|
|
---
|
|
|
license: mit
|
|
|
language:
|
|
|
- en
|
|
|
tags:
|
|
|
- random-forest
|
|
|
- binary-classification
|
|
|
- prompt-injection
|
|
|
- security
|
|
|
datasets:
|
|
|
- imoxto/prompt_injection_cleaned_dataset-v2
|
|
|
- reshabhs/SPML_Chatbot_Prompt_Injection
|
|
|
- Harelix/Prompt-Injection-Mixed-Techniques-2024
|
|
|
- JasperLS/prompt-injections
|
|
|
- fka/awesome-chatgpt-prompts
|
|
|
- rubend18/ChatGPT-Jailbreak-Prompts
|
|
|
metrics:
|
|
|
- recall
|
|
|
- precision
|
|
|
- f1
|
|
|
- auc
|
|
|
---
|
|
|
|
|
|
# Model Description
|
|
|
|
|
|
The purpose of our trained Random Forest models is to identify malicious prompts given the prompt embeddings derived from [OpenAI](https://huggingface.co/datasets/ahsanayub/malicious-prompts-openai-embeddings), [OctoAI](https://huggingface.co/datasets/ahsanayub/malicious-prompts-octoai-embeddings), and [MiniLM](https://huggingface.co/datasets/ahsanayub/malicious-prompts-minilm-embeddings). The models are trained with 373,598 benign and malicious prompts. We split this dataset into 80% training and 20% test sets. To ensure equal proportion of the malicious and benign labels across splits, we use stratified sampling.
|
|
|
|
|
|
Embeddings consist of fixed-length numerical representations. For example, OpenAI generates an embedding vector consisting of 1,536 floating-point numbers for each prompt. Similarly, the embedding datasets for OctoAI and MiniLM consist of 1,027 and 387 features, respectively.
|
|
|
|
|
|
## Model Evaluation
|
|
|
|
|
|
The binary classification performance of embedding-based random forest models is shared below:
|
|
|
|
|
|
| Embedding | Precision | Recall | F1-score | AUC |
|
|
|
|-----------|-----------|--------|----------|-------|
|
|
|
| OpenAI | 0.867 | 0.867 | 0.867 | 0.764 |
|
|
|
| OctoAI | 0.849 | 0.853 | 0.851 | 0.731 |
|
|
|
| MiniLM | 0.849 | 0.853 | 0.851 | 0.730 |
|
|
|
|
|
|
## How To Use The Model
|
|
|
|
|
|
We have shared three versions of random forest models in this repository. We used the following embedding models: `text-embedding-3-small` from OpenAI, and the open-source models `gte-large` hosted on OctoAI, as well as the well-known `all-MiniLM-L6-v2`. Therefore, you need to covert the prompts to its respective embeddings before querying the model to obtain its prediction: `0` for benign and `1` for malicous.
|
|
|
|
|
|
## Citing This Work
|
|
|
Our implementation, along with the curated datasets used for evaluation, is available on [GitHub](https://github.com/AhsanAyub/malicious-prompt-detection). Additionaly, if you use our implementation for scientific research, you are highly encouraged to cite [our paper](https://arxiv.org/abs/2410.22284).
|
|
|
|
|
|
```
|
|
|
@article{ayub2024embedding,
|
|
|
title={Embedding-based classifiers can detect prompt injection attacks},
|
|
|
author={Ayub, Md Ahsan and Majumdar, Subhabrata},
|
|
|
booktitle={CAMLIS},
|
|
|
year={2024}
|
|
|
}
|
|
|
``` |
