Trustworthy AI

Lecturer: Prof. Dr.-Ing. Alexandra Dmitrienko

Contact: Dr. Venkata Marella

Start date: 16.04.2026

Lecture: Thursday, 14:00 – 16:00, SH 601

Exercise group 1: Thursday, 16:00 – 18:00, SH 601

Exercise group 2: Monday, 14:00 – 16:00, SM 311

Exercise group 3: Wednesday, 14:00 – 16:00, SM 311

Language: English

Moodle: Lehr-Moodle UDE

LSF: Trustworthy AI
 

Description:

This lecture provides a comprehensive introduction to the security, robustness, privacy, and safety of modern Artificial Intelligence (AI) systems, with a particular focus on deep learning. While AI systems are increasingly deployed in safety-critical and security-sensitive domains, they remain vulnerable to a broad spectrum of adversarial, privacy, and misuse-related threats. The course systematically studies these vulnerabilities and presents state-of-the-art defense mechanisms, risk management approaches, and protection techniques.

Students will learn how to analyze AI systems from a cybersecurity perspective across the entire AI lifecycle, from data collection and model training to deployment and post-deployment monitoring. The course integrates theoretical foundations with hands-on exercises, enabling students to implement attacks and defenses in practical machine learning settings.

After successful completion of the module, students will be able to:

• Explain core concepts of Artificial Intelligence, Deep Learning, and the AI lifecycle
• Identify and classify security, privacy, and safety risks in AI systems
• Understand threat models and adversarial capabilities in machine learning
• Implement and evaluate evasion, poisoning, and privacy attacks
• Analyze misuse and abuse risks in large language models (LLMs)
• Apply protection techniques such as adversarial defenses, watermarking, and secure aggregation
• Assess trade-offs between robustness, privacy, utility, and safety
• Critically evaluate emerging trends in trustworthy and secure AI