Michael Aerni

Michael Aerni is a doctoral researcher at the Secure and Private AI (SPY) Lab within ETH Zurich, focusing on privacy and security challenges in machine learning and AI systems. His work bridges theoretical insights with practical applications.

• Education:
  • Doctoral student in Computer Science (ongoing), ETH Zurich
  • MSc in Computer Science, 2022, ETH Zurich
  • BSc in Computer Science, 2017, FHNW Windisch

His research emphasizes understanding and mitigating unintended memorization in large language models and evaluating empirical privacy defenses. Key themes include:

• Privacy leakage quantification
• Membership inference attack analysis
• Inductive bias impact on interpolation
• Robust margin behavior in noiseless data

Publications reveal trends in privacy-preserving machine learning, with a focus on exposing limitations of heuristic defenses and proposing rigorous evaluation protocols. Collaborations span institutions like ETH Zurich, with advisors like Florian Tramèr.