معرفی
Dr. Yuanyuan Yuan is a Researcher at the Department of Computer Science, ETH Zurich, Switzerland, based at CNB H 104.1, Universitätstrasse 6, 8092 Zurich. Her work bridges computer security and machine learning with a focus on practical vulnerabilities in deployed AI systems.
Her research centers on exposing and mitigating security flaws in deep learning deployments, particularly targeting trusted execution environments (TEEs) and on-device inference systems. Key contributions include pioneering side-channel attacks against TEE-shielded neural networks (CipherSteal, HyperTheft), bit-flip attack surfaces in DNN executables (BitShield), and novel testing methodologies for neural network robustness. She investigates cache/timing side channels, ciphertext analysis, privacy leakage in partitioned ML, and concept-based explainability.
Analysis of her 2023-2025 publications reveals a cohesive focus on offensive security research for AI infrastructures, with consistent contributions to top venues in security and machine learning. Her work demonstrates expertise in low-level system interactions (memory, cryptography) applied to ML security, spanning attack vectors, defensive mechanisms, and validation frameworks. The research trajectory shows increasing sophistication in exploiting hardware-software interfaces while developing practical hardening techniques for real-world deployments.
