cyber-threat-detector-v1

Overview

This model is specialized in detecting malicious intent within command-line strings and network log entries. It helps security analysts flag potentially harmful activity in real-time.

Model Architecture

• Base Model: RoBERTa-base
• Fine-tuning: Trained on custom cybersecurity datasets including obfuscated scripts and SQL injection attempts.

Intended Use

• Security Information and Event Management (SIEM) integration.
• Automated log auditing.
• DevSecOps pipeline monitoring.

Limitations

• May produce false positives on complex, valid administrative scripts.
• Only supports text-based logs; does not analyze binary packets.

Example Code

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

tokenizer = AutoTokenizer.from_pretrained("cyber-threat-detector-v1")
model = AutoModelForSequenceClassification.from_pretrained("cyber-threat-detector-v1")

log_line = "SELECT * FROM users WHERE id = '1' OR '1'='1';"
inputs = tokenizer(log_line, return_tensors="pt")
outputs = model(**inputs)
prediction = torch.argmax(outputs.logits)
print("Threat Detected" if prediction == 1 else "Safe")