Obstacle Detection for UAV & Drone using YOLO V8 and HC-SR04 Ultrasonic Sensor Chittaranjan Pradhan, Gourab Biswas, Jugul Kishor Gupta, Bibhuti Bhusan Dash, Utpal Chandra De, Sudhansu Shekhar Patra Proceedings of the 2026 International Conference on AI Driven Smart Systems and Ubiquitous Computing Icauc 2026, 2026 The article dives deep into the complications and inspects the development of a real-time obstacle detection system designed for UAVs and drones with the integration of advanced computer vision techniques and ultrasonic sensor technology. The combination of object detection, location, bounding box, and motion analysis conveys the exact state and shape of the existing object in the environment so that the drone or the remote pilot could locate and differentiate between stationary and moving objects. The YOLOv8 model is chosen due to its high precision for detection with added segmentation masks that outline the shape of obstacles, giving better situational awareness. By introducing the HCSR-04 ultrasonic sensor also complements this by measuring distance from the drone to detected obstacles. The HCSR-04 works on the principle of transmitting ultrasonic waves and then later using the time that reflected waves take to come back (SONAR) the sensor measures the accurate distance in real time. It uses an Arduino Uno micro controller so that it can control and process the outputs of the HCSR-04 in order to amplify the flexibility and responsiveness of the system. The proposed system demonstrates experiments which assures that it works perfectly in different circumstances, accurately segmenting and localizing obstacles. The autonomous UAVs or drones get feedback in correct time which help them avoid obstacles and make their way through safely in complex surroundings by using the high-performance, user-friendly and affordable solution of integrated object detection capability of YOLOv8 with Arduino-controlled ultrasonic sensing.
Mitigating data exfiltration from side-channel attacks on graphics processing units Nelson Lungu, Bibhuti Bhusan Dash, Binod Kumar Pattanayak, Rajen Bose, Utpal Chandra De, Sudhansu Shekhar Patra Futuristic Information and Communication A Multimodal Multidisciplinary Signal Analysis, 2026 Graphics Processing Units (GPUs) are progressively used to expedite compute-intensive applications. Adversaries may use the data parallelism intrinsic to GPUs to extract sensitive information via timing, power, and cache assaults. This study introduces a Secure Shader Execution Framework that addresses these vulnerabilities via the integration of randomized execution, power balancing, and cache partitioning. The proposed vendor-agnostic method aligns with current GPU programming paradigms, as shown by the GPUOwl benchmarks using OpenCL. Experimental findings indicate that the framework successfully conceals side-channel information leakage in security-sensitive data with a minimal performance cost. Randomized execution lowers the success rate of timing assaults by as much as 75%, while power balancing decreases leakage in power traces by over 60%.
Multi-Model Feature Fusion-based Ensemble of Pre-trained CNNs for Chest X-ray Classification Soumyarashmi Panigrahi, Dibya Ranjan Das Adhikary, Binod Kumar Pattanayak, Bibhuti Bhusan Dash, Utpal Chandra De, Sudhansu Shekhar Patra Proceeding of International Conference on Computing Communication Control and Cyber Physical Systems I5cps 2026, 2026 A correct diagnosis is necessary to confirm prompt and effective therapy for the proliferation of aberrant cells in the thoracic area and pulmonary nodules. Accurate classification of chest X-rays (CXR) is utmost necessary to generate effective treatment and improve patient condition. CXR are widely and fundamentally available diagnostic tool for initial findings. However, assessment of CXR is very challenging and leads to human error, which evolves the need for automatic and accurate pathology classification. The latest developments in deep learning (DL), specially convolutional neural networks (CNNs), have transformed medical image analysis. In this study, framework ensembling multiple pre-trained CNN models such as EfficientNetB7, InceptionV3, ResNet50V2, and VGG19 are proposed for image classification. The feature-fused model, including individual models are trained and evaluated on a particular dataset with variety of metrics such as accuracy, AUC, Precision, Specificity, and F1-Score. Our results demonstrate that the ensemble of these pre-trained models achieves an accuracy 98.88%. This shows the efficacy of advanced CNN architectures in giving reliable decision support for CXR classification.
Advanced Deep Learning OCR Algorithm for Deciphering and Digitizing Chola Period Tamil Inscriptions M Rajalakshmi, Shanthi Jeyabal, Manoj Ranjan Mishra, Bibhuti Bhusan Dash, Utpal Chandra De, Sudhansu Shekhar Patra 7th International Conference on Mobile Computing and Sustainable Informatics Icmcsi 2026, 2026 The Tamil inscriptions of the Chola are important in terms of conserving historical, cultural and linguistic history of the area. It could be tedious and time-consuming to read these inscriptions with the hand though. This paper discusses a customized Optical Character Recognition (OCR) system that facilitates easier digitization and transcription of old Tamil inscriptions. The solution is considered to be a combination of the YOLOv8, Tensorflow, and PyTorch systems that will promote the effectiveness of character recognition, segmentation, and detection. Our experiment with the method on a number of the inscriptions of the Chola-period was 94.7 percent, 93.5 percent, and 95.2 percent correct, precise, and recalled. The average time of processing each picture was 2.8 seconds. This model is more effective in dealing with broken inscriptions, complex backgrounds and mixed character styles than those that preceded it. The proposed OCR approach is appropriate to facilitate easier study and preservation of the Tamil cultural heritage as the epigraphers, archaeologists and researchers can scan and save these ancient manuscripts.
Deepsea: A Deep Learning-Based Secure and Energy-Aware Adaptive Protocol for Underwater Sensor Networks Smita Patra, Manoj Ranjan Mishra, Sudhansu Shekhar Patra, AparnaRajesh Atmakuri, Bibhuti Bhusan Dash, Utpal Chandra De Proceedings of the 5th International Conference on Sentiment Analysis and Deep Learning Icsadl 2026, 2026 Underwater Sensor Networks (UWSNs) are essential for marine exploration, environmental monitoring, and naval operations, but their performance is constrained by limited energy, dynamic communication conditions, and vulnerability to cyber attacks. This study introduces DeepSEA (Deep-learning Secured Energy-Aware Adaptive protocol), communication protocol designed to enhance energy efficiency, reliability, and security in UWSNs. The framework integrates deep learning and cyber security mechanisms to enable autonomous and resilient underwater networking.. The proposed framework incorporates deep learning and cyber security strategies to facilitate autonomous and robust underwater networking. Specifically, An LSTM-based energy management model predicts energy consumption and harvesting trends, while a Deep Q-Network (DQN) optimizes routing decisions dynamically based on residual energy, link quality, and trust scores. A lightweight cryptographic layer employing Elliptic Curve Cryptography (ECC) ensures data confidentiality with minimal overhead. Furthermore, an AIbased Intrusion Detection System (IDS) trained on the UNSWNB15 dataset detects attacks such as Sybil, replay, and wormhole using Random Forest, XGBoost, and SVM classifiers. Experimental simulations using NS-3 with AquaSim demonstrate that DeepSEA improves energy efficiency by 29.7 %, extends network lifetime by 43.1 %, and achieves an intrusion detection accuracy of 93.8 %, while maintaining low latency and high packet delivery. The results highlight that the integration of AI-driven adaptivity and lightweight cryptography makes DeepSEA a scalable and robust framework for secure autonomous underwater communications.
Leveraging Big Data Analytics and Applied Mathematics for Predictive Maintenance in Industrial Automation Systems Nelson Lungu, Sudhansu Shekhar Patra, Bibhuti Bhusan Dash, Utpal Chandra De 4th International Conference on Sentiment Analysis and Deep Learning Icsadl 2025 Proceedings, 2025 Predictive maintenance of industrial automation systems has attracted much interest recently, owing to the potential benefits of reducing downtime and improving production efficiency. However, existing solutions are affected by many different types of problems, such as managing heterogeneous and huge-scale data from dispersed industrial sensors, realising real-time analysis in high-speed data streams, and, at the same time, preserving interpretability while employing advanced techniques like deep learning and distributed machine learning. The recent approaches also use deep neural networks for fault detection and anomaly-based models for early warning signal identification with federated learning to ensure privacy-preserving data sharing. These have inherent complications such as a high rate of false positives and data imbalance handling difficulty; besides that, it is not flexible enough with the changing dynamics of the industrial processes. This paper addresses these limits by introducing an innovative architecture that merges big data analytics with applied mathematics to enhance predictive maintenance tasks effectively. To this end, one integrates machine-learning models with physics-based models across diverse datasets to enhance prediction accuracy while reducing false alarms and thus enabling proactive maintenance scheduling. In so doing, this approach will aim at shattering the present system's weaknesses by delivering robust anomaly detection along with decreased operational expenses in a scalable manner, thereby playing an important role towards further smarter, self-configurable manufacturing within the Industry 4.0 movement.
Computational Fluid Dynamics Modelling for Controlling Hybrid Nanoliquid Flow Distribution in Datacenters Satya Subha Shree Sen, Bibhuti Bhusan Dash, Utpal Chandra De, Mrutyunjay Das, Sudhansu Shekhar Patra 2025 IEEE 4th International Conference for Advancement in Technology Iconat 2025, 2025 Datacenters generate an enormous volume of heat. Due to this heat, the performance of the datacenters drops, and in the worst case, it may catch fire. There is a need to cool the datacenters. In this paper, the liquid cooling effect is studied through computational fluid dynamics (CFD). This study examines the impact of quadratic thermal radiation on magnetohydrodynamic (MHD) catalytic flow of a penta-hybrid nanoliquid past a stretching sheet. The nanoliquid is engineered by dispersing five different types of nanoparticles into a base fluid. It is designed to optimize heat and mass transfer rates. A coupled homogeneous-heterogeneous catalytic surface reaction is incorporated, and the effects of viscous dissipation are included within the energy equation. Departing from conventional linear radiation models, the quadratic radiation approach offers superior fidelity in characterizing thermal transport at elevated temperatures. The governing partial differential equations are transformed via similarity variables into a system of nonlinear ordinary differential equations, which are solved numerically using MATLAB's BVP4C solver. The findings underscore the significant role of magnetic strength in modulating the velocity, temperature, and concentration fields, offering actionable insights for the cooling effects in datacenters.
Modeling Adversarial Strategies in GPU Side-Channel Attacks: A Game-Theoretic Perspective Nelson Lungu, Suchismita Rout, Bibhuti Bhusan Dash, Sudhir Kumar Behera, Utpal Chandra De, Sudhansu Shekhar Patra Proceedings of 2025 International Conference on Computing for Sustainability and Intelligent Future Comp Sif 2025, 2025 The use of Graphics Processing Units (GPUs) in safety-critical applications is increasing; yet, their highly parallel designs provide weaknesses that facilitate side-channel attacks, enabling the theft of confidential information. Prior research has shown that time, contention, power, and access pattern side-channels are viable against GPU workloads. Nonetheless, their current protections remain insufficient to comprehensively safeguard real-world shader operations. A novel game-theoretical framework is proposed to model and evaluate the intricate dynamics between attackers and defenders in GPU side-channel assaults. The interactions are reformulated as a two-player, non-cooperative game, and the best tactics for both players have been established under various payment models and danger situations. Experimental results obtained from commercial GPUs corroborate that the suggested technique effectively represents hostile dynamics in reality, necessitating the development of resilient countermeasures. This project seeks to bridge the gap between theoretical security analysis and practical GPU protection methods. We provide a robust foundation for designing appropriate and secure GPU architectures for continuously changing side channel threadings. Human-centric approaches to driving work emphasise the need of considering human elements like as perception, judgement, and decision-making to effectively assess adversary techniques in cybersecurity. The game theoretical model offers a structured approach for forecasting probable assault vectors, assessing defence methods, and formulating effective counters customised to the adversary.
IoT Inspired Monitoring System for Air Pollution using Random Forest Classifier Pravat Kumar Rautaray, Binod Kumar Pattanayak, Bibhuprasad Mohanty, Bibhuti Bhusan Dash, Utpal Chanda De, Sudhansu Shekhar Patra Proceedings of 8th International Conference on Inventive Computation Technologies Icict 2025, 2025 Air pollution is affecting millions of people globally and is becoming a major health concern. Furthermore, it has harmful gasses and (PM2.5), or tiny microscopic particulate matter, which lower the air quality. Not only has this been the primary focus of scientific investigation, but it has also become a major social issue in the lives of the general population. This assertion is supported by the World Health Organization (WHO) estimate that the health effects due to air pollution result in 2.4 million fatalities each year. Chronic exposure to air pollution can lead to a range of serious health problems, including respiratory and lung conditions. Numerous organizations alert people to the fact that severe air pollution exists in many parts of the world. Governments and central agencies may find that these models have policy implications for preventing excessive pollution levels. Although there have been several attempts to simulate pollution levels in the literature, better data integration and more accurate prediction results are anticipated due to current developments in machine learning and deep learning approaches. This paper presents a thorough investigation into the modeling of actual air pollution data using machine learning frameworks. In this study five basic machine learning algorithms have been compared to predict the air quality. The Random Forest algorithm gives the best accuracy of 98%.
ResNet-GRU Hybrid Model for Brain Tumor Diagnosis: A Sequential Learning Framework Soumyarashmi Panigrahi, Dibya Ranjan Das Adhikary, Binod Kumar Pattanayak, Bibhuti Bhusan Dash, Utpal Chandra De, Sudhansu Shekhar Patra 2nd International Conference on Machine Learning and Autonomous Systems Icmlas 2025 Proceedings, 2025 This research investigates the effectiveness of integrating Gated Recurrent Units (GRUs), a variant of Recurrent Neural Network (RNN), into pre-trained Convolutional Neural Networks (CNNs) such as Residual Networks (ResNet50) for enhanced brain tumor classification using Magnetic Resonance Imaging (MRI) scans. RNNs are a class of Neural Networks (NNs) structured to analyze sequential data, while GRU is a specific type of RNN that efficiently learns long-term dependencies without suffering from vanishing gradients. ResNet50 is a robust architecture renowned for its exceptional image feature extraction capabilities. GRU specializes in modeling complex temporal relationships. By integrating these two, this approach effectively synergizes the strengths of both architectures to achieve enhanced performance. This combination enables the model to capture spatial and temporal dependencies in MRI scans, improving brain tumor classification accuracy. Using the publicly available “Br35H: Brain Tumor Detection 2020” dataset, we evaluate the effectiveness of the suggested ResNet50-GRU model, achieving state-of-the-art results. This comprehensive evaluation employs various metrics, such as Accuracy, Cohen's Kappa, Matthews Correlation Coefficient (MCC), Area Under the Curve (AUC), Precision, Recall, and F1-score. The results underscore the advantages of combining CNNs and GRUs for medical image analysis. This study provides valuable insights for developing more accurate and efficient Deep Learning (DL) based diagnostic tools, highlighting the prospects of CNN-RNN combinations in medical imaging applications.
Feature Evaluation of Nvidia R555-R560 Open-Source GPU Driver Transition Nelson Lungu, Bibhuti Bhusan Dash, Utpal Chandra De, Simon Tembo, Satyendr Singh, Sudhansu Shekhar Patra Proceedings of 2025 International Conference on Signal Processing Computation Electronics Power and Telecommunication Iconscept 2025, 2025
Intelligent Air Quality Control Through Continuous Policy Learning Jayashri Deb Sinha, Subir Gupta, Sayanti Samanta, Bibhuti Bhusan Dash, Utpal Chandra De, Sudhansu Shekhar Patra Proceedings of the 9th International Conference on Inventive Systems and Control Icisc 2025, 2025
Hybrid Binary SGO-GA for solving MAX-SAT problem Rhiddhi Prasad Das, Anuruddha Paul, Junali Jasmine Jena, Bibhuti Bhusan Dash, Utpal Chandra De, Mahendra Kumar Gourisaria Procedia Computer Science, 2025
Classifying Crisis Types and Urgency Levels in Tweets using TF-IDF and LSTM Basudev Nath, Suchismita Rout, Lalbihari Barik, Bibhuti Bhusan Dash, Utpal Chandra De, Sudhansu Shekhar Patra Proceedings of 3rd International Conference on Intelligent Cyber Physical Systems and Internet of Things Icoici 2025, 2025
Adaptive Gamified Cybersecurity Training for Enhanced User Engagement Nelson Lungu, Bibhuti Bhusan Dash, Utpal Chandra De, Suchismita Rout, Tanvir Habib Sardar, Sudhansu Shekhar Patra Proceedings of 5th International Conference on Soft Computing for Security Applications Icscsa 2025, 2025
Evaluating Human AI Collaboration Through Survey Based Random Forest Approach Subir Gupta, Priyanka Roy, Sudipta Hazra, Bibhuti Bhusan Dash, Utpal Chandra De, Sudhansu Shekhar Patra Proceedings of 2025 International Conference on Signal Processing Computation Electronics Power and Telecommunication Iconscept 2025, 2025
Efficient DDoS Detection in IoT Networks Using a CNN-GRU Hybrid Model Mohammad Osama Addas, Suprava Ranjan Laha, Susmita Panda, Binod Kumar Pattanayak, Bibhuti Bhusan Dash, Utpal Chandra De, Sudhansu Shekhar Patra 2025 5th Asian Conference on Innovation in Technology Asiancon 2025, 2025
Machine Learning Technology based Heart Disease Detection Model M. Subhashini, GR. Ashisha, X. Anitha Mary, Bibhuti Bhusan Dash, C. Karthik, Utpal Chandra De, Matam Mohan Babu, P. Jyotheeswari, Sudhansu Shekhar Patra 2025 2nd Asia Pacific Conference on Innovation in Technology Apcit 2025, 2025
Analyzing Security Implications for Artificial Intelligence Driven Medical Training Simulations Bibek Bikram Dash, Hameed Hassan Khalaf, Ahmed Read Al-Tameemi, Hiba ganem Hussain, Kadhim Abbas Jabbar, Amran Mezher Lawas, Bibhuti Bhusan Dash, Sudhansu Shekhar Patra, Utpal Chandra De 2025 International Conference on Next Generation of Green Information and Emerging Technologies Giet 2025, 2025
Blockchain-Based Framework for Enhanced Healthcare Data Accessibility Bibhuti Bhusan Dash, Hassan Abozibid, Hameed Hassan, Yaser Ahmad Ibrahim, Ali Ihsan Alanssari, Kadim A. Jabbar, Satyendr Singh, Utpal Chandra De, Sudhansu Shekhar Patra 2025 International Conference on Next Generation of Green Information and Emerging Technologies Giet 2025, 2025
Enhancing Breast Cancer Detection Using SVM and Explainable AI J. Anushree, GR. Ashisha, Utpal Chandra De, Bibhuti Bhusan Dash, X. Anitha Mary, C.Karthik, Matam Mohan Babu, P. Jyotheeswari, Sudhansu Shekhar Patra 2025 5th International Conference on Emerging Research in Electronics Computer Science and Technology Icerect 2025, 2025
THE ROLE OF CITIZEN ENGAGEMENT IN DEMOCRATIC GOVERNANCE ENHANCEMENT THROUGH E-GOVERNANCE: A CASE STUDY OF LUSAKA CITY COUNCIL, ZAMBIA Journal of Theoretical and Applied Information Technology, 2024
Multi-label Aspect Based Sentiment Analysis using Fuzzy LSTM Abinash Tripathy, Satyabrata Patro, Madhulika Tripathy, Utpal Chandra De 2024 IEEE 1st International Conference on Green Industrial Electronics and Sustainable Technologies Giest 2024, 2024
Performance Analysis of Classifiers in Predicting Car Insurance Claim Spoorthi Bhoji, Trilok Nath Pandey, Bibhuti Bhusan Dash, Rabinarayan Satpathy, Utpal Chandra De, Sudhansu Shekhar Patra Proceedings 2024 8th International Conference on Inventive Systems and Control Icisc 2024, 2024
Search Engine for QnA using Distributed Inverted Index System Snehasis Dey, Bhimasen Moharana, Utpal Chandra De, Tapaswini Samant, Trupti Mayee Behera, Shobhan Banerjee 2024 3rd International Conference for Innovation in Technology Inocon 2024, 2024
Advance Chess Engine: an use of ML Approach Kushagra Srivastava, Trilok Nath Pandey, Bibhuti Bhusan Dash, Sudhansu Shekhar Patra, Manoj Ranjan Mishra, Utpal Chandra De 2024 3rd International Conference for Innovation in Technology Inocon 2024, 2024
Probing Vulnerabilities in GPU Shader Execution Nelson Lungu, Simon Tembo, Sudhansu Shekhar Patra, Ngula Walubita, Bibhuti Bhusan Dash, Utpal Chandra De Proceedings 2nd IEEE International Conference on Device Intelligence Computing and Communication Technologies Dicct 2024, 2024
Performance of VM in SDN-Assisted Cloud Data Center with Working Vacations Lalbihari Barik, Bibhuti Bhusan Dash, Manoj Ranjan Mishra, Suchismita Rout, Utpal Chandra De, Sudhansu Shekhar Patra 7th International Conference on I Smac Iot in Social Mobile Analytics and Cloud I Smac 2023 Proceedings, 2023
A task offloading scheme with Queue Dependent VM in fog Center Sibananda Behera, Namita Panda, Utpal Chandra De, Bibhuti Bhusan Dash, Binita Dash, Sudhansu Shekhar Patra 2023 6th International Conference on Information Systems and Computer Networks Iscon 2023, 2023
Sentiment clustering using the unsupervised machine learning approach Abinash Tripathy, Utpal Chandra De, Bibhuti Bhusan Dash, Sudhansu Shekhar Patra, Binod Kumar Pattanayak, Trilok Nath Pandey Proceedings of the 2023 6th International Conference on Recent Trends in Advance Computing Icrtac 2023, 2023
Sentiment analysis of reviews using Nature Inspired Algorithm Abinash Tripathy, Utpal Chandra De, Bibhuti Bhusan Dash, Sudhansu Shekhar Patra, Binod Kumar Pattanayak, Trilok Nath Pandey 2023 4th IEEE Global Conference for Advancement in Technology Gcat 2023, 2023
Sentiment analysis of reviews using Nature Inspired PSO and ANN Abinash Tripathy, Utpal Chandra De, Bibhuti Bhusan Dash, Sudhansu Shekhar Patra, Ch. Chakradhara Rao, Trilok Nath Pandey 2023 Global Conference on Information Technologies and Communications Gcitc 2023, 2023
Energy-Efficiency in Software-Defined Networking: Rule Placement Approach Ngula Walubita, Bibhuti Bhusan Dash, Rabi Narayan Satapathy, Abinash Tripathy, Sudhansu Shekhar Patra, Utpal Chandra de International Conference on Self Sustainable Artificial Intelligence Systems Icssas 2023 Proceedings, 2023
Performance Evaluation of Drones in FANETs using Queueing Model Akash Ghosh, Bibhuti Bhusan Dash, Sudhansu Shekhar Patra, Trilok Nath Pandey, Binod Kumar Pattanayak, Utpal Chandra De International Conference on Sustainable Communication Networks and Application Icscna 2023 Proceedings, 2023
Leasing in IaaS Cloud Using Queuing Model Bibhuti Bhusan Dash, Utpal Chandra De, Manoj Ranjan Mishra, Rabinarayan Satapathy, Sibananda Behera, Namita Panda, Sudhansu Shekhar Patra Lecture Notes in Networks and Systems, 2023
