|
|
Methods for Realtime Voice Deepfakes Creation
Alakaev, Kambulat ; Pleško, Filip (oponent) ; Malinka, Kamil (vedoucí práce)
This thesis explores the possibility of achieving real-time voice deepfake generation using open-source tools. Through experiments, it was discovered that the generation rate of voice deepfakes is affected by the computing power of the devices running the speech creation tools. A deep learning model was identified to be capable of generating speech in near real time. However, limitations in the tool containing this model prevented continuous input data for real-time generation. To address this, a program was developed to overcome these limitations. The quality of the generated deepfakes was evaluated using both voice deepfake detection models and human online surveys. The results revealed that while the model could deceive detection models, it was not successful in fooling humans. This research highlights the accessibility of open-source voice synthesis tools and the potential for their misuse by individuals for fraudulent purposes.
|
|
|
Use of Diffusion Models in Deepfakes
Trúchly, Dominik ; Malinka, Kamil (oponent) ; Lapšanský, Tomáš (vedoucí práce)
A deepfake is a type of synthetic media created through sophisticated machine learning algorithms, particularly deep neural networks. As an example Generative adversarial neural networks (GANs), that are capable of generating images that are almost impossible for ordinary individuals to differentiate from genuine reality. Consequently, deepfake detection algorithms have been developed to address this growing concern. Leveraging advanced machine learning techniques, these algorithms analyze various features within images and videos to identify inconsistencies or anomalies indicative of manipulation. This thesis investigates the application of diffusion models, commonly utilized in digital image processing to enhance image quality by reducing noise and blurring, in bolstering the realism of deepfakes. By using these models, we test their effect on detecting deepfakes images using deepfake detectors.
|
|
|
Animal Identification Based on Biometric Information
Jančeková, Lucia ; Sakin, Martin (oponent) ; Dyk, Tomáš (vedoucí práce)
This thesis deals with wild animal identification, specifically wild boar, through photographs of their snouts. It focuses on the identification of individual animal, by utilizing the ridges found on the upper part of the snout. Within this work, a solution is designed and implemented, for the extraction of this biometric information and its comparison with other templates already stored in the system. The solution is tested on photographs from the same wild boar, as well as 49 other individuals.
|
|
|
Rozpoznávání izolovaných slov
Ondruška, Jiří ; Švrček, Martin (oponent) ; Kolářová, Jana (vedoucí práce)
Rozpoznávání lidské řeči v biometrických systémech je aktuální problematika, kterou se věda intenzivně zabývá. Mezi nejefektivnější metody spadá využití skrytých Markovových modelů. Při rozpoznávání izolovaných slov je pozornost zaměřena na získání charakteristických parametrů z řečových signálů umožňující co nejjednoznačnější identifikaci pomocí aplikace skrytých Markovových modelů. Tato práce se zabývá biometrickými systémy, jejich metodami a následně se zaměřuje na problematiku rozpoznávání izolovaných slov. Je navržen systém rozpoznávání metodou skrytých Markovových modelů, v němž jsou využity funkce systému Matlab. Návrh je zaměřen na získání charakteristických parametrů izolovaných slov, vytvoření kódové knihy prostřednictvím vektorové kvantizace, trénování modelů slov a nakonec vyhodnocení pravděpodobnosti shody pro pozorované slovo a daný model slova. Úspěšnost rozpoznání pro jednoho řečníka dosahuje 40%.
|
|
|
System for Recognition of 3D Hand Geometry
Svoboda, Jan ; Mráček, Štěpán (oponent) ; Drahanský, Martin (vedoucí práce)
In the last decade, there has been an increased interest in using 3D data for biometric person recognition. Perhaps the most widely researched application is 3D face recognition, where several commercial products are currently available on the market. There have been some research works on the 3D hand recognition as well, however, no commercially viable systems are currently known. Independently, in the recent years inexpensive 3D sensors have become a commodity, potentially enabling a wide range of 3D biometric applications. The main goal of this work is to develop a functioning prototype of a touchless 3D hand recognition system based on a new cheap RealSense 3D camera developed by Intel. One of the challenges in using the RealSense camera is that due to this small form factor, it produces relatively low quality samples in comparison to the more expensive acquisition hardware used in the previous research on the 3D hand biometrics. We analyze the robustness of different 2D and 3D features and study several methods for their fusion. We evaluate the performance of the system, showing that it achieves results comparable with the state-of-the-art.
|
|
|
Biometric System Security Using Blockchain Technology
Žiška, Marek ; Drahanský, Martin (oponent) ; Malaník, Petr (vedoucí práce)
This work analyzes existing protocols used to reach consensus in blockchain technologies, describes the concepts of biometric systems, identifies their security threats, and presents existing solutions for securing the biometric systems. Based on these findings, designs a decentralized version of the biometric system that makes use of the weighted PBFT protocol and the blockchain to improve the security of feature extraction and matching module of the classical biometric system. Blockchain is characterized as a system of recording information that assures immutability. The direct use of blockchain to secure sensitive data, such as biometric data, is not expected to be the most appropriate use, but its integration within the processes of individual components of biometric systems appears to be a good option. The proposed system was implemented and tested with a variety of test scenarios. Evaluation has shown that our design managed to mitigate direct attacks on the focused components and attacks on the channels that connect them together and the channel that connects the matcher to the external application.
|
|
|
Multikamerová biometrická brána pro identifikaci osob
Kosík, Dominik ; Orság, Filip (oponent) ; Goldmann, Tomáš (vedoucí práce)
Tato práce řeší vytvoření biometrické brány pro identifikaci osob. Identifikace probíhá za pomocí 5 barevných kamer a IR kamery. IR kamera zajišťuje detekci osoby a následně se ze snímku barevných kamer vytváří 3D model obličeje osoby. Na základě tohoto modelu se provádí identifikace. Jelikož při vytváření samotného 3D modelu docházelo k nepřesnostem, což má vliv na rozpoznání osoby, není výsledná identifikace dostatečně přesná. Z toho důvodu je zapotřebí upravit algoritmy zpracovávající 3D model, a tak dosáhnout dostatečné přesnosti.
|
|
|
Identifikace osob pomocí biometrie sítnice
Klimešová, Lenka ; Mézl, Martin (oponent) ; Odstrčilík, Jan (vedoucí práce)
Tato diplomová práce se zabývá identifikací osob pomocí biometrie sítnice. Cévní řečiště sítnice je neměnné a unikátní pro každého jedince, což jej předurčuje pro biometrické účely. První část práce se zabývá problematikou biometrie, biometrických systémů a hodnocením jejich spolehlivosti. Je uveden princip snímání pomocí experimentálního video oftalmoskopu a provedena rešerše využití snímků sítnice pro biometrii, metod extrakce příznaků a srovnávacích metrik. Dále jsou navrženy dva algoritmy pro využití zadaných dat a realizovány v programovém prostředí MATLAB®. Úspěšnost metod je otestována a vyhodnocena na snímcích z experimentálního video oftalmoskopu a na veřejně dostupných databázích STRaDe a DRIVE.
|
|
|
Generative Adversarial Networks Applied for Privacy Preservation in Bio-Metric-Based Authentication and Identification
Mjachky, Ľuboš ; Malinka, Kamil (oponent) ; Homoliak, Ivan (vedoucí práce)
Biometric-based authentication systems are getting broadly adopted in many areas. However, these systems do not allow participating users to influence the way their data will be used. Furthermore, the data may leak and can be misused without the users' knowledge. In this thesis, we propose a new authentication method which preserves the privacy of an individual and is based on a generative adversarial network (GAN). Concretely, we suggest using the GAN for translating images of faces to a visually private domain (e.g., flowers or shoes). Classifiers, which are used for authentication purposes, are then trained on the images from the visually private domain. Based on our experiments, the method is robust against attacks and still provides meaningful utility.
|
|
|
Security Implications of Deepfakes in Face Authentication
Šalko, Milan ; Goldmann, Tomáš (oponent) ; Firc, Anton (vedoucí práce)
Deepfakes, media generated by deep learning that are indistinguishable to humans from real ones, have experienced a huge boom in recent years. Several dozen papers have already been written about their ability to fool people. Equally, if not more, serious, may be the problem of the extent to which facial and voice recognition systems are vulnerable to them. The misuse of deepfakes against automated facial recognition systems can threaten many areas of our lives, such as finances and access to buildings. This topic is essentially an unexplored problem. This thesis aims to investigate the technical feasibility of an attack on facial recognition. The experiments described in the thesis show that this attack is not only feasible but moreover, the attacker does not need many resources for the attack. The scope of this problem is also described in the work. The conclusion also describes some proposed solutions to this problem, which may not be difficult to implement at all.
|
host ::
RSS.