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

Table of Contents

1. Fortifying Cybersecurity: A Deep Learning Paradigm for Comprehensive Threat Defense

   Nayini Sai Nithin, Gade Vishwas, Gudapareddy Mani Prakash, Peddineni Varshith, D. Radha

   Abstract

   In the current age of interconnectedness, individuals leverage the benefits of communication and collaboration, however, this interconnectedness also exposes systems to an escalating array of cybersecurity threats, posing challenges to security and integrity. This project addresses this concern by concentrating on the development of a defense system that leverages advanced deep-learning techniques. The primary objective is to accurately detect intrusions in real-time, classify malware. Going beyond conventional cybersecurity measures, this defense system integrates deep learning methods, ensuring flexibility and responsiveness to counter evolving tactics employed by malicious entities. The defense system encompasses both Packet Files and Executable Files, significantly augmenting cybersecurity capabilities. To seamlessly align with established practices, LSTM, GRU, and RNN Models are employed for both Packet Files and Executable Files. The harmonious integration of these technologies with existing infrastructure is imperative for cultivating a secure digital environment.

2. Design and Execution of High Reliability PDDC-Net Model for Classifying and Identifying Plant Leaf Disease Using Deep Learning

   Pinamala Sruthi, R. Venkateswara Reddy, L. Chandra Sekhar Reddy, K. Srinivas, K. Srinu, Rakshita

   Abstract

   The imaginative capacity of a nation is contingent upon the development and performance of its agriculture sector. Agricultural, as the primary source of raw materials and food, serves as the fundamental basis for all nations. Agriculture serves as a significant provider of sustenance for human populations. The identification of diseases of plants has become increasingly crucial as a result. The following are established methodologies for the identification of plant diseases. Nevertheless, the identification of leaf diseases frequently involves the utilization of visual inspection by plant pathologists or agricultural professionals. The process of identifying plant leaf disease using this approach can be subjective, costly, and time-intensive. It typically necessitates the involvement of a large team of experts possessing extensive knowledge of plant infections. The identification of plant leaf diseases can also be facilitated through the utilization of a software solution that has undergone rigorous experimental evaluation. In contemporary times, the utilization of deep learning and machine learning has become prevalent. The primary objective of this endeavor is to implement the utilization of deep learning models in the deployment of Plant Disease Identification and Classification (PDDC-Net). The procedure of preparation additionally encompasses the elimination of diverse sources of noise, thereby restoring the pictures within the information set. Moreover, the PDDC-Net employs Convolutional Neural Networks (CNNs) based on residual networks (ResNet-CNN) for the purpose of extraction of characteristics and classification. Based on empirical evidence, the proposed PDDC-Net model demonstrates a commendable level of reliability in the identification and classification of illnesses of the leaves.

3. Statistical Comparison Between Topological Indices and Toxicity Values of Natural Products

   G. Keerthi, M. Siva Parvathi, R. Lakshmi, T. Sukeerthi, E. Lavanya

   Abstract

   Graph theory is often applied to determine topological activity interactions of chemical substances by computing indices of topology. It has several uses in the development of in silico technologies. Machine learning, which has been extensively applied in numerous fields and is especially useful in the era of big data and artificial intelligence, can also be used to predict toxicity. However, topological indices need further consideration when used for predicting the toxicity of natural substances (flavonoids). The flavonoids are possessing 15 carbon atoms: C6 – C3 – C6 system. The basic molecular graphs of flavonoids are used to calculate various topological indices.

   In this paper, three types of metrics were developed per specified chemical components in natural products including the Statistical comparison amongst the topological index values (Wiener number, Polarity number Platt number) and the toxicity values of natural products was discussed.

4. Customer Segmentation

   T. Ammannamma, Shaik Sadaf Tabassum, G. Alekhya, G. Pallavi

   Abstract

   Customer segmentation is an important approach for modern firms since it allows them to categorize their customers based on shared traits. This approach optimizes marketing efforts by targeting specific customer segments, enhancing resource allocation efficiency, and fostering opportunities for cross and up-selling. By using large datasets, segmentation analysis helps marketers identify distinct client segments with precision, taking behavioral, demographic, and other factors into account. Crucially, segmentation should prioritize long-term customer lifetime value (CLV) over short-term gains, ensuring that marketing actions are tailored to maximize overall customer value to the business. By understanding how various marketing initiatives influence customer behavior, organizations can effectively engage with each segment, fostering enduring customer relationships and driving sustained revenue and profitability. Customer Segmentation thus emerges as a strategic imperative in contemporary business landscapes, guiding personalized interactions that resonate with diverse customer preferences and behaviors.

5. Enhancing Fetal Health Monitoring Through GAC Net

   Nagabothu Vimala, N. Srihari Rao, G. Deepika

   Abstract

   The main goal at this point in the pregnancy is to track fetal growth in order to identify any problems. Due to inherent characteristics of the fetus, automatic fetal head segmentation and biometric assessment of HC (Head Circumference) from ultrasound pictures are regarded as tough challenges. The study suggests GAC Net (Graph Attention Convolutional Network), a novel Convolutional Neural Network (CNN) intended to address the aforementioned problems. In order to enhance communication between the encoder and decoder and lessen the effect of defects in ultrasound picture quality on HC measurement, it integrates a Graph Convolutional Network (GCN) module. To improve border area detection performance of the network, a novel attention mechanism is presented. The HC18 dataset of fetal head ultrasound images was used for the experiments.

6. ECG Arrhythmia Classification Through Synergy of Pattern and Machine Learning with Fisher’s Criterion

   K. Chandrahasa Reddy, Y. Pavan Kumar Reddy, Y. Sunanda, K. Shankar, M. Ravi Kishore

   Abstract

   Using electrocardiogram (ECG) data, this study presents an algorithmic framework for identifying cardiac arrhythmias, specifically classifying Normal, Ventricular Tachycardia, and Fibrosis. The algorithm consists of several steps, such as preprocessing the data, visualizing the data using a histogram, and using Gaussian parameters to model data distributions. Finally, Bhattacharyya distance computations are performed to evaluate the degree of class separation. The system uses a variety of methods in the classification phase, including a 1-D convolutional deep residual neural networks, Principal Component Analysis (PCA) for reduction of size, Linear Discriminant Function (LDF), and K-Nearest Neighbhours (K-NN). The study is noteworthy for examining comparative Receiver Operating Characteristic (ROC) curves, which emphasize the performance of the approach with particular attention to sensitivity and specificity. This all-inclusive method integrates machine learning, statistical analysis, and visualization techniques in advancing the efficacy of arrhythmia diagnosis.

7. An Approach to Detect Copy Move Forgery Using Deep Learning Techniques

   O. Obulesu, Vaishnavi Pujala, Yakshitha Koulampeta, M. Mahendra

   Abstract

   The increasing sophistication of image manipulation techniques in the digital era has raised urgent concerns about the authenticity and reliability of digital images, particularly in sensitive fields such as journalism, legal investigations, and digital content verification. One of the most prevalent forms of manipulation is copy-move forgery, where a region of an image is copied and pasted elsewhere within the same image to conceal or duplicate content. In response to this challenge, this study proposes a novel deep learning-based methodology that integrates multiscale feature extraction, DBSCAN-based superpixel segmentation, and depth reconstruction to effectively detect copy-move forgeries. The proposed system utilizes the VGGNet-16 architecture for robust feature extraction and combines it with spatial depth cues to uncover subtle inconsistencies introduced during tampering.

   The key contributions of this work include the integration of density-based clustering for localized forgery region identification, the application of depth estimation to improve spatial analysis, and a modular framework that enhances both detection accuracy and computational efficiency. Experimental evaluation conducted on the CASIAv2 dataset demonstrates that the proposed approach achieves superior performance compared to state-of-the-art methods, with notable improvements in precision, recall, and F1-score—even in the presence of post-processing noise and geometric transformations. Additionally, the system maintains low computational overhead, making it suitable for real-time or large-scale forensic applications. This work lays the foundation for extending the methodology to detect other complex manipulations such as image splicing and multiple cloning, offering a scalable and reliable solution to uphold digital image integrity in an increasingly manipulated media landscape.

8. Comprehensive Review and Analysis of Formal Verification Methods for Smart Contracts

   G. Sowmya, R. Sridevi, K. S. Sadasiva Rao

   Abstract

   Smart contracts, fundamental to blockchain-based decentralized applications, require rigorous correctness and security assurance due to their immutable and self-executing nature. This review paper presents a comprehensive analysis of formal verification techniques for smart contracts, including static analysis, model checking, theorem proving, and symbolic execution. We analyze state-of-the-art tools such as Coq, Mythril, KEVM, and others, comparing their capabilities and limitations. Key challenges in formal verification are identified, including scalability, vulnerability coverage, and tool integration. We also detail real-world incidents such as the DAO Hack and Parity Wallet bug to demonstrate the necessity of these techniques. The paper concludes by highlighting current research gaps and emphasizing the critical need for standardized, scalable, and automated verification frameworks for smart contracts.

9. Design of Battery Charger by Using Interleaved Boost Type PFC and Phase-Shifted Full Bridge Converter

   Ramu Bhukya, T. Sravani, D. N. Apoorva, R. Lasya Priya, M. N. S. Suchithra, S. Deepthi

   Abstract

   Electric vehicles are gaining demand nowadays, as conventional fossil fuel vehicles cause environmental pollution and greenhouse gas emissions. As the automotive industry progresses toward a sustainable future, the use of electric vehicles (EVs) has become essential for cutting greenhouse gas emissions and dependence on fossil fuels. This project explores the critical importance of electric vehicle battery charging infrastructure. It highlights the significance of advancements in charging technology, including fast charging capabilities and smart grid integration, in addressing range anxiety, enhancing user convenience, and maximizing the overall viability of EVs. In this manuscript a comprehensive study centered on the modeling and construction of a battery charger integrating with interleaved boost converter (IBC) & using phase-shifted full-bridge converter (PSFB) topologies is presented. This proposed charger aims to enhance efficiency, power quality, and reliability in charging batteries for electric vehicles. Interleaving techniques are applied in the boost converter to adjust power factor and the full bridge converter for efficient voltage transformation. The output result shows the DC-DC conversion of power to enhance the battery charging through simulation.

10. Machine Learning in Cybersecurity: Techniques and Challenges

    Janga Prasad, E. Aparna, K. Mounika, Md. Shabaz khan, B. Ravikumar, L. Suneel

    Abstract

    - When it comes to providing appropriate protection against sophisticated assaults, standard cybersecurity methods often fall short in the constantly shifting environment of cyber threats. This article investigates the use of Machine Learning (ML) strategies in the field of cybersecurity with the goal of improving capabilities in the areas of threat detection, prevention, and response capacity. We address major machine learning methodologies, such as supervised learning, unsupervised learning, reinforcement learning, and deep learning, as well as their applications in areas such as the detection of intrusions, the analysis of malware, the detection of phishing, and the prevention of fraud. The obstacles of adopting machine learning in cybersecurity are also discussed in the research. These issues include challenges related to data quality, model performance, and interpretability. Through the use of machine learning, cybersecurity systems are able to reach improved levels of accuracy, flexibility, and resilience, therefore providing a significant defence against new threats.

11. Smart Garbage Segregation and Management Using IoT with Solar Cell

    Nirudi Santosh Kumar, Patturi Sai Priya, Nalla Vishnu Teja, Puppala Prem Kumar

    Abstract

    In populous areas, effective waste management is a critical challenge which is worsened by growing population and limited resources. This paper introduces a pioneering prototype aimed at revolutionizing garbage management through the integration of innovative advancements such as IoT and solar power. The prototype addresses key aspects of waste management, including segregation of dry and wet waste, real-time monitoring of garbage levels, and a rain-triggered lid opening mechanism. By harnessing the power of IoT, the system can automatically notify consent authorities when garbage levels exceed a certain threshold, facilitating timely intervention and efficient waste collection.

    This paper represents a significant advancement in sustainable waste management practices, offering a promising solution to the pervasive issue of improper waste disposal in urban environments. By promoting automated segregation and monitoring processes, the prototype streamlines waste management operations, reducing the manual monitoring and enhancing overall efficiency. Through the integration of IoT and solar power, the prototype exemplifies the ability of intelligent solutions to revolutionize conventional waste management methods, leading to cleaner and healthier urban environments.

12. PCA and GIS Analysis of Groundwater Quality Over Yadadri

    Nannaparaju Vasudha, Polisetty Venkateswara Rao

    Abstract

    Groundwater is one of the world's most vital resources. It is commonly utilized in agriculture in India and provides drinking water to millions of people, particularly those living in rural regions. Pre- and post-monsoon data were gathered from the government's publicly available website for the years 2021 and 2022. The data includes twelve polluting components collected at 32 locations throughout Telangana's Yadadri Bhuvanagiri district. The data was subjected to principal component analysis (PCA) to better understand the interrelationships between physiochemical components in groundwater, as well as to investigate the contribution of each component to the Water Quality Index (WQI). The Geographic Information System (GIS) was used to better comprehend and display the relationship. TDS, Cl, SO4, Na, Ca, Mg, and TH created Principal Component 1, which accounted for roughly 50% of the variation in WQI using PCA analysis. Inverse Distance Weighted (IDW) interpolated GIS values corresponding to Principal components showed the same levels in different locations of the research area and were consistent with WQI levels.

13. Analysis of the Medicine Learning Integrated Executive Support System for Clinical Prediction Using AI and ML

    T. Bhaskar, S. Kirubakaran, Aelgani Vivekanand, Voruganti Naresh Kumar, D. Maneiah, Vandhanapu Srinu

    Abstract

    This study suggests that data-driven technology is currently affecting professional decisions in health care through making forecasts or giving advice. In recent clinical research, there are a number of examples of how machine learning can be used, especially for prediction of outcomes algorithms. These effects can be everything from death and sudden cardiac arrest to irregular heartbeats and serious kidney damage. In this piece, we describe a way to make medical decisions when there isn't enough information. Its main building block is ontology-based automatic logic, and machine learning methods are added to improve the working models of patient records in order to deal with the problem of data that is missing. In this article, we give a summary of the most recent findings on predicting outcomes from associated research that look at data processing, inference, and model evaluation models made with data taken from electronic health records. In short, we show that machine learning has the potential to help with a job that is very important to medical professionals. This is done by dealing with lost or noise patient data and making it possible to use several clinical information’s.

14. DiabeXpert: A Comparative Machine Learning Framework for Diabetes Prediction

    Jacinth Peyyala Satvik, Chidurala Sanjana, Budigem Dhuhitha, Lingam Sunitha

    Abstract

    The central aim of this research encompassed the development and implementation of a machine learning based method for predicting diabetes, alongside an exploration of effective strategies to ensure its success. Diabetes ranks among the most severe global health conditions, resulting from a confluence of variables, including high blood sugar, obesity, and other causes. It causes improper metabolism and high blood sugar levels by interfering with the activity of the insulin hormone. The main objective of this program is to predict possible cases in order to reduce the risk of diabetes and encouraging individuals to adopt healthier dietary and lifestyle choices in the future by creating and implementing a machine learning based diabetes prediction using a diverse range of algorithms, including KNN,SVC, DT, RF, and GBC.

15. Detecting the Impact and Side Effects of Medicine on the Human Body Using Machine Learning Algorithms

    M. Rishith Reddy, K. Akshith Reddy, M. Vigneshwar reddy, Nirav Bhatt, Yugandhar Manchala, Nirmal Keshari Swain

    Abstract

    This study investigates predicting the adverse side effects of a drug holds paramount importance due to the potential impact it can have on a patient's well-being. When a treatment extends its effects beyond the intended cure for a particular ailment, it can lead to the emergence of side effects. These side effects can vary in intensity, ranging from mild discomfort to severe, and in the worst-case scenarios, they can even prove fatal. Treatments can take various forms, including medications, surgical procedures, and alternative therapies, all of which can induce these unintended reactions. To ensure that patients can fully reap the benefits of long-term treatments, it is imperative to educate them about the possibility of encountering adverse events and to provide precautionary instructions before recommencing the treatment. It is noteworthy that many individuals on daily medication regimens experience adverse side effects, often due to factors such as the introduction of new drugs or adjustments in dosage. These side effects can manifest with varying degrees of severity and implications for the patient’s overall health. While physicians can anticipate some of these adverse effects based on their knowledge and past patient experiences, there still is a realm of unknown possibilities. The objective here is to empower medical professionals to better foresee and prepare for the negative side effects that may arise because of the drugs they prescribe. Achieving this aim can be facilitated through the application of artificial intelligence, which has the potential to enhance our ability to predict and manage these unintended reactions more effectively. So, In this study we are proposing a Machine Learning (ML) model such as Logistic Regression (LR), K-Nearest Neighbor (KNN) and Random Forest (RF) to predict the various side effects caused by various drugs on the human body.

16. Botnet Detection Through Machine Learning: A Stacking Ensemble Model Approach

    A. Rahul Reddy, B. Rohith Reddy, K. Sankar Sai Kumar Reddy, Nirmal Keshari Swain, Yugandhar Manchala

    Abstract

    As cyber threats evolve, detecting botnet activities remains a critical challenge for digital ecosystem security. This research thoroughly investigates botnet detection techniques, introducing a novel stack-based ensemble model. Utilizing decision trees, random forests, XGBoost, and LightGBM as base learners, with a decision tree as the meta-learner, the ensemble demonstrates promising results in enhancing accuracy and resilience. The Stacking Ensemble Model achieved a high accuracy of 99.2%, with corresponding recall and F1 score values of 99.2% and 98.2%, respectively.

17. Design and Development of CanSat for Air Quality Monitoring with Advanced Node Communication Enhancement Strategy (ANCES)

    Hridhya Mehta, Aparna Taduri, Shyamala Merugu, Rakshit Raj Mandari, Vaishnavi Chavali, Narayana Mala, Vishwanath Kumar Panangipalli

    Abstract

    This study describes a new communication approach known as the advanced node communication enhancement strategy. Traditional communication methodologies used for competitors’ Cansat missions frequently have limits in dependability and range. ANCES improves communication at the node level, resulting in powerful communication. This article discusses the competition structure, the components chosen for our Cansat, and the new communication technique, ANCES. This technique is beneficial, as illustrated in the research because it increases competition reward points depending on communication dependability and range. It also describes the design and assembly of the Cansat that we built to monitor air quality, including the components, fabrication, and graphical user interface design.

18. Detection of Phishing Websites Using Machine Learning

    S. Mounasri, D. Bhargavi, K. Nikhitha, M. Akshitha

    Abstract

    Phishing remains a prevalent and evolving security threat, posing serious risks to both individuals and targeted brands. Despite its longstanding presence, phishing attacks persist as active and successful endeavors, with attackers continuously refining tactics to enhance their effectiveness. Detecting phishing websites is crucial in mitigating these threats. This paper provides an overview of the importance of such detection mechanisms and delves into the latest advancements in the area of study. Three primary kinds of detection methodologies are examined: list-based, similarity-based, and machine learning-based methods. The study reviews detection methodologies introduced in existing studies, alongside the data collections employed for its evaluation. By exploring these approaches and datasets, this research aims to contribute to a deeper understanding of phishing detection techniques, facilitating the development of more robust and effective countermeasures against this persistent cybersecurity menace.