|
--- |
|
license: apache-2.0 |
|
base_model: jackaduma/SecBERT |
|
tags: |
|
- generated_from_trainer |
|
metrics: |
|
- precision |
|
- recall |
|
- f1 |
|
- accuracy |
|
model-index: |
|
- name: Cyber-ThreaD/SecBERT-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/SecBERT-AttackER |
|
|
|
This model is a fine-tuned version of [jackaduma/SecBERT](https://huggingface.co/jackaduma/SecBERT) on an unknown dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 1.6932 |
|
- Precision: 0.3931 |
|
- Recall: 0.4987 |
|
- F1: 0.4397 |
|
- Accuracy: 0.7295 |
|
|
|
## 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 | |
|
|:-------------:|:-----:|:-----:|:---------------:|:---------:|:------:|:------:|:--------:| |
|
| 1.7927 | 0.4 | 500 | 1.5607 | 0.0956 | 0.0780 | 0.0859 | 0.6139 | |
|
| 1.3551 | 0.81 | 1000 | 1.3530 | 0.2064 | 0.2381 | 0.2211 | 0.6495 | |
|
| 1.0432 | 1.21 | 1500 | 1.3107 | 0.2269 | 0.3082 | 0.2614 | 0.6740 | |
|
| 0.8468 | 1.61 | 2000 | 1.2497 | 0.2447 | 0.3373 | 0.2836 | 0.6767 | |
|
| 0.7775 | 2.01 | 2500 | 1.2710 | 0.2895 | 0.3730 | 0.3260 | 0.6939 | |
|
| 0.5374 | 2.42 | 3000 | 1.3020 | 0.3006 | 0.4048 | 0.3450 | 0.7044 | |
|
| 0.5071 | 2.82 | 3500 | 1.2614 | 0.2959 | 0.4048 | 0.3419 | 0.7081 | |
|
| 0.4237 | 3.22 | 4000 | 1.3251 | 0.3367 | 0.4405 | 0.3817 | 0.7166 | |
|
| 0.3597 | 3.63 | 4500 | 1.3853 | 0.3423 | 0.4524 | 0.3897 | 0.7125 | |
|
| 0.3632 | 4.03 | 5000 | 1.4156 | 0.3559 | 0.4524 | 0.3984 | 0.7127 | |
|
| 0.2589 | 4.43 | 5500 | 1.4472 | 0.3504 | 0.4709 | 0.4018 | 0.7173 | |
|
| 0.323 | 4.83 | 6000 | 1.3997 | 0.3452 | 0.4603 | 0.3946 | 0.7222 | |
|
| 0.2167 | 5.24 | 6500 | 1.5194 | 0.3467 | 0.4590 | 0.3950 | 0.7233 | |
|
| 0.2363 | 5.64 | 7000 | 1.5585 | 0.3507 | 0.4722 | 0.4025 | 0.7222 | |
|
| 0.2721 | 6.04 | 7500 | 1.5420 | 0.3715 | 0.4854 | 0.4209 | 0.7210 | |
|
| 0.2073 | 6.45 | 8000 | 1.5878 | 0.3536 | 0.4854 | 0.4091 | 0.7147 | |
|
| 0.2021 | 6.85 | 8500 | 1.6637 | 0.3722 | 0.4854 | 0.4214 | 0.7197 | |
|
| 0.1648 | 7.25 | 9000 | 1.6724 | 0.3795 | 0.4788 | 0.4234 | 0.7255 | |
|
| 0.1927 | 7.66 | 9500 | 1.6891 | 0.3801 | 0.4947 | 0.4299 | 0.7245 | |
|
| 0.1958 | 8.06 | 10000 | 1.6774 | 0.3937 | 0.4974 | 0.4395 | 0.7281 | |
|
| 0.1508 | 8.46 | 10500 | 1.7379 | 0.3815 | 0.4854 | 0.4272 | 0.7259 | |
|
| 0.184 | 8.86 | 11000 | 1.7001 | 0.3863 | 0.5013 | 0.4364 | 0.7277 | |
|
| 0.1696 | 9.27 | 11500 | 1.6932 | 0.3931 | 0.4987 | 0.4397 | 0.7295 | |
|
| 0.1425 | 9.67 | 12000 | 1.7137 | 0.3824 | 0.5013 | 0.4339 | 0.7276 | |
|
|
|
|
|
### 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} |
|
} |
|
|
|
``` |
|
|
