Please use this identifier to cite or link to this item:
https://dspace.kmf.uz.ua/jspui/handle/123456789/5775| Title: | Evolving Security Paradigms: How AI Transforms Threats and Defenses |
| Authors: | Daria Hurtova Hanna Sorokun |
| Keywords: | Artificial Intelligence (AI);security;ChatGPT |
| Issue Date: | 2025 |
| Publisher: | II. Rákóczi Ferenc Kárpátaljai Magyar Egyetem |
| Type: | dc.type.conferenceAbstract |
| Citation: | In Csernicskó István, Maruszinec Marianna, Molnár D. Erzsébet, Mulesza Okszána és Melehánics Anna (szerk.): A biztonság szerepe a határon átnyúló és nemzetközi együttműködésben. Nemzetközi tudományos és szakmai konferencia Beregszász, 2025. október 8–9. Absztraktkötet. Beregszász, II. Rákóczi Ferenc Kárpátaljai Magyar Egyetem, 2025. pp. 44-45. |
| Abstract: | Abstract. Artificial intelligence continues to reshape security landscapes around the world, enabling systems that are more adaptive, predictive, and efficient than ever before. In recent months, several breakthroughs have highlighted both the opportunities and fresh threats posed by AI in the security domain. One unexpected vulnerability discovered in ChatGPT’s Deep Research agent, known as “ShadowLeak,” allows attackers to siphon sensitive information without any visible user interaction. This zero-click, server-side flaw underscores the fact that AI tools themselves can be weak links in enterprises’ security posture. At the same time, generative AI attacks are accelerating at an alarming rate. A Gartner report from September 2025 found that 29% of organizations experienced attacks targeting their AI application infrastructure within the past year. Deepfake incidents – through audio, voice biometric spoofing, or video – are growing in prevalence, introducing new vectors for disinformation and fraud. Another concern is that malware is now being developed that can evade even advanced protection tools. In one experiment, a malware strain trained via reinforcement learning managed to bypass Microsoft Defender’s security checks around 8% of the time after only three months of training, using a relatively small budget. This kind of AI-generated malware shows attackers’ growing ability to automate and refine their methods very quickly. On the defense side, there are some promising developments. In Ukraine, for example, the Institute for Information Registration Problems is building an AI/ML-based system to assess the combat effectiveness of units in real time, forecast outcomes, and support decision-making under pressure. Ukraine’s military AI ecosystem has also benefitted from a large number of companies and academic programs: over 243 AI-focused companies, more than 307,000 specialists in IT and software, and 106 AI/ML academic programs across 42 higher education institutions. Despite infrastructural and resource challenges, such numbers show strong local capacity and commitment. |
| Description: | Teljes kiadvány: https://kme.org.ua/uk/publications/rol-bezpeki-v-transkordonnomu-ta-mizhnarodnomu-spivrobitnictvi/ |
| URI: | https://dspace.kmf.uz.ua/jspui/handle/123456789/5775 |
| ISBN: | 978-617-8143-50-3 (puhatáblás) 978-617-8143-51-0 (PDF) |
| metadata.dc.rights.uri: | http://creativecommons.org/licenses/by-nc-nd/3.0/us/ |
| Appears in Collections: | A biztonság szerepe a határon átnyúló és nemzetközi együttműködésben |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| Evolving_security_paradigms_how_transforms_threats_defenses_2025.pdf | In Csernicskó István, Maruszinec Marianna, Molnár D. Erzsébet, Mulesza Okszána és Melehánics Anna (szerk.): A biztonság szerepe a határon átnyúló és nemzetközi együttműködésben. Nemzetközi tudományos és szakmai konferencia Beregszász, 2025. október 8–9. Absztraktkötet. Beregszász, II. Rákóczi Ferenc Kárpátaljai Magyar Egyetem, 2025. pp. 44-45. | 10 MB | Adobe PDF | View/Open |
This item is licensed under a Creative Commons License



