|
|
Návrh počítačových cvičení pro zabezpečení informačních systémů
Nakládalová, Kateřina ; Tomašov, Adrián (oponent) ; Myška, Vojtěch (vedoucí práce)
V dnešní době je spousta služeb poskytována v internetovém prostředí prostřednictvím webových aplikací. Kyberprostorem se tak přenáší velké množství dat a citlivých informací, které mohou být útočníky odcizeny při nedostatečném zabezpečení systémů, které s nimi pracují. Vzdělání v oblasti Informační bezpečnosti je proto stěžejní. Za tímto účelem byla vytvořena webová aplikace v programovacím jazyce Java k demonstraci útoků na ni mířených. Aplikace je sestrojena s využitím frameworku Spring a připojena k databázi PostgreSQL. V práci jsou předvedeny a popsány nejběžnější útoky na webové aplikace, kterými jsou Útok hrubou silou, útoky SQL injection, útok Cross-site scripting a Cross-site request forgery, chyby v řízení přístupu, expozice citlivých dat, nekonzistence dat ve webové aplikaci a zásah do manipulace s parametry webové aplikace. Proti těmto útokům jsou uvedeny konkrétní způsoby implementace ochrany k zajištění jejich bezpečného používání. Tato práce může sloužit jako užitečný zdroj informací pro vývojáře i uživatele webových aplikací, kteří chtějí získat vědomosti o problematice počítačových útoků a snižovat tím riziko jejich výskytu.
|
|
|
Accelerated sensor data analysis using an embedded system with a graphics processing unit
Maczkó, Adam ; Myška, Vojtěch (oponent) ; Tomašov, Adrián (vedoucí práce)
The thesis deals with two main goals, namely the acceleration of data analysis and the subsequent visualization of this data. The purpose is to speed up the supplied application so that it is suitable for protecting optical infrastructures through real-time analysis of polarization state changes and visualization of its outputs. The thesis describes topics that are important in terms of accelerating computations on the graphics processor, particularly principles of parallelism, parallel programming, processes, threads, and parallel architectures. In addition, the thesis describes the capabilities of the Jetson Nano platform. The output of the thesis is an application that is capable of performing computations on the graphics processor and has a web interface for visualizing the analyzed data. The PyTorch library was used for acceleration on the graphics processor. Visualisation was achieved through the React framework in conjunction with the react-spectrogram and ApexCharts libraries.
|
|
|
Mobilní aplikace pro inteligentní systém chovu včel
Pecár, Martin ; Myška, Vojtěch (oponent) ; Kiac, Martin (vedoucí práce)
Cílem práce je navrhnout a vytvořit aplikaci, která umožňuje včelařům správu včelstva pomocí mobilního telefonu. Důvodem je centralizace a zpřehlednění všech nasbíraných dat z kontrol včelstva, přičemž tato data bude později možné použít při tvorbě statistik. Dále tato aplikace obsahuje možnosti, jak včelaře upozornit na potřebu zásahu do včelstva pomocí jím vytvořených upozornění a statistiky o vybraných sledovaných vlastnostech včelstva. Výstupem bude popsaná mobilní aplikace.
|
|
|
Universal data storage for IoT applications
Kadlíček, Matej ; Myška, Vojtěch (oponent) ; Šteffan, Pavel (vedoucí práce)
The master’s thesis is written about the topic of data warehouse and the technology associated with it. The aim of the theses was to select programming tools for creating a website. Next aim was to design the functionality, appearence and datastorage. Subsequent implementation of design was done after. Theses was divided into 5 chapters. The first one includes selections of programming tools for a creating a backend. Second chapter includes design of database storage. Third chapter includes design of functionality and user interface. Fourth chapter includes implementation of the application and the last one includes testing.
|
|
|
Mobilní aplikace pro systém monitorující optická vlákna
Vaverka, Jan ; Myška, Vojtěch (oponent) ; Dejdar, Petr (vedoucí práce)
Hlavním cílem této diplomové práce je vývoj multiplatformní mobilní aplikace ve frameworku Flutter pro platformy iOS a Adnroid. Aplikace bude vzdáleně komunikovat s~analyzátorem optických vláken skrz server. V úvodu práce je popsána problematika optických vláken a možnosti zabezpečení. V další části jsou popsány možnosti komunikace mezi aplikací a serverem, rovněž je zde popsán framework Flutter. V praktické častí je navrhnuta komunikace mezi aplikací a serverem. Velká část diplomové práce se věnuje návrhu a vývoji multiplatformní aplikace. Poslední část se věnuje testování komunikace mezi aplikací a serverem.
|
|
|
Detection Of Road Surface Defects From Data Acquired By A Laser Scanner
Myska, Vojtech
Research in the field of automatic detection of road surface defects has been relativelywidespread in recent years. Most of the existing works solve this issue by processing the imageacquired by camera technology. The contribution of this study is the proposal of the LRS-CNN algorithmfor the detection of defects on road surfaces based on their laser scans. The advantage ofLRS-CNN is the ability to detect so-called microcracks, which can not be recognized from camerarecordings. We have also found that transfer learning methods are not suitable for the use of road defectdetection from their laser scans. Our LRS-CNN algorithm has been trained on unique nonpublicdata and is able to achieve up to 99.33% of success depending on the type of task.
|
|
|
Graph Convolutional Neural Networks For Sentiment Analysis
Myska, Vojtech
Commonly used approaches based on deep learning for sentiment analysis task operating over data in Euclidean space. In contrast with them, this paper presents, a novel approach for sentiment analysis task based on a graph convolutional neural networks (GCNs) operating with data in Non-Euclidean space. Text data processed by the approach have to be converted to a graph structure. Our GCNs models have been trained on 25 000 data samples and evaluated 5 000 samples. The Yelp data set has been used. The experiment is focused on polarity sentiment analysis task. Nevertheless, a relatively small training data set has been used, our best model achieved 86.12% accuracy.
|
|
|
Interactive web presentation of audiovisual works
Paulech, Matúš ; Myška, Vojtěch (oponent) ; Sikora, Pavel (vedoucí práce)
The bachelor thesis contains a presentation of audiovisual works processed by a calibrator based on neural networks (deep learning and artificial intelligence). The work deals with uploading video files to a web application and then playing them on a website. Data, that have been processed by neural networks, are added to the videos. This data is divided according to the classification into given tags, in which it is written what is currently in the image of video. These tags can be used to classify videos, and these classifications are also displayed in real time during video playback.
|
|
|
Segmentace obrazových dat využitím hlubokých neuronových sítí
Hrdý, Martin ; Myška, Vojtěch (oponent) ; Kiac, Martin (vedoucí práce)
Cílem této diplomové práce je seznámit se a nastudovat teorii současných segmentačních metod, které používají hluboké učení. Na základě teoretických znalostí bude navržena a vytvořena segmentační neuronová síť, která bude schopna segmentovat jednotlivé instance objektů. Segmentační neuronová síť bude zaměřena na detekci elektronických součástek na deskách plošných spojů.
|
|
|
Segmentation of multiple sclerosis lesions using deep neural networks
Sasko, Dominik ; Myška, Vojtěch (oponent) ; Kolařík, Martin (vedoucí práce)
This master thesis focused on automatic segmentation of Multiple Sclerosis (MS) lesions on MRI images. We tested the latest methods of segmentation using Deep Neural Networks and compared the approaches of weight initialization by transfer learning and self-supervised learning. The automatic segmentation of MS lesions is a very challenging task, primarily due to the high imbalance of the dataset (brain scans usually contain only a small amount of damaged tissue). Another challenge is a manual annotation of these lesions, as two different doctors can mark other parts of the brain as damaged and the Dice Coefficient of these annotations is approximately 0.86, which further underlines the complexity of this task. The possibility of simplifying the annotation process by automatization could improve the lesion load determination and might lead to better diagnostic of each individual patient. Our goal was to propose two techniques that use transfer learning to pre-train weights to later improve the performance of existing segmentation models. The theoretical part describes the division of artificial intelligence, machine learning and deep neural networks and their use in image segmentation. Afterwards, the work provides a description of Multiple Sclerosis, its types, symptoms, diagnosis and treatment. The practical part begins with data preprocessing. Firstly, brain scans were adjusted to the same resolution with the same voxel size. This was needed due to the usage of three different datasets, in which the scans had been created by devices from different manufacturers. One dataset also included the skull, therefore it was necessary to remove it by an FSL tool, leaving only the patient's brain in the scan. The preprocessed data were 3D scans (FLAIR, T1 and T2 modalities), which were cut into individual 2D slices and used as an input for the neural network with encoder-decoder architecture. The whole dataset consisted a total of 6,720 slices with a resolution of 192 x 192 pixels for training (after removing slices where the mask was empty). Loss function was Combo loss (combination of Dice Loss with modified Cross-Entropy). The first technique was to use the pre-trained weights from the ImageNet dataset on encoder in U-Net network, with and without locked encoder weights, respectively, and compare the results with random weight initialization. In this case, we used only the FLAIR modality. Transfer learning has proven to increase the metrics from approximately 0.4 to 0.6. The difference between encoder with and without locked weights was about 0.02. The second proposed technique was to use a self-supervised context encoder with Generative Adversarial Networks (GAN) to pre-train the weights. This network used all three modalities also with the empty slices (23,040 slices in total). The purpose of GAN was to recreate the brain image, which was covered by a checkerboard. Weights learned during this training were later loaded for the encoder to apply to our segmentation problem. The following experiment did not show any improvement, with a DSC value of 0.29 and 0.09, with and without a locked encoder, respectively. Such a decrease in performance might have been caused by the use of weights pre-trained on two distant problems (segmentation and self-supervised context encoder) or by difficulty of the task considering the hugely unbalanced dataset.
|
host ::
RSS.