Cyber-ThreaD/SecureBERT-AttackER

This model is a fine-tuned version of ehsanaghaei/SecureBERT on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.4668
• Precision: 0.4762
• Recall: 0.5291
• F1: 0.5013
• Accuracy: 0.7376

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
1.9177	0.4	500	1.6839	0.06	0.0278	0.0380	0.6004
1.4976	0.81	1000	1.4936	0.2281	0.2659	0.2456	0.6313
1.2309	1.21	1500	1.2915	0.2650	0.3148	0.2878	0.6657
1.0546	1.61	2000	1.2454	0.2950	0.3796	0.3320	0.6804
0.9405	2.01	2500	1.2377	0.3613	0.3532	0.3572	0.6916
0.7501	2.42	3000	1.1723	0.3607	0.4180	0.3873	0.7171
0.7133	2.82	3500	1.1584	0.3632	0.4444	0.3998	0.7160
0.5896	3.22	4000	1.2288	0.4103	0.4444	0.4267	0.7306
0.5353	3.63	4500	1.2319	0.3978	0.4815	0.4357	0.7254
0.5432	4.03	5000	1.2173	0.4269	0.4868	0.4549	0.7306
0.4062	4.43	5500	1.2832	0.4398	0.5026	0.4691	0.7272
0.4485	4.83	6000	1.2196	0.4212	0.5093	0.4611	0.7412
0.3614	5.24	6500	1.3155	0.4325	0.4960	0.4621	0.7325
0.3308	5.64	7000	1.3501	0.4184	0.5119	0.4604	0.7354
0.3645	6.04	7500	1.3391	0.4359	0.5172	0.4731	0.7366
0.2982	6.45	8000	1.3889	0.4093	0.5225	0.4590	0.7315
0.2845	6.85	8500	1.4109	0.4452	0.5159	0.4779	0.7377
0.2482	7.25	9000	1.4668	0.4762	0.5291	0.5013	0.7376
0.2636	7.66	9500	1.4925	0.4540	0.5357	0.4915	0.7341
0.2605	8.06	10000	1.4916	0.4586	0.5344	0.4936	0.7405
0.1989	8.46	10500	1.5096	0.4661	0.5370	0.4991	0.7387
0.2415	8.86	11000	1.4698	0.4603	0.5450	0.4991	0.7443
0.2488	9.27	11500	1.4736	0.4578	0.5304	0.4914	0.7455
0.2129	9.67	12000	1.5067	0.4640	0.5450	0.5012	0.7439

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}
}