MAIA: A Taxonomy for AI-Generated Attacks

Le résumé fourni par la source

Significant parts of the cyber threat landscape have changed with the usage of AI based tools. The most significant change is that AI is used by threat actors to trick people that lead to a variety of cyber incidents. Thus, a new taxonomy is needed to categorize these AI-based modern-day attacks to provide key insights into the modalities involved, what causes cybercrimes can they lead to and propose the mitigation strategies accordingly. In our paper, we are working to bridge this gap and propose a taxonomy describing the AI driven attack methodologies used by threat actors in the current cyber space. This taxonomy outlines the various modalities exploited by attackers, including AI-generated videos, audio, images, and text. A required dimension in this classification is whether AI is used in a static or dynamic manner-whether the attack occurs in real-time or not. Our analysis is substantiated by a dataset derived from open-source intelligence (OSINT), which serves to validate the claims and classifications presented. This systematic approach aims to enhance our understanding of AI-enabled cyber threats, the diverse ways in which they are executed across various domains, and the possible mitigation strategies.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.