Curriculum
Computer Science for Societal Challenges and Innovation, XL series
Grant sponsor
Unismart - Cariparo Foundation- Intesa Sanpaolo
Supervisor
Mauro Conti
Co-supervisor
to be defined
Contact
anastasiia.belousova@phd.unipd.it
Project description
Artificial intelligence (AI) is increasingly integrated into complex software systems, where AI components interact with traditional software and hardware. However, current security evaluation methods still assess AI models in isolation and overlook how attacks can exploit interactions between different parts of an AI system. This project aims to develop a systematic framework for evaluating the security of AI systems as a whole, identifying new types of threats that arise from the interplay of multiple components. Using a multi-stage AI-based Windows malware detection pipeline as a case study, the research seeks to model these threats, understand how they can impact system behavior, and propose methodologies to improve the security and reliability of real-world AI systems.