http://dspace.mist.ac.bd:8080/xmlui/handle/123456789/64 2026-04-21T14:51:37Z http://dspace.mist.ac.bd:8080/xmlui/handle/123456789/1061 IDENTIFICATION OF DDOS ATTACKTHROUGH INTRUSION DETECTION MODELUSING ENSEMBLEMACHINELEARNING ANIS, ADEEBA A distributed denial of service (DDoS) attack targets at hindering authorized individu als from accessing a server or website by flooding it with traffic from many sources. To avoid a DDoS attack from damaging the target system, detection is required. The sys tem becomes unsafe as a result of this attack. The aim of this thesis work is to provide an ensemble machine learning technique to detect DDoS attack. Another objective is to select optimal features of the dataset. In this thesis dataset is collected from Kaggle repository which contains 42 columns and 17171 rows. Firstly, three feature selection techniques—ANOVA, Mutual Information, and Feature Importance have been used to re duce the dataset and increase the performance. Then, optimal features have been selected using domain knowledge. The traditional machine learning methods K-Nearest Neigh bors (KNN), Support Vector Machine (SVM), Decision Tree (DT), and Naive Bayes (NB) are then used with the chosen features. Next, five ensemble models have been created by all the combinations of four traditional models- (KNN, SVM, DT), (KNN, SVM, NB), (KNN, NB, DT), (SVM, NB, DT) and (KNN, SVM, NB, DT). By evaluating accuracy, precision, recall, and F1-score, the experiment’s outcome is determined. After all the experiments, the result shows that the ensemble voting classifier by the combinations of KNN, SVMand DTgives the highest accuracy. Among the feature selection techniques, feature importance technique gives the maximum accuracy that is 98.86% and by using the optimal features, highest accuracy to detect the DDoS attack is determined which is 99.4% IDENTIFICATION OF DDOS ATTACKTHROUGH INTRUSION DETECTION MODELUSINGENSEMBLE MACHINELEARNING 2024-01-01T00:00:00Z http://dspace.mist.ac.bd:8080/xmlui/handle/123456789/1056 DEVELOPMENTOFANIOTANDBLOCKCHAIN INTEGRATED VERTICALFARMINGSYSTEM MD.MUNIM, KAZI The demand for sustainable and efficient procedures to produce safe food is increasing due to ongoing global challenges like population growth, resource deficiency, climate change, andglobal warming. Thusvertical farming, apioneeremergingtechnology, mayplayavital role in addressing such challenges in agriculture. Vertical farming systems need continuous monitoring of the plants and farm environment. There are scopes to improve efficiency, reduce human effort, and ensure proper yield. This research aims to explore the parameters required to monitor a vertical farm and to propose an IoT and blockchain based automated and efficient food production system. The IoT devices allow real time monitoring and con trolling of essential parameters like temperature, moisture, light, nutrient levels, and alike, while the Blockchain technology provides a transparent and immutable ledger of the records regarding crop growth and resource consumption of the farm, enhancing the framework’s traceability. A prototypical system of this framework was developed and simulated in the lab environment to evaluate its performance. By generating the required control instructions to maintain the ideal environment, the prototypical system proved to be effective, efficient, and reliable. Development of an IoT and Blockchain Integrated Vertical Farming System 2024-03-01T00:00:00Z http://dspace.mist.ac.bd:8080/xmlui/handle/123456789/1055 ISCHEMIC STROKELESIONSEGMENTATIONFROM DIFFUSION-WEIGHTED MRI USINGDEEPENSEMBLE LEARNING HALDER, RATHIN Strokes are a leading cause of premature mortality in developed and developing countries, and early treatment assistance can significantly prolong a patient’s life. How quickly the lesion is determined from MRI images is the primary rehabilitative step in stroke therapy. However, manual lesion identification takes time and is susceptible to both intra- and inter observer inconsistencies. Since manual lesion detection is highly time-consuming, it can negatively impact patient outcomes and overall experience. To aid in diagnosis, treatment planning, and analysis, medical image segmentation separates a medical image into indi vidual parts or segments, each corresponding to a particular anatomical structure or tissue, such as organs, lesions, or other areas of interest. In light of this, computerized estimation of the outcome of the ischemic stroke lesion can assist physicians in better evaluating the stroke and providing information on tissue outcomes. So, this will be an essential tool for assessing the extent of brain cell damage. Convolutional neural networks (CNN) have been extensively utilized in the categorization of abnormalities from brain images in recent years. So, this can be achieved by accurately classifying the characteristics of ischemic stroke le sions employing a convolutional neural network with convolutional layers. Consequently, to separate the Ischemia in the current study, a deep-learning network is employed in this thesis. The primary discovery of this work is the extraction of ischemic lesion character istics by utilizing the InceptionV3 network and the preservation of Z-axis information by employing a conventional 3D U-NET architecture. The proposed model was trained and tested using the ISLES-2017 dataset, and the experimental results obtained an overall seg mentation Dice Coefficient of 0.43. The results of this investigation show that the proposed technique is superior to earlier studies. ISCHEMICSTROKELESIONSEGMENTATIONFROM DIFFUSION-WEIGHTEDMRIUSINGDEEPENSEMBLE LEARNING 2024-09-01T00:00:00Z http://dspace.mist.ac.bd:8080/xmlui/handle/123456789/1054 TRANSFORMERANDPOSEGRAMMARBASEDDEEP NEURALNETWORKFOR3DHUMANPOSE ESTIMATION SULTANA, ZINIA For computers to understand human activity or behavior in a variety of scenarios, reli able 3D human posture estimation is a prerequisite. A number of difficulties have made such work more complex as it is influenced by various factors, including image quality, background, garment texture and diversity, body shape, and the presence of other objects alongside persons in the image which has depicted the necessity of adopting the technique of computer vision. While much work has been done on 2D human pose estimation, show ing state-of-the-art performance, the objective of this research is to estimate 3D human pose from 2D joint positions. We have investigated deep neural networks comprising of linear layers with residual blocks and proposed a hybrid deep learning framework in order to achieve this objective. We experimented the proposed by raising the number of residual blocks to anlaysis the performance. The final proposed architecture (HEpose) comprises of three parallel models, one model is base one only the linear layers concept, second one is based on the residul connection without normalization, and third model gathers the in formation of connection among the joints. We combined ouputs of the three model and f inally used a fully connected linear layer to estimate 3D pose. We also showed compar ative training results. Finally, the proposed architecture was evaluated on H3WB dataset and presented the evaluation results considering the evaluation metrics of the mean per joint position error (MPJPE) and the percentage of correct keypoints (PCK). The proposed architecture performed about 50% better in terms of MPJPE and PCK@150mm for three residual block. We had also compared the performance of HEpose with other state-of-the art methods of 3D pose estimator and achieved inevitable performance. Transformer and Pose Grammar Based Deep Neural Network for 3D HumanPose Estimation 2024-03-01T00:00:00Z