|
|
Evaluation of Sources of Human Speech for Deepfake Creation
Frič, Michal ; Malinka, Kamil (oponent) ; Firc, Anton (vedoucí práce)
Voice deepfakes, powered by rapid advancements in artificial intelligence and machine learning, represent a dual-edge technology with significant benefits and risks. These synthetic voice outputs are increasingly realistic due to the easy access to vast amounts of digital speech data from various sources. This thesis analyses these sources’ suitability for creating convincing deepfakes. We identified and evaluated numerous speech sources and developed methodologies for assessing their quality, accessibility, diversity, and update frequency. The evaluation extended to analyzing the impact of source characteristics on deepfake quality and the effectiveness of detection by software and human evaluators. Findings indicate that all identified sources can provide sufficiently high-quality recordings to create high-quality deepfakes, often indistinguishable. Additionally, they highlight each source’s particular strengths and weaknesses (measured properties) grade. An anomaly in detection software was discovered, allowing deepfakes to be modified to evade detection. Furthermore, less than 10 seconds of human speech could suffice to create a high-quality deepfake, directly correlating the length and quality of input recordings to the fidelity of the output. The thesis concludes with a discussion of the risks associated with these sources and proposes measures for prevention and mitigation.
|
|
|
Communication System for the Volunteer Fire Department
Čajka, Peter ; Firc, Anton (oponent) ; Malinka, Kamil (vedoucí práce)
The process of gathering volunteer firefighters in case of an emergency is often not optimal. In critical situations, every second counts. The goal of this thesis is to create a mobile application that simplifies the process of informing volunteer firefighters about an emergency event. The resulting product will be a mobile application for the Android operating system, capable of rapidly sending SMS messages to all members of the volunteer fire department in case of an emergency and subsequently receiving their responses.
|
|
|
Reversibility of Voice Change Methods
Lička, Zbyněk ; Firc, Anton (oponent) ; Malinka, Kamil (vedoucí práce)
State-of-the-art voice-changing methods allow inexperienced users to create convincing voice recordings of famous individuals with just a few seconds of recorded speech. There are two major approaches to voice generation: voice conversion and text-to-speech. Voice conversion methods require the user to input source speech to be converted to the target voice. A trend with voice conversion methods, especially those requiring only mere seconds of reference speech, has been restricting the amount of information about the original speaker in the converted speech. This work focuses on studying the amount of information extractable about the original speaker from artificial speech and potentially reconstructing the original speech. The results of this work shed light on an unstudied property of voice-changing methods.
|
|
|
Detection of fraudulent real estate advertisements
Frátrik, Tomáš ; Homoliak, Ivan (oponent) ; Firc, Anton (vedoucí práce)
The advent of the Internet has led to a rise in the number of Internet users. Internet ad fraud refers to a form of online crime where individuals or businesses publish deceptive real estate advertisements to defraud unsuspecting victims. The purpose of this thesis is to provide an overview of real estate ad frauds found on websites that allow anyone to post real estate ads on their websites. And also provide a tool, that determines the probability of fraudulent advertisement of real estate chosen by user from supported web portals.
|
|
|
Creating Novel Deepfake Speech Dataset
Sztolarik, Maroš ; Homoliak, Ivan (oponent) ; Firc, Anton (vedoucí práce)
In the recent years, deepfake technology has advanced to a point where it can convincingly mimic human speech, posing significant challenges in distinguishing between real and synthetic voices. In this thesis, we introduce a novel dataset comprising speech deepfakes generated using diffusion models. This dataset, created with two sophisticated text-to-speech tools, DiffSpeech and ProDiff, aims to provide insight into the threat that these new tools pose. Two more datasets are created with more mature tools, Glow-TTS and Tacotron2, to provide a point of comparison. Then all the generated samples are analyzed through two deepfake detectors in order to provide a direct comparison into how much of a threat each tool is to these detectors. The results show that even though the tools utilizing the diffusion models are threatening, the use of diffusion models did not provide these tools any meaningful advantage in evading the detection.
|
|
|
Support Tools for Verifying Human Ability to Detect Deepfakes
Potančok, Patrik ; Malinka, Kamil (oponent) ; Firc, Anton (vedoucí práce)
The aim of this thesis is to create a web application using PHP and MySQL, that will test the human ability to detect deepfake recordings while collecting their data like date of birth, native language, proficiency in other languages, how many times and how long did they listen to a recording and a number of correct answers. This application includes management of the recordings and the users and the ability to export user data in CSV format. The application was implemented using Laravel, Vue.js and MySQL.
|
|
|
Vyhodnocení kvality a dostupnosti zdrojů obrazových médií pro tvorbu deepfakes
Andrýsek, Kryštof ; Malinka, Kamil (oponent) ; Firc, Anton (vedoucí práce)
Umělá inteligence je v posledních letech na vzestupu a začíná se stávat běžnou součástí našich životů. Tato práce se zaměřuje na nebezpečí související s obrazovými deepfakes. V současnosti se totiž ukazuje, že se netýká pouze vysoce postavených politických osobností nebo hollywoodských hvězd, ale může ovlivnit život kohokoliv z nás. K tomu přispívá mimo jiné zvyšující se množství aplikací, s jejichž pomocí dokáže deepfake vytvořit i člověk bez odborných znalostí a zkušeností. Primárním úkolem této práce je vyhodnotit zdroje, které lze využít k získání obrazových médií vhodných pro tvorbu obličejových deepfakes, a navrhnout metody, pomocí nichž je možné zhodnotit kvalitu těchto zdrojů z pohledu vhodnosti pro tvorbu deepfakes. Další cíl představuje identifikace skupin lidí, jimž hrozí vyšší riziko v souvislosti s obrazovými deepfakes, což může napomoci dokonalejší prevenci a obecnému prohloubení znalostí. Práce nabízí rozbor zdrojů použitelných k tvorbě obličejových deepfakes, navrhuje metody hodnocení jejich kvality, detailní vyhodnocení zdrojů pomocí těchto metod a analýzu zranitelnosti jednotlivých skupin osob. Nakonec je pomocí dotazníkového šetření, zaměřeného především na zdroj sociální sítě, zkoumáno, v jakém množství a jaký druh fotografií sdílejí běžní uživatelé nejčastěji, a na základě těchto údajů jsou upřesněny informace o rizikových skupinách.
|
|
|
Protection against profiling with random multilingual search
Klusáček, Jan ; Zbořil, František (oponent) ; Firc, Anton (vedoucí práce)
Search engines must first create a user profile to deliver tailored results. However, in order to create such a profile, they need to collect a significant amount of information about users, which they obtain by monitoring their activities. As a result, privacy may be compromised and search results may become biased and subjective. This thesis addresses the issue of user profiling in online search engines and aims to investigate whether random multilingual search can be used to stop or at least mitigate the profiling process. The results obtained by conducting the experiments suggest that although the search results hardly change when random search is used, the search engine identified different interests in users who used random search compared to other users. In addition, this work experimentally verified that to reduce user profiling, it is better to use fewer languages and a higher ratio of random searches to real user searches. Among other things, random searches were also found to change already identified interests in users who had been using the search engine for some time without random searching.
|
|
|
Differential-based deepfake speech detection
Staněk, Vojtěch ; Černocký, Jan (oponent) ; Firc, Anton (vedoucí práce)
Deepfake speech technology, which can create highly realistic fake audio, poses significant challenges, from enabling multi-million dollar scams to complicating legal evidence's reliability. This work introduces a novel method for detecting such deepfakes by leveraging bonafide speech samples. Unlike previous strategies, the approach uses trusted ground truth speech samples to identify spoofs, providing critical information that common methods lack. By comparing the bonafide samples with potentially manipulated ones, the aim is to effectively and reliably determine the authenticity of the speech. Results suggest that this innovative approach could be a valuable tool in identifying deepfake speech, especially recordings created using Voice Conversion techniques, offering a new line of defence against this emerging threat.
|
|
|
Advanced Web-based Tool for Managing Security Correlation Rules and Cybersecurity Responses
Hemza, Martin ; Firc, Anton (oponent) ; Malinka, Kamil (vedoucí práce)
The aim of this bachelor thesis is to develop an advanced web-based tool for managing security correlation rules and cybersecurity responses, focusing on SIEM and SOAR technologies. The reason for this thesis is the lack of a standardized resource format for these technologies. As part of this work, these formats were found and a web tool was created to manage and design them. The interface for managing SOAR scenarios includes a visualization in the form of a decision tree. The application uses a microservices architecture with integration of the Git versioning system. Testing included a description of an attack and the use of the developed tool. The created tool allows security analysts to quickly design and manage resources for detecting and responding to security threats.
|
host ::
RSS.