Αντιμετώπιση των deepfakes μέσω προηγμένων τεχνολογιών κυβερνοασφάλειας

Κουτσογιάννης, Παναγιώτης

Cybersecurity advanced technologies for competing against deepfakes

Master Thesis

Author

Κουτσογιάννης, Παναγιώτης

Date

2025-03

Abstract

The rapid development of artificial intelligence algorithms in recent years has introduced a new threat to the world of the internet, the realistic synthetic multimedia. These are multimedia such as videos, images, texts, which have been produced by artificial intelligence and can hardly be categorized as artificial or not. The purpose of this paper is to highlight the importance of the problem of the distribution of artificial videos and images by highlighting the way in which society can be aoected through disinformation, while also delving into the technical interventions that cybersecurity can ooer to address these threats. Initially, the necessary background is given for the definition of deepfakes by citing examples of threats through real incidents of the past. Then, through the analysis of how deepfakes are created, the basis for the continuation of the work is set, more specifically, the basic algorithms that create synthetic content Generative Adversarial Networks and Autoencoders are analyzed. Focusing on a system that will constitute a comprehensive solution for the authentication and ensuring the integrity of genuine material that has not been processed, an analysis of the most basic synthetic content detection techniques in the literature is made, focusing on the Xception model. Then, through watermarking techniques, it is explained how the authentication of the material and the assurance of the integrity of multimedia can be achieved through modern techniques. Based on these, a code was created that detects synthetic content through the Xception model and then with the LSB technique an encrypted watermark is introduced through which the integrity of the content and authenticity are ensured. Finally, a brief analysis is provided regarding the data provided for training artificial intelligence models and how they can be protected by the GDPR and the Artificial Intelligence Act.

Postgraduate Studies Programme

Ασφάλεια Ψηφιακών Συστημάτων

Department

Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα Ψηφιακών Συστημάτων

Number of pages

Language

Greek

URI

https://dione.lib.unipi.gr/xmlui/handle/unipi/17586

Collections

Τμήμα Ψηφιακών Συστημάτων

Show full item record

Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα