, Decision Tree, Voting Classifier and other algorithms. This methodology is used to group diabetic retinopathy into one of five severity-based classifications (0,1,2,3,4). The proposed system will facilitate the removal of ambiguous diagnoses done by ophthalmologists. This would enable the faster and more accurate prediction and diagnosis of patients’ condition.

9. Exploring Hybrid Approaches in NLP: Enhancing Text Summarization and Classification Techniques

   Gadi Mounica, Suneetha Eluri

   Abstract

   Classifying and categorizing texts is a paramount responsibility in applications of autonomous natural language processing (NLP), like subject categorization, analysis of feelings, identification of purpose, prevention of spam, and routing of emails. Classifying texts using machine learning is capable of helping companies review and organize their paperwork quickly, efficiently, and affordably, automating steps and improving judgments based on data. The study explores text-processing hybrid architectures. It discusses the benefits and drawbacks of integrating derivative and generalized summarization with different classification algorithms and how hybrid models can capture factual content and key semantic relationships in text to produce more informative summaries with better classification accuracy. The report critiques hybrid approach research on varied datasets and compares it to standard methodologies. The article discusses vital evaluation metrics, domain-specific hybrid model adaption, and outstanding research concerns. Finally, the study examines the future of this promising field. Hybrid NLP techniques can transform how we extract knowledge and make sense of the rising universe of text data by investigating unique hybrid architectures, including deeper semantic understanding, and establishing robust assessment frameworks.

10. OTP Based Transaction Using Facial Recognition

    D. Srinivasa Rao, V. Sravan Kiran, M. Koteswara Rao, Kristam Naga Sai Nithin, Madhavaram Ashithesh Kumar, Nirmala Sriya, Varsha Bandla

    Abstract

    Banks provide various services for their customers. Most customers use ATM cards to access the features given by the bank. However, there are issues faced by physical cards in our daily lives. Some security measures need to be implemented to increase security. Several researchers have created an abstract model for card less electronic ATMs that permits users to do several functions, including fund transfers, balance checks, and cash withdrawals. In this proposed method, we are implementing three steps. First, the user has to enter his details. After these details are verified with the database, the user is allowed to the next step if and only if the details match. The next step is an OTP verification. After the OTP is verified, the user is asked for face recognition. If all the above three steps are cleared then the user is allowed for the banking operations. The proposed method's primary goal is to use facial recognition technology to conduct secure transactions in ATM settings.

11. Scalable Machine Learning for Big Data Mining: Challenges and Opportunities

    Sargari Swapna, Vemula Shiva Kumar, Pechetti Sujani, Maggidi Mounika, Sadula Sai Prasanna, Appari Lakshmi Prasanna

    Abstract

    This paper addresses the challenges and opportunities in scalable machine learning for big data mining. We explore the foundational concepts of machine learning and delve into the unique characteristics and challenges posed by big data. Our focus is on scalable algorithms, distributed computing, and parallelization techniques to enable efficient processing of large-scale datasets. The paper also examines the importance of scalable data preprocessing, feature engineering, and model training and evaluation. Deep learning approaches in the context of big data are discussed, along with considerations for real-time analytics and cloud-based solutions. We highlight domain-specific applications and ethical considerations in scalable machine learning. The paper concludes with an outlook on future trends and challenges, providing insights for researchers and practitioners in the evolving landscape of big data and machine learning.

12. Design and Establishment of Solar Photovoltaic System on Roof Top of Campus Electric Vehicle

    Gouthami Eragamreddy, B. Sahithi, B. Tanuja, K. Aparnaa, A. Gayathri, Sai Sreeja

    Abstract

    Solar energy, harnessed through solar panels, is a widely utilized renewable resource. Solar panels are most efficient when maintained at a consistent temperature, yielding optimal energy output. Solar photovoltaic energy finds a wide range in power generation. This research paper contributes to utilizing generated solar power to charge the electric vehicle battery. Establishing a solar charging system for electric vehicles (EVs) utilizing a charge controller, aiming to reduce electricity costs, and charging losses. Implementation of this system on EV roof top contributes to sustainability efforts. The integration of solar panels onto EVs and subsequent simulation using MATLAB Simulink is explored to evaluate the feasibility and efficacy of solar energy utilization. This paper also shows the hardware implementation results of the solar panel on EV roof top in terms of reduced power usage in charging EV.

13. Criminal Identification During Activity Through Videos Using Python

    Swathi Gowroju, J. Varun Kumar, H. Rohan, G. Samhith, K. Abishek

    Abstract

    Since there has been a rising number of anti-social activities recently, security has become increasingly important. In this article, we create and implement a software for identifying criminals. This means that the same has to be automated. To help with the quicker determination that the odd activity is abnormal, it is also necessary to indicate which frame and what portion of those contains the unpredicted activity. In order to accomplish this, video is divided into frames, and each processed frame’s people and activities are examined. Machine learning techniques and algorithms help us to make a wide range of things possible. Strong and effective security systems are very essential in today’s world, as crime rates are skyrocketing using state-of-the-art face recognition algorithms and technologies, the system is able to precisely identify and locate human faces in a variety of dynamic environments, including subjects, locations, positions, noise levels, and lighting, all of which are frequently seen in surveillance footage.

14. A Deep Learning-Based Approach for Early Detection of Oral Cancer Using Class Activation Maps

    Madiha Sadaf, Amtul B. Ifra

    Abstract

    Oral squamous cell carcinoma, another name for oral cancer, is one of the ten most common diseases worldwide, accounting for approximately 500,000 new cases and 350,000 fatalities annually, with India accounting for one-third of these instances. There is an urgent need for objective, distinctive technologies that enable early, precise diagnosis. Through transfer learning on Inception-ResNet-V2, we built a method to identify images as “suspect” and “normal” and automated heat maps to highlight the area of the images most likely to be engaged in decision-making. Using the Kaggle datasets with photographic shots of 87 and 44 cases, we assessed the feasibility of the created approach. The system was tested using both 10-fold cross-validation and leave-one-patient-out validation procedure. The study’s novel findings and methods include developing and validating our styles on two datasets gathered from various locations in India, demonstrating that using patches rather than the full lesion image improves performance, and determining which areas of the images are predictive of the classes using class activation map analysis.

15. Streamlining Data Mining Processes with Machine Learning: Automated Tools and Frameworks

    Maggidi Mounika, Appari Lakshmi Prasanna, Sadula Sai Prasanna, Sargari Swapna, Pechetti Sujani, Vemula Shiva Kumar

    Abstract

    This paper explores the study of automated tools and frameworks that are aimed to promote efficiency and effectiveness in the process of knowledge discovery from massive datasets. The paper investigates the integration of machine learning approaches to expedite data mining procedures. The purpose of this research is to analyze the synergy that exists between conventional data mining techniques and cutting-edge machine learning algorithms, with the goal of showing the potential for these two types of methodologies to work together to tackle difficult analytical problems. A detailed overview of the functions and advantages offered by the automated tools and frameworks that permit smooth integration is provided in this article, which analyzes the most important automated tools and frameworks. Because of the increasing expansion of data in a variety of fields, it is necessary to have data mining techniques that are both efficient and automated to extract meaningful insights.

16. Transforming Legal Document Generation: A BERT and BART Approach

    V. Surya Narayana Reddy, Maddimsetty Sesha Satya Priyanka, Kala Krishna Jyothika, Ravi Kanth Motupalli, S. Annapoorna

    Abstract

    Introducing a system driven by artificial intelligence that aims to simplify legal documentation and ensure compliance with Indian laws. By automating the creation of papers in plain English, the approach gets beyond the challenges presented by complex legalese. It also contains features that facilitate legal compliance inspections for individuals and small businesses. By bridging the gap between user accessibility and legal obstacles, this approach has the potential to increase the accessibility of justice and legal services in India.

17. Safeguarding Android Devices Through Deep Learning Against Malware

    Neda Fatima, M. Deepthi

    Abstract

    Recent advancements in computer technology have shifted human experiences from the physical to the virtual realm. However, this transition has also brought about an increase in malicious software, or malware, used in cyberattacks. These malware variants continuously evolve, employing sophisticated techniques like advanced packing and obfuscation, posing challenges for traditional detection methods. To address this, an approach called Safeguarding Android Devices through Deep learning against Malware is proposed. This technique integrates data pre-processing and ensemble learning, leveraging two machine learning models—Random forest Classifier (RFC), Extra Tree Classifier (ETC), along with deep learning methods Convolutional neural network (CNN)and Artificial neural network (ANN). Through comprehensive experimental analysis, the CNN deep learning technique demonstrates superior performance compared to existing methods in detecting Android malware.

18. Identifying Malicious Software by Analysing Distinct Artifacts Using Machine Learning and Deep Learning Algorithms

    Nimra Mohammed Abdul Quddus, I. Ravi Prakash Reddy

    Abstract

    There are a number of websites that fulfil a range of roles in the modern digital world. These functions include the transmission of information, the construction of connectivity, and other functions. The identification of dangerous API calls and photos was the primary focus of this research, with the ultimate goal of discovering the precise risks that were addressed by the files. The investigation was carried out with the intention of completing the inquiry. The examination of application programming interface (API) calls and graphics makes it possible to discriminate between programmes that are hazardous and those that are not harmful. In this particular investigation, classification analysis was employed for the objective of classifying malware based on textual information. The classification models that were utilised consisted of five unique models, including support vector machine, naïve bayes, and random forest, KNN, and decision tree. These models were selected because of the numerous benefits that machine learning and deep learning techniques offer, as well as the widespread usage of these approaches. During the course of the study project, the Convolutional Neural Networks technique was employed for the goal of categorising the pictures that were included in the Malimg dataset. Both the Malimg dataset and the APIMDS dataset were utilised in order to complete the evaluation of this approach. The results of the trials reveal that the suggested technique is capable of efficiently categorising malware with a high level of accuracy, which is superior to the methods.