International Journal of Communication Networks and Information Security (IJCNIS)

Improving performance of consensus protocol in blockchain applications using "proof of reputation" method

Layla Nadhim Naser, Behrouz Tousi, Mohammad Farhadi-Kangarlu — 2024-11-13

Blockchain technology enables transactions among parties who might not completely trust one another. It employs consensus algorithms within its networks to guarantee dependability, despite some nodes being unreliable. Blockchain, a technology that provides a safe, immutable distributed ledger, has facilitated the development of solutions for trust issues in decentralized networks. While primarily recognized for its role in cryptocurrencies, blockchain's scope extends to protecting other elements, like reputations. However, the majority of existing studies on enhancing reputation systems via blockchain are limited to cryptocurrencies and are plagued by inefficiencies and high energy demands. We introduce a new reputation-based consensus method called "proof of reputation," which effectively upholds the reliability and integrity of transaction outputs. This method is utilized in permissioned blockchains that include a management layer for node access. Here, participants authenticate one another's credentials using asymmetric cryptography. This system, unlike Bitcoin’s proof of work, is more economical by removing the need for mining and hash calculations in block validations. The encrypted nature of blockchain also enhances the security and trustworthiness of our approach. We have created a prototype, and our evaluations show that this protocol meets expected scalability, bandwidth, and throughput requirements of the reputation algorithm. In the future, we aim to expand our study to create an all-encompassing reputation-based framework that integrates access control, identity verification, and further security features.

Learning an efficient text augmentation strategy and DQN and XLNet to improve sentiment analysis applications

Hussein Ali Salman, MortezaValizadeh, Vahid Talavat — 2024-11-13

Modern machine learning models like deep neural networks contain numerous parameters and thus necessitate large labeled datasets for training. Often, there is a shortage of labeled data in many scenarios, leading to overfitting. A popular solution to this scarcity is data augmentation, which involves expanding training datasets by altering data points in a way that retains their class labels. This method assumes that additional information can be leveraged from the existing dataset through modifications, which enhance the training set size either by altering data forms or by oversampling. Data augmentation techniques enhance the diversity of training data without the need to gather new data and have become essential in boosting the performance of deep neural networks across various areas, including computer vision. However, the application of these techniques in natural language processing has been limited due to its complexity. In our study, we employ the DQN method, a blend of reinforcement learning and deep learning for augmenting text data. We used various neural networks like CNN, Bi-LSTM+Attention, Transformer, BERT, and XLNet. Our tests on the SemEval dataset showed improvements across most networks, with the combination of DQN and XLNet achieving the highest accuracy at 66%.

Performance Evaluation and Optimization of Software-Defined Networking in Large-Scale Computer Networks

2024-11-13

Traditional network structures could not meet the increasing needs of large organizations, infrastructure developers, and end-users in terms of load balancing. This need is addressed by the development of software-defined networks (SDNs) and by enhancing their efficiency in avoiding congestion, which leads to improved load balancing on the network. The BalCon algorithm is a heuristic algorithm in this field that is capable of detecting and solving the load balancing problem by effectively migrating switches to other controllers.In this study, based on the variables affecting the performance of software controllers, the Invasive Weed Optimization (IWO) algorithm is used to determine an improved solution for the migration of switches to the appropriate controller. The input parameters of the optimization algorithm include the number of migrations required to achieve the optimal arrangement, the maximum load of the controllers, the correlation index of the controller switches, and the load variance. In this research, the performance evaluation variables were the load variance, total and maximum load of controllers, and the number of migrations of switches from one controller to another. the proposed method leads to a better load balancing and smaller number of required migrations between controllers compared to the base design.

Software Defect Prediction Using an Intelligent Ensemble-Based Model

Dr K Subba Reddy, Badam Rajesh — 2024-11-13

Predicting software defects is essential for improving software quality and lowering testing costs. Finding and sending only faulty modules to the testing step is its main goal. An intelligent ensemble-based software defect prediction model that integrates several classifiers is presented in this study. To identify faulty modules, the suggested approach uses a two-step prediction procedure. Four supervised machine learning algorithms—Random Forest, Support Vector Machine, Naïve Bayes, and Artificial Neural Network—are used in the first phase. To get the best accuracy possible, these algorithms are tuned via iterative parameter optimization. The final predictions are made in the second step by combining the predictive accuracy of each classifier into a voting ensemble. The precision and dependability of the defect forecasts are significantly enhanced by this ensemble technique. To construct and assess the suggested defect prediction system, seven historical defect datasets—CM1, JM1, MC2, MW1, PC1, PC3, and PC4—from the NASA MDP repository were used. The findings show that each dataset's suggested intelligent system outperformed twenty cutting-edge defect prediction approaches, such as ensemble methods and base classifiers, with amazing accuracy.

ENHANCED FLOOD FORECASTING MODEL USING CNN

K. Mohana Rao, Kota Venkata Supraja — 2024-11-13

One of the most frequent natural catastrophes that causes significant harm to property, crops, the economy, and human lives is flooding. For scholars who have been trying to forecast floods for a long time, flood prediction presents a significant difficulty. This article proposes a flood forecasting model that makes use of the federated learning approach. By preventing data from being exchanged over the network for model training, Federated Learning, the most sophisticated machine learning (ML) approach, addresses network latency issues that arise in flood prediction while guaranteeing data privacy, availability, and security. Instead of transferring large data sets to a central server for local model aggregation and global data model training at the central server, the Federated Learning approach encourages onsite training of local data models and concentrates on transmitting these local models across the network. The suggested model in this article combines locally trained models of eighteen customers, determines which station is likely to experience flooding, and provides a flood alarm with a five-day lead time for a particular client. At the client station where the flood is anticipated, a local feed forward neural network (FFNN) model is developed. The local FFNN model's flood forecasting module uses a number of regional factors as input to estimate the anticipated water level. From 2015 to 2021, a dataset of five distinct rivers and barrages was gathered, taking into account four factors: hydrodynamics, flow routing, rainfall-runoff, and snow melting. The suggested flood forecasting model has an 84% accuracy rate in predicting past floods that occurred in the chosen area between 2010 and 2015.

FRAUDAUDITOR A VISUAL ANALYTICS APPROACH FOR COLLUSIVE FRAUD IN HEALTH INSURANCE

M Malakondrayudu, Shaik Neeha — 2024-11-13

The functioning of the healthcare system is threatened by collusive fraud, in which many con artists band together to steal money from health insurance. However, owing to the absence of labeled data and the great resemblance of fraudulent actions to routine medical visits, current statistical and machine learning-based algorithms are restricted in their capacity to identify fraud in the context of health insurance. Expert knowledge must be included into the fraud detection process in order to guarantee the accuracy of the detection findings. We present FraudAuditor, a three-stage visual analytics method for collusive fraud detection in health insurance, developed in close collaboration with audit specialists in the field. In particular, in order to represent the visit connections of various patients holistically, we initially let users build a co-visit network interactively. Second, suspected fraudulent groupings are identified using an enhanced community identification algorithm that takes the level of fraud possibility into account. Lastly, with customized visualizations that accommodate various time scales, users may compare, examine, and validate questionable patient behavior using our visual interface. In order to identify the true fraud group and rule out the false positive group, we performed case studies in a real-world healthcare setting. The outcomes and professional opinions demonstrated the method's efficacy and practicality.

Propounding First Artificial Intelligence Approach for Predicting Robbery Behavior Potential in an Indoor Security Camera

S Sreenivasulu, Gonugunta venkata Bharathkumar — 2024-11-13

For video surveillance systems to avoid incidents and safeguard assets, crime prediction is necessary. In this regard, our paper suggests the first artificial intelligence method for predicting and detecting Robbery Behavior Potential (RBP) in an interior camera. Three detection modules—head cover, crowd, and loitering detection modules—are the foundation of our approach, which enables prompt response and deters robberies. The YOLOV5 model is retrained using the manually annotated dataset we collected to create the first two modules. Furthermore, using the Deep SORT algorithm, we provide a new definition for the loitering detection module. After converting expert information into rules, a fuzzy inference system makes a final determination about the likelihood of robbery. The robber's various style, the security camera's variable viewpoint, and the poor quality of the video photos make this tedious. We successfully completed our experiment using actual video surveillance footage, achieving an F1-score of 0.537. Therefore, we design a threshold value for RBP to assess video pictures as a robbery detection issue in order to do an experimental comparison with the other relevant research. Assuming this, the experimental findings clearly reflect an F1-score of 0.607, indicating that the suggested approach performs much better in identifying the heist than the robbery detection methods. We firmly think that by anticipating and averting robbery incidents, the implementation of the suggested strategy might reduce the harm caused by robberies in a security camera control center. However, the human operator's situational awareness improves and additional cameras may be controlled.

Machine Learning Approach to Classification of Online Users by Exploiting Information Seeking Behavior

Dr. K.V.Srinivasa Rao, Pakala Yedukondalu — 2024-11-13

The internet is overflowing with unfiltered, impromptu, and continuous material from many sources due to technology in today's environment. In order to successfully deliver information depending on user intent, complex algorithms are created. Users' online experiences include a number of information-seeking activities, such as sharing, searching, and information verification. However, a thorough investigation of this complex user behavior is still pending. This study helps to identify different types of users based on their online engagement, propose a user intent-machine learning model for classifying users based on their online search, share, and verification behavior, and show that dynamic online interactions can be categorized based on their searching, sharing, and verifying behavior. Participants from a wide range of age, gender, and vocational backgrounds complete a questionnaire designed to collect user input on online behavior and practices. After a thorough feature engineering process, K-Mean clustering is used to discover user intent classes or profiles and their attributes based on the key features. To identify these classes, data is subsequently used to train a supervised learning Linear Discriminant Analysis Classifier (LDAC). With 80% accuracy, the suggested framework was able to predict the user intent class. A second user study collects data on users' dynamic interactions, which is used to test the model further. The user profiles that emerge from clustering are used by human raters to label the information search, exchange, and verify activity data after it has been converted to suit the model. While the model predicted the user with 67% accuracy, the study obtains an Inter-rater reliability (IRR) of 60%. According to this study, a user's motivation for looking for information, their propensity to share information on social media, and their propensity to believe information to be reliable can all help to understand their intentions, spot behavioral patterns, and identify intent through dynamic interactions that can be utilized for search engine optimization and targeted marketing.

Novel convolutional neural network structure for Diagnosing Autism Spectrum Disorder by Deep Learning

Entihaa Nsaif Jasim, Behrouz Tousi, Mohammad Farhadi-Kangarlu — 2024-11-14

Purpose: Autism spectrum disorder (ASD) affects approximately 1% of the population and is characterized by restricted and repetitive behaviors. Early diagnosis of ASD has been shown to improve outcomes and management of the disorder. Machine learning, which leverages patterns and common features, offers a promising approach for diagnosing ASD through brain imaging and distinguishing patients from control groups. This study aims to identify early indicators of ASD and reduce diagnostic complexity, utilizing the ABIDE dataset.

Method: A supervised learning model was applied to 3D brain images using the TensorFlow library. The model was trained on 1,300 samples with a learning rate of 0.0001 and a dropout rate of 0.5. Additionally, 120 images, representing 1% of the total data, were used for testing. The model comprised three convolutional layers, each using the ReLU activation function, followed by max-pooling layers to isolate active brain regions at rest. After the final max-pooling layer, a fully connected layer was used for classification, and a dense layer with a dropout rate was employed to eliminate unnecessary nodes. A final fully connected layer, utilizing the SoftMax function, was used for prediction and decision-making.

Findings and Conclusion: The model was trained 50 times using the entire test dataset with random shuffle algorithms. Results showed an increase in accuracy from 60% to 100%, and a reduction in the cost function from 36% to 20%, achieving an average prediction accuracy of 80%.

Collaborative Approaches in Data Engineering and Analytics

2024-11-18

This research paper explores the evolving landscape of collaborative approaches in data engineering and analytics. As the volume, variety, and velocity of data continue to grow exponentially, organizations are increasingly recognizing the need for more efficient and effective methods of data management and analysis. This study examines various collaborative strategies and tools that have emerged to address these challenges, focusing on their impact on productivity, innovation, and decision-making processes. Through a comprehensive literature review, case studies, and empirical analysis, we investigate the benefits and challenges of collaborative data engineering and analytics approaches. Our findings suggest that these collaborative methods significantly enhance the efficiency of data processing, improve the quality of insights derived from analytics, and foster a more agile and responsive organizational culture. However, we also identify potential pitfalls and offer recommendations for successful implementation of collaborative data practices.

Optimizing SQL for BI in Diverse Engineering Fields

2024-11-18

This research paper explores the optimization of Structured Query Language (SQL) for Business Intelligence (BI) applications across various engineering disciplines. The study investigates the challenges faced by engineers in different fields when using SQL for data analysis and decision-making processes. Through a comprehensive analysis of existing literature, case studies, and original research, this paper proposes novel approaches to enhance SQL performance in BI contexts specific to engineering domains. The research findings demonstrate significant improvements in query execution times, data processing efficiency, and overall system performance when implementing the proposed optimization techniques. The paper concludes with recommendations for future research and practical applications of optimized SQL in engineering BI systems.

Presenting an optimal method for computational resource management of computational cloud data centers using a load balance approach and optimization of energy consumption

2024-11-18

Cloud computing technology has attracted the attention of many researchers, organizations, governments, etc. In cloud computing, load balance, resource allocation management, and scheduling execution of tasks are among the most important challenges. In a data center, there are still confrontations for load balancing, scheduling tasks, and resource allocation at the level of hosts and VMs. Considering the challenges in cloud computing,this article uses an advanced Cat Swarm Optimization (CSO) algorithm to allocate resources to virtual machines while maintaining a load balance. The mentioned method has several stages including initialization of the algorithm and cloud computing, determining the number of virtual machines and number of tasks, implementing the advanced cat swarm optimization algorithm, allocating resources, and scheduling tasks while maintaining the load balance of virtual machines. Through simulation of the method presented here, it was observed that the extent of improvement of the advanced cat swarm optimization algorithm is considerable in allocating resources and maintaining load balance between virtual machines compared to other methods based on the criteria of execution time, response time, and the number of migrated tasks.

In this paper, we have investigated different criteria for evaluating the extent of efficiency of the advanced cat swarm optimization algorithm. The obtained results indicated that the execution time of the advanced cat swarm algorithm with the number of virtual machines of 5 to 40 and 40 tasks is0.2084s on average, while for the best-performed algorithm(HBB-LB), we know at the time of writing this paperis0.4004s. Accordingly, the extent of improvement ratio of the proposed algorithm compared to the former best method is1.92. Further, the average response time to the requests in the advanced cat swarm algorithm with 5 to 40 virtual machines and 40 tasks is 0.2624s. while for the HBB-LBalgorithm, the response time is0.4128s on average. Based on the results obtained from implementing the suggested algorithm, it is observed that the number of tasks when the advanced cat swarm algorithm is implemented with 5 to 40 virtual machines and 40 tasks is 3 tasks on average. On the other hand, usingHBB-LB the number of migrated tasks is about 5.6on average, showing that the extent of improvement by the proposed method is good.

Special Issue of Proceedings of Development Engineering Conference Center

DECC,Tehran, Iran — 2024-11-18

This Proceedings contains of selected abstracts of Development Engineering Conference that hold in Tehran, Iran.

The Power of Big Data Analytics: How Organizations Are Leveraging Data-Driven Insights to Make Smarter Decisions and Drive Innovation

Dr V. Sudarsan Rao — 2024-11-21

This research aims at determining the role that organizations play in the utilization of big data analysis to improve decision making ad foster innovation in different sectors. Therefore, this study examines how the quantitative methods including clustering, regression, and neural network analyses as well as decision tree support the use of big data in improving the operational and strategic performance in organisations. As can be observed from the results obtained the use of the neural networks yielded a high accuracy of 92% compared to that of the decision trees with 87% and regression model with 85% accuracy. Traditional decision-making techniques and the clustering algorithms performed well for segmenting customer data with an accuracy of 89 %, enabling targeted marketing communication. To support this argument, this research concerns itself with a comparison with studies that have been conducted before to identify that the application of big data analytics enables the redesign of decision-making processes and the promotion of innovation. However, the study also reveals some of the limitations, including data privacy issues and shortage of experienced people to work on big data analytics that ought to be dealt with to realize the full potential of big data analytics. They lead to insistence on establishment of data literate organizations and technological support systems to support organizational competiveness in current world.

E-Store Power Efficiency: IoT Solutions for Advanced Electrical Energy Management in Retail

2024-11-22

Every day, hundreds of customers visit a shopping mall to make many purchases. Shopping malls are growing quickly these days since everything you need is readily available in one spot, including groceries, clothing, fruits, vegetables, and other products. To gather the things at the mall, a trolley is needed. Pulling the trolley forward or backward is necessary for gathering the things. Following product purchase, customers must wait in a lengthy line to get their purchases billed. We are creating an intelligent shopping trolley to solve this issue. Every single item in the mall has an RFID tag, and the trolley has an RFID reader installed. The item and its quantity are displayed on the LCD installed in the cart when the consumer places the product there. The reader scans the tag. When the customer has finished shopping, they push the finish button on the trolley, which is carry to the bill counter computer and shows the total amount due on the LCD. The customer's entire bill will be delivered as a message.

LSTM Autoencoder-based Ransomware detection using deep anomaly detection

Aqeel Jameel Kaduam, Morteza Valizadeh, Mehdi Chehel Amirani — 2024-11-26

In recent years, with the development of various encryption methods and non-recoverable payments, ransomware has become one of the most critical cyber threats, targeting a wide range of victims. Consequently, ransomware detection has become a significant research field among malware detection researchers. The study of ransomware detection mechanisms and approaches shows that it is impossible to detect unknown ransomware early using existing methods. To address this challenge, this study proposes an in-depth learning method for the early detection of ransomware. In the proposed method, the LSTM Autoencoder neural network is used to model the sequence of input/output request (IRP) packets of normal processes. Processes created outside the model are classified as anomalous processes.For this purpose, several attributes are first extracted from the events recorded by the IRPLogger program. The extracted features are then used to train the model with the LSTM Autoencoder network. Finally, by determining a threshold value, the model training obtained from normal processes is used to detect anomalies. To evaluate the proposed method, datasets including 15 common samples (220 anomalous processes) and 30 common ransomware samples have been generated. The experimental results show that the proposed method, for 5000 initial input/output request packets of each process, has a true positive rate of 97.8% and a false positive rate of 2.1%, making it suitable for the early detection of unknown ransomware.

Scheduling Tasks in Fog Environment: A Competitive Analyses of Greedy, FCFS, CNN Algorithms

Wurood AL-Shadood, Mohsen Nickray, Salam Alyasseri — 2024-11-23

CNN)Convolutional Neural Network(is found to be effective for scheduling text files in a fog computing environment than both the greedy algorithm and FCFS) First come first served). From the results tabulated and analyzed above, the proposed algorithm has achieved enhanced accuracy in both the allocation of task and utilization of the available resources resulting to an improved system performance. Despite the fact that using CNN may take Suitable time to train and to process, the advantages that it brings such as increase in the accuracy of the outcomes and the alleviation of potential bottlenecks of the fog computing system makes this method suitable for use in applications that emphasizes on large and complex text data. Of course, CNN is the best fit in terms of accuracy/speed/resource trade-off and, therefore, CNN would be preferred in the future for disseminating tasks in new-generation fog systems.

Skin Cancer Prediction Model using Deep Learning

2024-11-23

Abstract—This research uses the ISIC 2024 dataset to classify skin lesions using deep learning and convolutional neuralnetworks (CNNs). Skin cancer is one of the most common healthproblems in the world, and successful treatment depends onearly detection. A CNN model that could identify the seven different forms of skin lesions—including dermatofibroma, actinickeratoses, vascular lesions, basal cell carcinoma, squamous cellcarcinoma, melanocytic nevi, and melanoma—was presented in

this work. Oversampling and data augmentation were employedto address the class imbalance.As a result, the constructed modelperformed exceptionally well on its categorization tasks. The model should ideally allow clinicians to promptly identify realskin malignancies because it was tested on a different testset to validate the results. The study suggests that AI-basedinterventions may improve test accuracy and streamline thescreening process if used in medical practice. Betterpatient-oriented results in dermatological treatment should followfrom this.

A REVIEW ON MACHINE LEARNING MODELS FOR QUALITY OF SERVICE IN CLOUD COMPUTING

Nisha, Dr. Deepak Nandal, Dr. Sunil Kumar Nandal — 2024-11-23

One of the most popular ways that services are now provided via the internet is through cloud computing. Factors including resource contention, fluctuating workloads, and varied user needs make guaranteeing QoS in cloud settings a key task. In this research, we provide a new approach that uses machine learning to improve cloud computing's quality of service. The model uses state-of-the-art machine learning algorithms to assess past data, forecast future workload patterns, and distribute resources dynamically to successfully fulfill quality-of-service criteria. By continuously learning from past performance metrics and user feedback, the proposed model adapts to changing conditions, optimizing resource utilization and improving service reliability. We conduct comprehensive experiments using real-world cloud datasets to evaluate the performance of our model. Results demonstrate significant improvements in key QoS metrics such as response time, throughput, and availability compared to traditional approaches. Our findings highlight the potential of machine learning techniques in addressing QoS challenges in cloud computing, paving the way for more efficient and reliable cloud services.

PREDICTIVE MODELING AND CATEGORIZATION OF CYBER THREATS USING DATA MINING TECHNIQUES

Mohammed Javed Hussain, Dr. Sarwesh P — 2024-11-23

Numerous IoT devices and the quick advancement of technology make cyberspace unsafe, which ultimately results in Significant Cyber Incidents (SCI). One method for preventing SCI in online systems is cyber security. Cyber security relies heavily on Data Mining and Machine Learning (DM-ML) for SCI detection, prevention, and prediction. Pre-pandemic and post-pandemic SCI are the two subgroups of the dataset (SCI according to the Center for Strategic and International Studies (CSIS) study). Well-known machine learning classifiers including Naïve Bayes (NB), Support Vector Machine (SVM), Logistic Regression (LR), and Random Forest (RF) are utilized for classification, while data mining (DM) techniques are employed for feature extraction. One consolidated dataset is maintained using a centralized classifier technique by using contributions from the world's six continents. This research concludes with improved accuracy and the prediction of which form of SCI can occur in which portion of the world by comparing the results of the pre-pandemic and post-pandemic datasets. Asia is anticipated to be the continent most impacted by SCI, and it is determined that SVM and RF are far superior classifiers than others.

EDGE COMPUTING SECURITY: ADVANCED FEATURE SELECTION ADOPTED SUPERVISED LEARNING MODELS FOR REAL-TIME INTRUSION DETECTION

Dr. P. Karunakar Reddy , T. Nalini Devi , Shaik Asiya, R.Swathi — 2024-12-25

India has experienced significant growth in cyberattacks in recent years, with nearly 1.39 million cyber incidents reported in 2022 alone, according to CERT
In (Indian Computer Emergency Response Team). The adoption of digital transformation and IoT devices has made edge computing critical for
decentralized data processing. However, this distributed nature of edge devices makes them vulnerable to cyber threats, especially in real-time environments.
Edge computing enhances data processing at the source, enabling low-latency, real-time applications like autonomous vehicles and smart cities. Ensuring the
security of these decentralized systems is crucial for maintaining trust and performance.To develop an AI-driven intrusion detection system leveraging advanced feature selection techniques and supervised learning models for real time threat detection in edge computing environments. Before machine
learning, traditional intrusion detection relied on rule-based systems, firewalls, and signature-based antivirus software

Comparative Analysis of Advanced Ensemble Learning Techniques for Accurate and Interpretable Heart Disease Prediction

2024-11-24

Cardiovascular diseases (CVDs) are a significant global health challenge, contributing to
millions of deaths annually. Accurate and interpretable prediction of heart disease, a major subset of CVDs, is critical for early diagnosis and effective clinical decision-making. This study
presents a comparative analysis of advanced ensemble learning techniques for heart disease prediction using the UCI Heart Disease dataset, consisting of 920 instances and 14 clinically relevant attributes. Data preprocessing steps, including handling missing values, feature selection, and standard scaling, were employed to enhance data quality. We implemented five classification algorithms—Random Forest, Extra Trees Classifier, AdaBoost Classifier, XGBoost Classifier, and Support Vector Machine (SVM)—to evaluate individual performance. Among these, XGBoost and SVM demonstrated the highest standalone accuracy and precision of 82% and 83%, respectively. To further enhance predictive accuracy, a meta-model ensemble learning approach was adopted. By integrating the predictions of base learners through a majority voting mechanism, the proposed model achieved exceptional performance with an accuracy of 96.43%, precision of 97%, recall of 98%, and F1-score of 96%. These results underscore the potential of ensemble techniques to mitigate individual model errors and improve diagnostic reliability. This research highlights the importance of leveraging ensemble learning for heart disease prediction and its implications for improving clinical outcomes. Future studies may focus on refining these models, incorporating additional datasets, and extending the approach to other medical conditions for broader applicability.

ARTIFICIAL INTELLIGENCE (AI)TOOLS FOR COST OVERRUNS ON CONSTRUCTION PROJECTS

Lisandra Seecharan, Aaron Anil Chadee — 2024-11-27

The application of artificial intelligence (AI) to cost overrun mitigation in construction projects is explored in this literature review, especially in light of the nature of construction environments, which are very complex and dynamic. There are many underlying reasons for cost overruns, such as project mismanagement, design changes, and site conditions not encountered in the design. In contrast, AI provides unique solutions on the basis of predictive analytics, machine learning, and computer vision in order to facilitate cost estimation accuracy and reduce risks, allocate resources more efficiently, and improve decision-making processes. AI-driven models analyze past data to predict possible budget problems so managers can manage cost related risks in an online or proactive way. AI case studies show cost overrun reduction across projects worldwide. The gaps in the current AI applications and the suggested improvement in its scalability, real-time data integration, and domain-specific knowledge in the AI models for construction cost management are reviewed in this paper. These findings reveal the importance of AI becoming part of the construction landscape to bring cost efficiency and enhance project resilience.

A framework for Stakeholders Analysis in Digital Sharing Economy Platforms

2024-12-03

Alongside the growth and evolution of sharing economy platforms, stakeholders management challenges such as platform owners, users, providers, and platform promoters have also increased. Given the diversity of perspectives on the nature and essence of the sharing economy, there is a need to collect and identify important subjects in the field of the sharing economy from a business perspective. This research turns into examining the concept of the sharing economy, the influential components in sharing economy platform, and finally analyzing sharing economy platforms from the stakeholder theory perspective using a systematic literature review method. Finding of this research include reviewing important theories and trends in the sharing economy field, and presenting a framework for analyzing key stakeholders in sharing economy platforms using stakeholder theory. Applying this framework in managing interactions with stakeholders will have a significant impact on the sustainability of the sharing economy platform.

Balancing Privacy and Utility in AI Training: Comparative Analysis of Privacy-Preserving Techniques

Arpita Maheriya, Dr. Shailesh Panchal — 2024-12-04

The growing need for data in Artificial Intelligence (AI) training and analysis presents a significant challenge on safeguarding individual privacy. Conventional approaches frequently risk exposing sensitive information when sharing data. This research paper explores privacy-preserving machine learning techniques, comparing k-anonymity, differential privacy, noise addition, data sampling, and feature selection. Through sensitivity analysis, it was found that while privacy measures can reduce accuracy, careful parameter adjustments can mitigate these effects. Considerations such as computational overhead, communication costs, scalability, and privacy guarantees are crucial. The study emphasizes balancing privacy and model utility and suggests developing advanced techniques that minimize information loss without compromising performance. This work aims to enhance privacy-preserving machine learning by optimizing the balance between privacy, utility, and scalability.Our hypothesis suggests that our security model minimizes data breach risks and unauthorized access, particularly in healthcare, prioritizing patient data privacy. This approach aims to advance responsible AI development by balancing data analysis with privacy protection.

Optimization of employee performance and Efficient Improvement of multispecialty health care Industry using AI

M.Savitha, Dr S. Praveen Kumar — 2024-12-05

Efficient management and optimization of employee performance are essential components in ensuring the smooth operation and delivery of quality healthcare in multispecialty hospitals. Hospitals face several challenges in maintaining operational efficiency, such as balancing workloads, ensuring the optimal use of staff skills, and making timely decisions that can impact patient care and overall hospital functioning. In this context, Artificial Intelligence (AI) can play a transformative role in addressing these issues, ultimately enhancing operational efficiency across various hospital functions. By incorporating AI-driven strategies, hospitals can streamline processes, improve the allocation of resources, and foster a more effective working environment.

This study aims to explore the potential of AI technologies in improving hospital operations by focusing on key areas like performance management, workload distribution, and decision-making. AI-based solutions can help hospitals automate routine tasks, analyze large sets of data, and predict patient needs or staffing requirements in real time. Such automation and predictive analytics not only reduce human error but also allow hospital staff to focus on more complex tasks that require human expertise, ultimately leading to enhanced productivity and service quality.

Additionally, AI can optimize employee performance by identifying skill gaps and ensuring that the right personnel are assigned to appropriate tasks. By matching staff members' skills with their respective duties, hospitals can prevent inefficiencies such as underutilization or overburdening of staff. AI also facilitates better decision-making by providing real-time insights into patient conditions, hospital resources, and staff availability, which can help in making informed, timely decisions that enhance patient care while maintaining operational flow.

A Descriptive Business Model on Integration of Cutting Edge Technologies in Healthcare Industry

M.Savitha, Dr S. Praveen Kumar — 2024-12-05

The integration of cutting-edge technologies such as the Internet of Things (IoT), Big Data, Artificial Intelligence (AI), and Telematics is transforming the healthcare sector. These technologies promise to significantly improve efficiency and effectiveness in healthcare providing more autonomous, data-driven solutions. This paper proposes the business modelling that can serve as a key strategy for successfully advancing the integration of emerging technologies in healthcare. By combining frameworks for AI, IoT, and other communication technologies, a unified system can be established that connects all healthcare service providers. Digital technology is instrumental in streamlining healthcare processes, making life easier for both patients and providers. This paper alsoproposes an architectural framework required for seamless technology of integration, with examples illustrating how these systems can be applied in practice for a remote patient with chronic illness. Health care practice remains limited, with many AI products, IoT devices and smart systems being at the design and develop stage itself. However, automating healthcare functions through these innovations comes with several challenges, including data privacy concerns, infrastructure limitations, investment barriers, and uncertainty around sustainability, reliability, and efficiency. Despite these hurdles, these challenges can be addressed through strategic planning and developing a robust business model. An extended business modelling approach can help healthcare providers navigate the complexities of technology integration by providing clear frameworks for implementation, identifying potential risks, and devising solutions for overcoming obstacles. Furthermore, innovations in healthcare will emerge from a human-centred approach that fully understands the complexities of patient journeys and care pathways. Given the challenges involved, the development of a thoughtful business model is essential to mitigate risks, ensure societal benefits, and improve the return on investment (ROI) for healthcare providers. Ultimately, a well-designed business model and integration strategy can help realize the full potential of these cutting-edge technologies in healthcare, driving both clinical and operational advancements.

ARTIFICIAL INTELLIGENCE BASED CUSTOMER CHURN PREDICTION MODEL FOR POST- GRADUATION STUDENTS IN SAUDI UNIVERSITIES

Hussein Moselhy Syaed Ahmed, Safaa Sayed Mahmoud, Nabil Mohamed Alabsy — 2024-12-08

Despite improvements in undergraduate student retention, the problem of student attrition continues to exist and remains a challenge for the academic institutes, but especially for the post graduate programs where retention is essential for academic success and growth of the institute. It is an Artificial Intelligence (AI) based Churn Prediction model to identify the risk of disengagement or dropout from the postgraduate students of Saudi Universities. The dataset used in the research consists of academic, demographic, behavioral, and financial factors. Finally, we apply some machine learning algorithms like Random Forest and Gradient Boosting to predict churn with very high accuracy. We evaluate the performance of the model using metrics such as precision, recall, and ROC-AUC to validate that the model is able to identify at-risk students. Burke finds that academic performance, attendance pattern, and financial stability have a strong positive effect on churn likelihood. This first model has important implications for proactive interventions by universities administrators, such as targeted support strategies. The study contributes to foster academic excellence and education institutions sustainability by improving retention rates.

Anti-Antiforensic Techniques in Steganalysis: Bifurcation-Based CNN Optimization

Amadeus Pondera Purnacandra, Yudi Prayudi — 2024-12-09

Information technology has become an indispensable need for humans today, from working, learning, trading, to communicating using information technology. Digital images are an asset that cannot be separated from human communication when using information technology. Cyber security is an important aspect that must be considered when using information technology. One method of sending data securely is steganography, where a secret message is embedded in digital media. Steganography itself is one of the anti-forensic techniques, whose main purpose is to hide secret messages to make it difficult during the digital forensic investigation process. As the opposite of Steganography, Steganalysis is one of the digital forensic techniques that focuses on the detection of hidden messages (stego) in digital media. In this research, we propose an innovative steganalysis approach using Convolutional Neural Network (CNN) as a method of steganographic detection. This research aims to develop a reliable steganalysis system in detecting hidden messages with modern steganography techniques. We utilize the strength and ability of CNN in recognizing complex patterns in image data and even extracting the data from specific hiding method. This approach has the novelty of integrating the previous method with a bifurcation architecture that aims to improve the sensitivity and generalization of steganography detection, which is proven to increase the accuracy up to 92.53%.

Towards Efficient DDoS Attack Detection for SDN: A Survey of Hybrid Deep Learning Frameworks

Nawar Jumaah, Asghar Tajoddin, Ahmed Aljhayyish — 2024-12-09

Software-defined networking (SDN) is a revolutionary innovation in the technology of networks ?with a lot of demanded attributes such as manageability and flexibility. SDN shows new risks of ?privacy as well as safety such as distributed denial-of-service (DDoS) attacks.?? ?DDoS attacks ?refer to a basic SDN network threat that leads to strict network performance disruptions. ?Therefore, multiple deep learning (DL) models usage like recurrent neural networks (RNN) and ?convolutional neural networks (CNN), efficiently raises the capability for recognizing ?complicated attacks. In addition, Explainable AI (XAI) methods cooperate with DL models, and ?transparency and interpretability also strengthen trust in security systems. Here, DDoS attack’ ?detection concerns and responses in SDN networks applying multiple architectures given the DL ?were examined. In addition, it provides novel fault-tolerant SDN frameworks that could cope ?with DDoS attacks and guarantee the fixing of the network in crucial conditions. At last, the ?present paper outcomes illustrate that multiple architectures could be strong means to diagnose ?and counter DDoS attacks in SDN networks also show study and improvement demand in the ?present domain for developing new network security systems performance.?

An Efficient Wrapper-KNN Based Feature Selection Method for High Dimensional Data Using Machine Learning on a Ray Framework

Subhash Kamble, Arunalatha J S, Venugopal K R — 2024-12-06

A variety of disciplines have handled large databases that contain
an enormous number of features. While several attempts have
been made to design an optimized model for Feature Selection
(FS) in high dimensional data systems, the difficulty of
processing such data remains a great challenge. As a result, the
high dimensionality of large datasets can hinder the Machine
Learning (ML) process. Feature selection is aimed at reducing
features that are unnecessary, noisy or redundant that may affect
classification. To address this issue Parallelized Exhaustive
Wrapper-Based Feature Selection Technique using a ray
distributed framework is proposed. In this, K-Nearest Neighbour
(KNN) classifier algorithm and Cross Validation (CV) approach is
utilized to deal with overfitting problems to obtain better results.
The experimental outcomes show that the proposed approach
retains the exceptional efficiency of the model while reducing
computation time considerably. The overall average accuracy of
the proposed work is 92.24%, and the execution time taken is
better than as compared with the state-of-art techniques.

FABRICATION & MECHANICAL CHARACTERIZATION OF ALUMINIUM 7075 METAL MATRIX REINFORCED WITH SILICON CARBIDE & GRAPHITE

B. RAMESH, Dr. D.R. SRINIVASAN — 2024-12-15

Metal Matrix Composite (MMCs) had been used in several applications in aerospace and automotive module industries. Although several technical challenges exist with casting technology achieving a uniform distribution of reinforcement within the matrix is a major challenge, that affects the properties and quality of composite. In this work an attempt was made to develop a low cost method used in production of metal matrix composites having homogeneous dispersion of ceramic materials. After fabrication an experimental investigation was carried out to determine the effects of silicon carbide with step size of 2% variation and graphite fraction of 2% is kept constant. The results which obtained from the investigation had analyzed for various parameters to predict the effects of variation of silicon carbide in relation to other constituents. The experimental test comprises of tensile test, compressive test, hardness test to evaluate its mechanical properties.

Methodical Review of Deep Learning Macro and Micro Face Emotion Detection

Meenakshi Kondal, Dr. Sushil Kumar Bansal, Dr. Amandeep Verma Puri — 2024-12-18

Facial expression recognition (FER) has added weighty power in the ground of deep learning (DL) owed to its wide-ranging tenders in human-computer interaction (HCI) mainly within healthcare. psychology and entertainment. The paper presents a well-organized review of FER methodologies leveraging deep learning techniques. We survey and evaluate 56 papers a several range of studies, meeting on their objectives, methods, scope and datasets. The analysis classifies the studies based on the types of expressions used for emotion detection and focus on investigating revisions in the arena of emotion recognition using deep learning actions. Traditional methods, counting feature-based methods, such as geometric and appearance-based techniques, are weighed with current deep learning models like convolutional neural networks (CNN) and generative adversarial networks (GAN). The review judgmentally examines the accuracy, performance and adaptability of these methods across various datasets and situations. It discovers how deep learning has transformed facial expression recognition, offering important growths in handling complex expressions, miscellaneous datasets, and real-time applications. Our study is built on a type the natures of languages used for emotion detection, the proper DL approaches and the datasets dynamic. We organize appearance modules into macro and micro facemask expressions recognition. Whereas many studies remain to employ old-style emotion recognition methods, our efficient works review emphasizes DL approaches. The techniques, such as Convolutional neural networks (CNN), Faster region-based convolutional neural network (R-CNN) and further CNN are beneficial since they abstract reality by identifying outlines in images, allowing them to complete advanced performance in object recognition, feature extraction, and classification which are prevalent in the reviewed literature. We also discover attention mechanisms, region-based CNN object recognition and the broader use of CNN in emotion recognition. By providing a comprehensive overview of state-of-the-art deep learning algorithms and datasets, this review goals to enhance researchers' understanding of the fundamental components and emerging leanings in facial expression recognition.

Unbridling the Power of Deep Learning Methodologies for Early Diagnosis of Skin Diseases: A Critical Review

Ritika Sharma and Sushil Kumar Bansal — 2024-12-23

Skin diseases affect millions of individuals around the globe and it is a serious health problem. Effective management and treatment of skin disorders depend on a prompt and correct diagnosis; however, access to dermatological care is still restricted, especially in underprivileged areas. The study shows the development of various predictive modeling techniques, whichhave completely changed the interpretation of medical images and opened up exciting new possibilities for the automated diagnosis of skin conditions. Deep learning algorithms have demonstrated outstanding efficacy in the analysis of intricate visual data, including photographs related to dermatology. It may also make it possible for people to self-assess skin lesions, enabling prompt referrals and early diagnosis of potential anomalies. This paper underscores the potential influence of deep learning methods such as convolution neural networks (CNN) which have gained high popularity in accomplishing sustainable healthcare objectives while offering a thorough review of their use in the diagnosis of skin conditions. Recurrent neural networks (RNN), and long- short-term memory (LSTM) models are also examined in this paper. The model demonstrates the advancements made in the automated systems that can correctly diagnose a range of skin disorders, such as eczema, dermatitis, psoriasis, and melanoma. The research supports more general goals of guaranteeing universal access to high-quality healthcare and encouraging sustainable growth.

Resource Allocation in Computation Offloading Based on improved two-layer Q-network Reinforcement Learning

Maryam Thabit Hussein, Ehsan Shoja — 2024-12-27

Collaborative edge computing (CEC) and mobile-edge computing (MEC)have recently emerged as a popular approach that allows for the sharing of computational resources among various edge devices. In this model, multiple parties, including mobile users, edge computing servers, and cloud servers, work together to enhance computational capabilities, enabling the efficient handling of computation-heavy tasks at the edge. The paper presents a model where each task and its associated data, whether processed in the cloud or at the edge, is represented as a network of interdependent sub-tasks. A critical issue in CEC is the allocation of resources for offloading tasks, which involves determining the timing and location for executing each subtask. This work establishes a mathematical framework for optimizing resource allocation to reduce the average delay in completing tasks. The MEC network incorporates mobile users, edge computing servers, and cloud servers. Efforts have been made to develop both static and dynamic strategies aimed at minimizing the average computation delay for tasks within the network. The system was modeled as a Queuing Network to derive optimal routing probabilities under a static policy. Additionally, the problem was modeled using a Continuous-Time Markov Chain (CTMC), but due to the vast state space and dynamics of the CTMC, coupled with a reward structure that becomes apparent after several state transitions, Deep Reinforcement Learning algorithms were employed to determine the most effective dynamic policy. While the static policy conserves energy by not requiring system state measurements, the dynamic policy provides superior performance in reducing average delay compared to benchmark algorithms.

Advancing Biometric Security Through AI and ML: A Comprehensive Analysis of Neural Network Architectures for Multimodal Authentication Systems

Venkateswaranaidu Kolluri, Souratn Jain, Manjeet Malaga, Jyotipriya Das — 2024-12-26

Biometric authentication is now a key element has emerged as a cornerstone of modern security systems, which use physiological and behavioral characteristics to identify a person based on fingerprints, face, voice, and iris. However, typical unimodal biometric systems have received a lot of acceptance; they remain liable to problems such as spoofing, variability across environments, and low recognition rate thus limiting their application in safety sensitive activities. This article offers a review of how AI and ML more specifically, neural network architectures, are changing biometric security. Multimodal authentication systems, therefore, present data from multiple biometric traits through the application of complex techniques such as CNNs, RNNs, and GANs for processing, greatly enhancing accuracy, enhanced invulnerability to adversarial attacks. During this study, early fusion, late fusion, hybrid fusion, and deep fusion are evaluated to integrate biometric data to improve security and convenience. Common applications of AI in healthcare, finance, and business security are highlighted as well as potential major issues, including ethics, privacy, algorithms’ bias nature, and computational intensity of AI systems. Overall, the results highlight AI and ML’s role in biometric authentication, focusing on how they can help overcome the challenges inherent in unimodal systems while driving the development of multimodal solutions that are secure, scalable and highly user-friendly. Thus, this research contributes to the history of how biometrics can gradually become wiser, more ethical and stronger to help facilitate the quest to build safe digital environments in a world that is becoming even more connected.

Transforming Decision-Making with Big Data Analytics: Advanced Approaches to Real-Time Insights, Predictive Modeling, and Scalable Data Integration

Srinivas Murri, Manoj Bhoyar, Guru Prasad Selvarajan, Manjeet Malaga — 2024-12-30

Huge progress has been achieved in big data decision-making throughout different industries due to real-time insights and data scalability, complicated models. This article focuses on state-of-art solutions can be adopted by organisations when dealing with very large and complex data volumes in an effort to improve instaneneous decision-making accuracy. Real time data streams can also be incorporated to respond to new trends of operations as well as challenges. Prescriptive modeling also use other tools as machine learning and statistical to try to look forward and this is important in order to take preventive measures. Further, solutions are also implemented to facilitate the efficient integration of large volumes of multi-structured data for improved end-to-end business visibility. Based on the findings of the study the enormous benefits of using big data analytics in enhancing the quality of decisions made, the efficiency of operations conducted and the competitive advantage gained with consideration for the difficulties of managing data and its implementation.

Modern Permissions Management Strategies for Enforcing Least Privilege in Cloud: A Comparative Assessment

Anant Wairagade — 2024-12-30

Cloud computing over the years has gained widespread attention due to features like scalability, cost-effectiveness, and on-demand access to resources. Despite these advantages, it faces challenges associated with security, particularly managing permissions effectively. The principle of least privilege is considered a critical strategy to handle issues like these by minimizing the access required for a certain task. Modern permission management methods such as Role-based access control, attribute-based access control, context-aware access control, time-based access, and identity verification and authentication incorporate this principle to provide efficient methods of allowing permission in the cloud. In this work, we delve into the path of giving comparative assessment of modern permission management method in a cloud environment. We carried out an extensive literature review to assess their performance based on strengths and limitations, evaluation criteria used, and attack it provide robustness against. Moreover, we also conducted a comparative analysis of permission management approaches using the criteria of scalability, complexity, and other attributes. The findings provided us with detailed knowledge regarding the challenges encountered by the existing method in the cloud. Building on these challenges, we proposed future solutions for more robust and adaptable permission management methods in cloud infrastructure.

The Convergence of Blockchain and IoT: A New Paradigm for Secure Connected Devices

2024-12-30

This paper investigates the potential of Blockchain technology to enhance security within the Internet of Things (IoT) networks. By examining the integration of Blockchain's core features—decentralization, immutability, and transparency—it addresses significant IoT security challenges such as data tampering and unauthorized access. The study includes practical case studies from sectors like smart grids and healthcare, highlighting improvements in operational efficiency and data integrity. Despite promising outcomes, challenges such as scalability, resource constraints, and privacy issues persist. The paper concludes with future research directions aiming to optimize Blockchain integration in IoT. This exploration sets a foundation for advancing IoT security through Blockchain technology.

Multi - Security Authenticated Personal Device Access System using IoT

2024-12-31

The main goal of this project is to design and implement a locker security system based on fingerprint and IoT technology for use in banks, offices, and homes.Only authenticated individuals can access the locker, thereby ensuring security for stored documents and valuables. The system workflow begins with a user enrolling a username and password; once verified, the user’s fingerprint is scanned and stored with an ID. If this ID matches, a four-digit code is sent to the authorized user’s mobile device to unlock the locker. This system maintains a log of check-in and checkout activities, along with basic user information, enhancing accountability.

Dynamic Communication Platform for College Campus: Arduino and GSM Based LED Notice Board for Real- Time Updates

2024-12-31

In today's fast-paced world, effective communication is essential for keeping individuals informed and connected. Traditional notice boards often fall short
in meeting the dynamic communication needs of modern environments. To address this challenge, we present an innovative solution: an Arduino and
GSM based LED notice board. This project leverages the power of Arduino microcontrollers and GSM (Global System for Mobile Communications)
technology to create a versatile and interactive communication platform. By combining hardware and wireless communication capabilities, this LED
notice board offers real-time updates and remote-control functionalities,revolutionizing the way information is disseminated in various contexts,
including educational institutions, public spaces, and businesses.

IOT And RFID Based Bus Identification System With Voice Alert For Enhanced Accessibility To Visually Impaired

2024-12-31

A bus detection system using RFID technology that aims to ease the traveling and movement of blind people. The proposed system consists of two detection
subsystems, one on the buses and the other on the bus stations and a website. In the bus detection subsystem, the nearby stations will be easily detected and then announced through a voice message inside the bus . In the bus station subsystem, the coming buses will be detected and then announced in the station in order to alert the blind people. The bus details will be posted on the website using IoT. This system is used to help blind people to travel smoothly and independently from one place to another by providing complete and clear information. Every status of project is monitor in LCD using 16*2 modules. The proposed system is designed using ARDUINO microcontroller using Arduino IDE software. 5V regulated power supply used to control ARDUINO microcontroller.

Design and Fabrication of Serving Robot for Old Age Home

2024-12-31

Restaurants often face challenges in efficiently serving meals to their customers. Limited human staff and the repetitive nature of serving tasks can be time-consuming and exhausting. To address these issues, this project focuses on the development of a serving robot specifically designed for restaurants. The robot incorporates key components such as an Arduino microcontroller, a 30A motor driver, powerful left and right DC motors, a relay module, DC lights, and a 6-CH transmitter. The primary objective is to create a versatile and efficient robot capable of seamlessly delivering food and other items within a restaurant setting,
thereby streamlining processes and reducing the need for manual labor.

MACHINE LEARNING-BASED FAULT DIAGNOSIS IN ENGINES USING IOT BASED THERMODYNAMIC FEATURE VECTORS

Dr. Kishore Kumar, Tadikonda Jyothilatha , Ankitha Sangani , Thota Priyadarshini — 2024-12-02

The integration of Machine Learning (ML) and Internet of Things (IoT) technologies has revolutionized engine fault diagnosis, leveraging real-time data from IoT sensors to enhance predictive maintenance practices. Therefore this project explores the development and implementation of an ML-based fault diagnosis system using thermodynamic feature vector collected from IoT sensors. These feature vectors, including temperature,pressure, and flow rates, provide a comprehensive overview of engine conditions, enabling the detection and prediction of faults with high accuracy. Traditional diagnostic systems, reliant on manual inspections, rule based methods, and Diagnostic Trouble Codes (DTC), often fail to handle the complexity and volume of data generated by modern engines. These methods are not only time-consuming and prone to human error but also limited in their ability to provide proactive maintenance solutions. This project addresses these limitations by employing advanced ML algorithms, such as neural networks, decision trees, and support vector machines, to
analyze and interpret the high-dimensional data from IoT sensors. The proposed system is designed to provide real-time monitoring and predictive
maintenance, significantly reducing downtime and maintenance costs while enhancing engine performance and safety. The ML models are trained on a
diverse dataset of thermodynamic features, enabling them to detect subtle anomalies and predict potential faults before they escalate into major issues.
Through rigorous testing and validation, the system demonstrated superior diagnostic accuracy and adaptability to varying operational conditions. This
project underscores the potential of ML and IoT in transforming engine fault diagnosis, offering a scalable, efficient, and reliable solution for modern
industrial and automotive applications. The findings contribute to the advancement of predictive maintenance strategies, fostering greater
operational reliability and safety in engine management systems.

DEEPFAKE DETECTION ON SOCIAL MEDIA: LEVERAGING DEEP LEARNING AND FASTTEXT EMBEDDINGS FOR IDENTIFYING MACHINE-GENERATED TWEETS

S. Vinod Kumar , Santosh Taruni Annapa Reddy , S. Lakshmi devi , V. Kalyani — 2024-12-25

Deepfake technology, which uses AI to create manipulated media, poses a significant threat to information integrity on social media platforms. In India, the rise of deepfake content has grown exponentially, especially in the political and entertainment domains, where fake news and AI generated videos have gone viral, leading to misinformation. The
primary objective is to develop a robust AI model that accurately detects deepfake content on social media platforms, focusing on identifying machine-generated tweets using FastText embeddings. Traditional methods involved human moderation, fact-checking agencies, and manual filtering of social media posts based on predefined rules and
keyword matching.

REMAINING USEFUL LIFE PREDICTOR FOR EV BATTERIES USING MACHINE LEARNING

G.Anjali ,T. Sravani , T. Ajanta Sravani , T. Vennela — 2024-12-25

Electric vehicles (EVs) are a key solution to combat rising carbon emissions and reduce dependence on fossil fuels. In India, the government has implemented policies such as the Faster Adoption and Manufacturing of Hybrid and Electric Vehicles (FAME) scheme to promote EV adoption Predicting the Remaining Useful Life (RUL) of EV batteries using machine learning ensures better battery health management and enhances operational efficiency. Applications include EV fleet management, battery recycling, and cost effective maintenance. To develop a machine learning model that accurately predicts the Remaining Useful Life (RUL) of EV batteries to improve operational reliability, reduce maintenance costs, and support sustainable energy practices.

A SUSPICIOUS FINANCIAL TRANSACTION DETECTION MODEL USING ML AND RISK-BASED APPROACH

2024-12-25

The detection of suspicious financial transactions has been a critical focus in the financial industry for decades. Traditionally, financial institutions employed rule-based systems for identifying potentially fraudulent activities. These systems rely on predefined thresholds and patterns, such as large transactions or frequent deposits, to flag suspicious activities. While effective to some extent, traditional systems face significant limitations. They often generate a high rate of false positives, requiring manual intervention to review flagged transactions. Additionally, these systems struggle to adapt to evolving fraud patterns, making them less effective in detecting sophisticated financial crimes.

AI- Driven Predictive Maintenance for Factory Equipment

Dr.G.Tirupathi Rao, Mahmad Sana, M.Sravani, M.Akhila — 2024-12-25

The advent of Industry 4.0 has revolutionized the manufacturing sector, emphasizing automation and data exchange in production technologies.In this context, the detection and classification of electrical equpiment faults are critical for maintaining optimal operational efficiency. Statistics indicate that unplanned downtimes due to electrical equipment failures can cost manufacturers up to $50 billion annually. Accurate fault detection can mitigate these losses by enabling predictive maintenance strategies. As manufacturing systems become more complex, traditional fault detection methods struggle to keep up with
the volume and variety of data generated. There is a pressing need for automated solutions that can analyze this data in real-time. Machine learning offers a promising avenue to address these challenges by leveraging historical and real-time data to predict and classify faults accurately.

Enhancing Cyber-Physical System Security: A Data-Driven Approach for Attack Detection Using DNN

Dr.K.Vijay Baskar, M.Poojitha, P.Akhila , P.Sushma — 2024-12-25

Cyber-physical systems enabled by the Internet of Things (IoT), such as industrial equipment and operational IT, facilitate the transmission and reception of data via the internet. This equipment will be equipped with sensors to monitor its status and transmit data to a centralized server via an internet connection. Occasionally, malevolent people may compromise or infiltrate such sensors, subsequently manipulating their data, which will then be sent to a centralized server, resulting in erroneous actions being executed. Many countries' equipment and production systems have failed due to erroneous data, prompting the development of numerous algorithms to detect attacks. However, these algorithms are hindered by data imbalance, where one class may contain a
substantial number of records (e.g., NORMAL records), while the other class, such as attacks, may have only a few records. This imbalance can result in the
failure of detection algorithms to predict accurately. To address data imbalance, current methods included OVER and UNDER sampling, which create new
records exclusively for the minority class. To address this problem, we are using a unique approach that does not use any under- or oversampling techniques. The suggested methodology has two components. An autoencoder will be trained on an unbalanced dataset to extract features, which will then be used to train a decision tree algorithm for predicting labels for both known and unknown assaults.

Enhanced security for Banking Transactions using Image Based Steganography

2024-12-25

In an age of increasing digital communication and data transfer, ensuring the security and privacy of sensitive information is paramount. Steganography, the art of hiding information within other data, has been used for centuries. In the digital realm, it plays a critical role in secure communication and information
concealment. Traditional steganography methods often involve embedding information within a single image. While effective, this approach may be susceptible to detection, as single-image steganography can leave detectable traces,especially under sophisticated analysis. The primary challenge is to
develop a robust system for multiple image steganography that can securely hide sensitive files within a set of images. This involves designing algorithms that distribute the information effectively across the images while maintaining imperceptibility and ensuring reliable extraction.

Secure Card Generation and Access Scheme for e-Seva Applications

G.Karunakar, M.Trisha , N.Sri Latha, N.Alekhya — 2024-12-25

In today's digital age, ensuring secure access to Automated Teller Machines (ATMs) is of paramount importance. The existing ATM access systems face significant security challenges due to the limitations of traditional authentication methods. Instances of card fraud, identity theft, and unauthorized access to bank accounts highlight the urgent need for more secure and robust authentication mechanisms. Additionally, with the increasing adoption of digital banking services, there is a growing demand for ATM access systems that can provide enhanced security without compromising user convenience

Safeguarding Against URL Attacks: Strategies for Chatbot Security in Phishing Web Activity

Dr.M.V.Subba Rao , L.Harshini, P.Usha Sri, P.Laxmi Sanjana — 2024-12-25

The rapid adoption of chatbots by organizations to efficiently manage user queries has brought significant advancements, but it has also introduced new risks. Traditionally, before the integration of machine learning (ML) and artificial intelligence (AI), phishing prevention relied on manual techniques such as blacklists, rule-based filters, and heuristic analysis, which were often slow and insufficient against evolving threats. The primary issue was the manual nature of these systems, which struggled to keep up with the sophisticated tactics used by malicious entities, leading to the exploitation of chatbots for phishing attacks. This challenge highlighted the need for more intelligent and adaptive security measures. The objective of this research is to design, develop, and integrate a self-defensive chatbot capable of identifying and neutralizing phishing attempts by inspecting URLs embedded in user interactions. The motivation behind this study stems from the increasing incidents where chatbots are manipulated to deliver phishing links that, when clicked, install malicious software to steal
sensitive data such as cookies and session passwords. This is particularly concerning for sectors like banking and finance, where compromised data can lead to significant user losses. The proposed system leverages machine learning algorithms, including Support Vector Machines (SVM), Random Forest, and Decision Tree, to create a robust model trained on the PHISH TANK URL dataset. This model can accurately distinguish between normal and malicious URLs in real-time, thereby enhancing the security of chatbot interactions. By evaluating each algorithm's performance through metrics such as accuracy, precision, recall, F-score, and confusion matrices, the system ensures optimal phishing detection capabilities.

An Innovative Approach of Intrusion Combat Model for Industrial Internet of Things: An Application of Industry 4.0

Dr.N.Baskar, M.Jahnavika , M.Sonalika , M.Tejasvi — 2024-12-25

Machine learning techniques are being widely used to develop an intrusion detection system (IDS) for detecting and classifying cyber-attacks at the network-level and host-level in a timely and automatic manner. However,many challenges arise since malicious attacks are continually changing and are occurring in very large volumes requiring a scalable solution. There are different malware datasets available publicly for further research by cyber security community. However, no existing study has shown the detailed analysis of the performance of various machine learning algorithms on various publicly available datasets.

Deep Learning Based Forgery Detection in Digital Images for Identifying Copy Move Abnormalities

M. Syamala SaiSree, J. Sanjana, K. Srija, K. Vineela — 2024-12-25

Copy-move forgery is a prevalent form of image manipulation where a part of an image is duplicated and pasted onto another area, often to conceal or alter
information. It's commonly encountered in digital media forensics, with applications in detecting tampered images, verifying authenticity, and ensuring
integrity in legal and journalistic contexts. Currently, detecting copy-move forgery relies heavily on manual analysis by forensic experts. This process involves visually inspecting images, looking for inconsistencies in textures, lighting, and patterns. Despite its reliability, manual analysis is time consuming and resource-intensive, limiting its scalability and efficiency in handling large datasets. Manual analysis suffers from subjectivity and human error, leading to potential inaccuracies in identifying forged regions. It's impractical for processing a vast number of images quickly, hindering its applicability in real-time scenarios. Additionally, the increasing sophistication of forgery techniques demands more robust and automated solutions. VGG 16, a convolutional neural network (CNN), offers a promising solution for automating copy-move forgery detection. Trained on extensive datasets, VGG 16 excels in feature extraction, enabling it to recognize patterns indicative of tampering with high accuracy. Its hierarchical architecture allows for the detection of both global and local inconsistencies in images, enhancing its versatility and effectiveness in identifying forged regions.

Smart Online Voting System using Block Chain Technology

David Livingston, K. Keerthi, G. Sahithi, G. Ramya — 2024-12-25

The rapid advancements in technology have paved the way for an efficient and secure online voting system, addressing challenges faced in traditional voting mechanisms. Online voting systems aim to provide a convenient, accessible, and efficient alternative to manual voting processes. However, the history of online voting systems reveals persistent challenges related to security, transparency, and voter privacy. Traditional systems, such as paper ballots and early electronic voting systems, were prone to issues like ballot tampering, vote manipulation, and lack of trust due to opaque processes.

ML-Enabled Intelligent Bot Detection in Network Communications

Manasu Madhavi, M. Soumya, M. Vani, M. Pavani — 2024-12-25

As the Internet of Things (IoT) continues to proliferate across diverse sectors, the security of IoT networks has become increasingly critical.Distributed Denial of Service (DDoS) BOT attacks pose a significant threat by overwhelming networks with malicious traffic, resulting in service disruptions, data breaches, and financial losses. Traditional detection systems rely heavily on signature-based approaches, which are ill-suited for zero-day attacks and lack adaptability to evolving threat landscapes.

Blockchain-Based Secure KYC Sharing System

B. Haritha Lakshmi, G. Sindhu, G. Geetha, K. Soniya — 2024-12-25

The Blockchain-Based Secure KYC (Know Your Customer) Sharing System leverages blockchain technology to securely streamline the management and sharing of customer identity data. KYC processes are essential for verifying customer identities in industries such as banking, finance, and telecommunications to ensure compliance with regulations and prevent fraud. Historically, KYC practices relied on centralized systems where institutions independently gathered and stored customer data, often leading to duplication, inefficiency, and heightened risks of data breaches.

Enhancing Insider Threat Detection in Cloud Environments Through Ensemble Learning

M. Vanitha, M. Navya Patel, K. Madhumitha, J. Sathvika — 2024-12-25

Cloud computing has drastically transformed data storage and accessibility,but it has also heightened concerns about insider threats—where employees
or privileged users misuse their access to sensitive information. According to the 2022 Cloud Security Report, insider attacks account for approximately 35% of global cloud data breaches, underscoring the urgency of addressing this threat vector. Traditional methods—such as rule-based systems, manual audits, and access log reviews—are not only reactive and labor-intensive but also prone to oversight and human error. These limitations become especially pronounced in large-scale cloud environments, where immense data volumes can overwhelm conventional security measures and delay or obscure detection. In response, this research advocates for a machine learning–driven solution, specifically harnessing ensemble learning techniques. By incorporating algorithms like Random Forest, AdaBoost, and CatBoost, the proposed system can process vast amounts of cloud logs and user activities more efficiently, enabling real
time anomaly detection. This approach significantly reduces false positives, enhances response times, and ultimately strengthens overall security posture against insider threats. Through integrated and automated analysis, ensemble learning models offer a proactive defense mechanism, ensuring greater integrity, confidentiality, and continuity of cloud-based operations.

Decentralizing Legal Documents Management Enhancing Security and Integrity Through Blockchain Technology.

Dr. Ranjith, K. Tejaswi, K. Sreeja, K. Sneha Reddy — 2024-12-25

Decentralizing the management of legal documents using blockchain technology offers a transformative solution to enhance security, integrity, and
transparency. Traditionally, legal document management relied on centralized systems, often susceptible to data breaches, unauthorized modifications, and
loss due to corruption or mismanagement. These conventional approaches necessitated trust in intermediaries, which occasionally led to inefficiencies,
tampering, and disputes. Historically, the advent of digital documentation was an advancement, but it lacked the robustness to address escalating challenges
in data integrity and security.

ML-Enabled IoT-Driven Smart Meter Data Analysis for Precise Energy Consumption Prediction

K. Kumaraswamy, Viramallu Sri Varshini, Satnoor Rohini Swamy, Siripuram Ruthvika — 2024-12-25

Countries are placing a greater emphasis on energy efficiency in an effort to reduce their carbon emissions and make the most of the resources they have available inside their borders. The tracking of energy use in the past consisted of either hand-reading meters or crude automated devices that provided only a limited amount of data. It is common for conventional energy consumption prediction systems to rely on too simplistic models that do not make full use of the opportunities presented by data from smart meters

A Machine Learning–Driven Framework for Predictive Risk Assessment of Automotive Air Pressure Systems

Rajkumar P, Yalabaka Jahnavi, S. Poojitha, Pokala Sreeja — 2024-12-25

Air pressure systems play a critical role in modern vehicles, underpinning essential functions such as braking and suspension. Despite their importance, conventional approaches to managing air pressure system failures—primarily scheduled maintenance and manual inspections—remain largely reactive and fail to fully utilize the vast data streams generated by today’s vehicles. This gap results in unplanned downtimes, increased maintenance costs, and
heightened safety risks. In response, this research proposes a proactive, machine learning–based risk assessment framework that integrates real-time sensor data to predict and prevent potential failures in air pressure systems.

ML–Driven Framework for Proactive Detection and Mitigation of Privilege Escalation Attacks in Cloud Environments

Suvarna Sunil Kumar, Nareddy Neha Sri, Penchala Siri, Nomula Anusha — 2024-12-25

Privilege escalation attacks in cloud environments pose a significant threat to the integrity and confidentiality of sensitive data. These attacks enable unauthorized users to gain elevated access privileges, thereby jeopardizing organizational security. Historically, such threats were addressed using manual monitoring and rule-based systems that relied on predefined thresholds and basic anomaly detection methods. However, these traditional approaches have proven both time-consuming and prone to high false-positive rates, especially when applied to large, dynamic datasets.

Cyber Security Threat Detection Model using Artificial Intelligence Technology

O. Ramyateja, Padilam Navya Sahithi, Samanthula Vaishnavi, Perolla Shirisha — 2024-12-25

Botnets—networks of compromised computers controlled by malicious actors—continue to pose a major threat to cybersecurity. Their impact is amplified by the use of Domain Generation Algorithms (DGAs), which dynamically produce new domain names for command-and-control servers, making traditional signature- and rule-based detection methods increasingly ineffective. This challenge underscores the need to enhance botnet DGA detection, particularly as attackers refine their techniques to outmaneuver existing defenses.

ML Driven Anomaly Detection for IoT Edge Devices: Insights from ADMM-Based Frequency Management

G. Siva Parvathi, Vadlamudi Alekya, Nasa Sridevi, Pendyala Archana — 2024-12-25

The rapid proliferation of Internet of Things (IoT) devices has necessitated the development of efficient and reliable anomaly detection mechanisms to ensure
system integrity, security, and performance. Traditional centralized anomaly detection systems are increasingly inadequate due to their scalability issues,
latency, and inability to handle the diverse and voluminous data generated by IoT edge devices. This project proposes an innovative Machine Learning (ML)-driven anomaly detection framework specifically designed for IoT edge devices, leveraging the Alternating Direction Method of Multipliers (ADMM) for effective
frequency management.

Revolutionizing Email Security with Quantum Key Distribution for Enhanced Data Protection In Communication Systems

Geetha Prathibha, Srinivasan Sowmya, Vaddepalli Likitha, Sitala Sanjana — 2024-12-25

In a digital age where sensitive material is often sent via email, safe communication is crucial. This project uses quantum cryptography to rethink email security. Quantum Key Distribution (QKD) uses quantum mechanics to construct unbreakable cryptographic keys, providing unprecedented cyber security. Email security developed with the internet. Early systems used passwords or unencrypted communications for protection. S/MIME and PGP encrypted email content, a major improvement.

The Synergy Between AI and Performance Engineering Enhancing Resiliency and Scalability

Sudhakar Reddy Narra — 2024-12-31

Context and Significance: Artificial Intelligence (AI) is driving a paradigm shift in performance engineering, offering advanced capabilities that go beyond traditional methods. As systems become more distributed, dynamic, and data-driven, AI plays a central role in enhancing scalability, resiliency, and efficiency.
In this era of complex infrastructures, AI techniques—such as predictive analytics, anomaly detection, and self-healing systems—are essential for
achieving real-time optimization and adaptive decision-making, significantly outpacing conventional approaches to performance engineering. AI’s integration into critical sectors like cloud computing, Internet of Things (IoT), 5G networks, and blockchain technologies is creating a foundation for next-generation self-optimizing, autonomous systems capable of continuously improving performance and reliability.