Thorsten Holz — accepted AI-conference papers

2026

An Empirical Study of Memory Poisoning Defenses for LLM Agents

Qianshan Wei, Tengchao Yang, Yaochen Wang, Xinfeng Li, Lijun Li, Zhenfei Yin +4

ICML 2026poster

Large Language Model (LLM) agents use memory to learn from past interactions. However, this reliance on memory introduces a critical security risk: an adversary can inject seemingly harmless records into an agent's memory to manipulate its future behavior. This vulnerability is characterized by two …

Cited by 0 Source ↗Scholar ↗

2020

Leveraging Frequency Analysis for Deep Fake Image Recognition

Joel Frank, Thorsten Eisenhofer, Lea Schönherr, Asja Fischer, Dorothea Kolossa, Thorsten Holz

ICML 2020poster

Deep neural networks can generate images that are astonishingly realistic, so much so that it is often hard for humans to distinguish them from actual photos. These achievements have been largely made possible by Generative Adversarial Networks (GANs). While deep fake images have been thoroughly inv…

Cited by 708 Source ↗PDF ↗code ↗Scholar ↗