|
|
--- |
|
|
library_name: transformers |
|
|
license: mit |
|
|
base_model: microsoft/mdeberta-v3-base |
|
|
tags: |
|
|
- prompt-injection |
|
|
- injection |
|
|
- security |
|
|
- llm-security |
|
|
- generated_from_trainer |
|
|
metrics: |
|
|
- accuracy |
|
|
- precision |
|
|
- recall |
|
|
- f1 |
|
|
model-index: |
|
|
- name: mdeberta-v3-base-prompt-injection |
|
|
results: [] |
|
|
--- |
|
|
|
|
|
# mdeberta-v3-base-prompt-injection |
|
|
|
|
|
This model is a fine-tuned version of [microsoft/mdeberta-v3-base](https://huggingface.co/microsoft/mdeberta-v3-base) on a combination of [jackhhao/jailbreak-classification](https://huggingface.co/datasets/jackhhao/jailbreak-classification), [deepset/prompt-injections](https://huggingface.co/datasets/deepset/prompt-injections/viewer/default/test?views%5B%5D=test), a custom datasets containing known attacks, and injections nested in legitimate content like websites and articles. |
|
|
|
|
|
## Usage |
|
|
|
|
|
```Python |
|
|
from transformers import pipeline |
|
|
|
|
|
classifier = pipeline( |
|
|
"text-classification", |
|
|
model="proventra/mdeberta-v3-base-prompt-injection" |
|
|
) |
|
|
|
|
|
print(classifier("Your text to scan")) |
|
|
``` |
|
|
|
|
|
## Use in Proventra Core |
|
|
[proventra-core](https://github.com/proventra/proventra-core) python library |
|
|
|
|
|
check out [Proventra](https://www.proventra-ai.com) |
