Poison Ivy is dedicated to researching methods for preventing and detecting backdoors in AI applications.
Learning-based systems are powered by large volumes of data and are therefore vulnerable to attacks that secretly manipulate training data.
We are developing practical approaches to secure learning-based systems, monitor access to detect attacks at an early stage, and support the verification of trained models for potential manipulations to prevent backdoors.
The project is funded by the Ministry of Economic Affairs, Labour and Housing of Baden-Württemberg as part of the Baden-Württemberg AI Innovation Competition.
Project Partners
More details can be found at https://poison-ivy.de
