Faculty
Adam Dziedzic ist Tenure-Track Faculty am CISPA, wo er die SprintML-Gruppe mit dem Forschungsschwerpunkt „Sicheres, privates, robustes, interpretierbares und vertrauenswürdiges maschinelles Lernen“ mitleitet. Bevor er zum CISPA kam, war er Postdoktorand am Vector Institute und an der University of Toronto sowie Mitglied des CleverHans Lab unter der Betreuung von Prof. Nicolas Papernot. Er promovierte in Informatik an der University of Chicago, wo er von Prof. Sanjay Krishnan betreut wurde und sich mit der Komprimierung von Eingaben und Modellen für adaptive und robuste neuronale Netze befasste.
International Conference on Machine Learning (ICML) Unlocking Post-hoc Dataset Inference with Synthetic Data
International Conference on Machine Learning (ICML) Efficient and Privacy-Preserving Soft Prompt Transfer for LLMs
ACM Conference on Computer and Communications Security (CCS) Secure Noise Sampling for Differentially Private Collaborative Learning
International Conference on Machine Learning (ICML) Privacy Attacks on Image AutoRegressive Models
International Conference on Learning Representations (ICLR) Precise Parameter Localization for Textual Generation in Diffusion Models
National Conference of the American Association for Artificial Intelligence (AAAI) Differentially Private Prototypes for Imbalanced Transfer Learning
International Conference on Learning Representations (ICLR) Captured by Captions: On Memorization and its Mitigation in CLIP Models
International Conference on Learning Representations (ICLR) Differentially Private Federated Learning with Time-Adaptive Privacy Spending
European Conference on Artificial Intelligence (ECAI) Efficient Model-Stealing Attacks Against Inductive Graph Neural Networks
NeurIPS-Workshop (NeurIPS-W) Auditing Empirical Privacy Protection for Adaptations of Large Language Models