This repository contains the code for the paper "FDI: Attack Neural Code Generation Systems through User Feedback Channel"
The data for the prompt injection experiment is available in the prompt_search/dataset
folder, while the data for the continual learning experiments can be downloaded from this link.
For the continual learning experiments, the running scripts are available in the continual_learning/scripts
.
Please revise the scripts to set the correct paths to the data.
For the prompt injection experiments, it can be run using the following command:
python run.py