|
--- |
|
license: mit |
|
base_model: microsoft/deberta-v3-base |
|
tags: |
|
- generated_from_trainer |
|
metrics: |
|
- precision |
|
- recall |
|
- f1 |
|
- accuracy |
|
model-index: |
|
- name: AttackER |
|
results: [] |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
# Cyber-ThreaD/DeBERTa-v3-AttackER |
|
|
|
This model is a fine-tuned version of [microsoft/deberta-v3-base](https://huggingface.co/microsoft/deberta-v3-base) on an unknown dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 1.5468 |
|
- Precision: 0.4730 |
|
- Recall: 0.5569 |
|
- F1: 0.5115 |
|
- Accuracy: 0.7401 |
|
|
|
## Model description |
|
|
|
More information needed |
|
|
|
## Intended uses & limitations |
|
|
|
More information needed |
|
|
|
## Training and evaluation data |
|
|
|
More information needed |
|
|
|
## Training procedure |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 2e-05 |
|
- train_batch_size: 2 |
|
- eval_batch_size: 2 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- num_epochs: 10.0 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy | |
|
|:-------------:|:-----:|:-----:|:---------------:|:---------:|:------:|:------:|:--------:| |
|
| 1.7886 | 0.4 | 500 | 1.5075 | 0.1842 | 0.2103 | 0.1964 | 0.6169 | |
|
| 1.3644 | 0.81 | 1000 | 1.3342 | 0.2364 | 0.3056 | 0.2666 | 0.6492 | |
|
| 1.1181 | 1.21 | 1500 | 1.2655 | 0.2959 | 0.3585 | 0.3242 | 0.6812 | |
|
| 0.9833 | 1.61 | 2000 | 1.2368 | 0.2941 | 0.3902 | 0.3354 | 0.6778 | |
|
| 0.9036 | 2.01 | 2500 | 1.2682 | 0.3551 | 0.4021 | 0.3772 | 0.7023 | |
|
| 0.7102 | 2.42 | 3000 | 1.2176 | 0.3668 | 0.4590 | 0.4078 | 0.7159 | |
|
| 0.6868 | 2.82 | 3500 | 1.2170 | 0.3794 | 0.4683 | 0.4192 | 0.7147 | |
|
| 0.5671 | 3.22 | 4000 | 1.2603 | 0.3951 | 0.4881 | 0.4367 | 0.7259 | |
|
| 0.4878 | 3.63 | 4500 | 1.2460 | 0.3925 | 0.5093 | 0.4433 | 0.7333 | |
|
| 0.4942 | 4.03 | 5000 | 1.3147 | 0.4047 | 0.4802 | 0.4392 | 0.7284 | |
|
| 0.3812 | 4.43 | 5500 | 1.3308 | 0.4205 | 0.5146 | 0.4628 | 0.7351 | |
|
| 0.421 | 4.83 | 6000 | 1.3031 | 0.4275 | 0.5225 | 0.4702 | 0.7386 | |
|
| 0.3157 | 5.24 | 6500 | 1.3943 | 0.4132 | 0.5040 | 0.4541 | 0.7293 | |
|
| 0.3072 | 5.64 | 7000 | 1.4087 | 0.4303 | 0.5185 | 0.4703 | 0.7396 | |
|
| 0.3436 | 6.04 | 7500 | 1.4197 | 0.4461 | 0.5251 | 0.4824 | 0.7363 | |
|
| 0.2774 | 6.45 | 8000 | 1.4249 | 0.4275 | 0.5225 | 0.4702 | 0.7377 | |
|
| 0.2629 | 6.85 | 8500 | 1.4811 | 0.4580 | 0.5344 | 0.4933 | 0.7327 | |
|
| 0.2271 | 7.25 | 9000 | 1.5576 | 0.4733 | 0.5397 | 0.5043 | 0.7415 | |
|
| 0.235 | 7.66 | 9500 | 1.5468 | 0.4730 | 0.5569 | 0.5115 | 0.7401 | |
|
| 0.2415 | 8.06 | 10000 | 1.5956 | 0.4730 | 0.5437 | 0.5058 | 0.7433 | |
|
| 0.1826 | 8.46 | 10500 | 1.6168 | 0.4455 | 0.5410 | 0.4886 | 0.7413 | |
|
| 0.2083 | 8.86 | 11000 | 1.5866 | 0.4505 | 0.5423 | 0.4922 | 0.7413 | |
|
| 0.2169 | 9.27 | 11500 | 1.5974 | 0.4708 | 0.5437 | 0.5046 | 0.7468 | |
|
| 0.1747 | 9.67 | 12000 | 1.6219 | 0.4567 | 0.5437 | 0.4964 | 0.7405 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.36.0.dev0 |
|
- Pytorch 2.1.0+cu118 |
|
- Datasets 2.15.0 |
|
- Tokenizers 0.15.0 |
|
|
|
|
|
### Citing & Authors |
|
|
|
If you use the model kindly cite the following work |
|
|
|
``` |
|
@inproceedings{deka2024attacker, |
|
title={AttackER: Towards Enhancing Cyber-Attack Attribution with a Named Entity Recognition Dataset}, |
|
author={Deka, Pritam and Rajapaksha, Sampath and Rani, Ruby and Almutairi, Amirah and Karafili, Erisa}, |
|
booktitle={International Conference on Web Information Systems Engineering}, |
|
pages={255--270}, |
|
year={2024}, |
|
organization={Springer} |
|
} |
|
|
|
``` |
|
|
