WraPbiT/secbert-casie-extraction

SecBERT for Cyber Threat Intelligence Extraction (CASIE)

This model is a fine-tuned checkpoint of SecBERT (jackaduma/SecBERT) trained on the CASIE dataset (AAAI 2020) for Cybersecurity Event and Argument Extraction (EAE).

Model Details

• Base Architecture: SecBERT (Domain-adapted BERT for Cybersecurity)
• Task: Fine-grained Token Classification (Event Triggers & Argument Roles)
• Dataset: CASIE (Cybersecurity Event Detection and Argument Extraction)
• Official Project Repository: GitHub: cyber-event-argument-extraction

Quick Usage

import torch
from transformers import AutoTokenizer, AutoModelForTokenClassification

model_id = "WraPbiT/secbert-casie-extraction"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForTokenClassification.from_pretrained(model_id)

text = "LockBit ransomware group encrypted internal databases at the hospital demanding $2 million."
inputs = tokenizer(text, return_tensors="pt")
with torch.no_grad():
    outputs = model(**inputs)