E³ - ECR Master Class: Imaging Informatics and Artificial Intelligence
E³ 226a - Cybersecurity in AI-driven radiology systems
Lectures
1
Chairperson's introduction
05:00Anton Becker, New York City / US
2
Security threats posed by AI systems in radiology
15:00Brendan S Kelly, United Kingdom / UK
3
Can large language models (LLMs) be a security threat?
15:00Tugba Akinci D’Antonoli, Basel / CH
4
Generative AI: identifying threats and aligning risks to data
15:00Renato Cuocolo, Naples / IT
5
Panel discussion: Dissecting a cyberattack
10:00Panel discussion: Dissecting a cyberattack
5 min
Chairperson's introduction
Anton Becker, New York City / United States
15 min
Security threats posed by AI systems in radiology
Brendan S Kelly, United Kingdom / United Kingdom
- To identify the cybersecurity vulnerabilities in AI-powered radiology systems.
- To describe how integrating AI tools into radiological workflows introduces new points of vulnerability.
- To evaluate the real-world of compromised cybersecurity incidents and case studies.
15 min
Can large language models (LLMs) be a security threat?
Tugba Akinci D’Antonoli, Basel / Switzerland
- To recognise the unique risks that LLMs introduce to radiology systems.
- To differentiate between traditional cybersecurity threats and those specifically associated with LLMs.
- To explore the mitigation strategies against LLM-related breaches.
15 min
Generative AI: identifying threats and aligning risks to data
Renato Cuocolo, Naples / Italy
- To examine how generative AI models can generate misleading images or manipulate patient data.
- To identify generative AI-related risks for regulatory and institutional environments.
- To propose mitigation strategies and governance controls.
10 min
Panel discussion: Dissecting a cyberattack