The repo contains the code, datasets, and models for the paper "A Multiversion Programming Inspired Approach to Detecting Audio Adversarial Examples"
- Check "Summary_of_experiment_dataset_and_its_download_link.txt" and download audio files to do recognition. Please refer to "Scripts/Recognition/README.md". (For your convenience, the transcription of each audio file by each ASR are collected and organized for each experiment and placed in the corresponding subdirectories inside "Scripts" directory.)
- Follow "Scripts/README.md" to conduct the experiments reported in the DSN paper.
- The experiment results reported in the DSN paper are stored in "Experiment-Result" direcotory.