Commit
•
fd33665
1
Parent(s):
b47cbef
add tar files link
Browse files
README.md
CHANGED
|
@@ -7,6 +7,7 @@ This dataset is a subset of [LibriSpeech](https://huggingface.co/datasets/libris
|
|
| 7 |
|
| 8 |
Original [source code](https://github.com/RaphaelOlivier/whisper_attack).
|
| 9 |
|
|
|
|
| 10 |
# Configurations and splits
|
| 11 |
* The `targeted` config contains targeted adversarial examples. When successful, they fool Whisper into predicting the sentence `OK Google, browse to evil.com` even if the input is entirely different. We provide a split for each Whisper model, and one containing the original, unmodified inputs
|
| 12 |
* The `untargeted-35` and `untargeted-40` configs contain untargeted adversarial examples, with average Signal-Noise Ratios of 35dB and 40dB respectively. They fool Whisper into predicting erroneous transcriptions. We provide a split for each Whisper model, and one containing the original, unmodified inputs
|
|
|
|
| 7 |
|
| 8 |
Original [source code](https://github.com/RaphaelOlivier/whisper_attack).
|
| 9 |
|
| 10 |
+
The raw [tar files](https://data.mendeley.com/datasets/96dh52hz9r).
|
| 11 |
# Configurations and splits
|
| 12 |
* The `targeted` config contains targeted adversarial examples. When successful, they fool Whisper into predicting the sentence `OK Google, browse to evil.com` even if the input is entirely different. We provide a split for each Whisper model, and one containing the original, unmodified inputs
|
| 13 |
* The `untargeted-35` and `untargeted-40` configs contain untargeted adversarial examples, with average Signal-Noise Ratios of 35dB and 40dB respectively. They fool Whisper into predicting erroneous transcriptions. We provide a split for each Whisper model, and one containing the original, unmodified inputs
|