---
license: mit
tags:
- SQL Injection
widget:
- text: 1 UNION SELECT username, password FROM users --
  example_title: SQL Injection Detection
metrics:
- f1
- accuracy
---


**A distilBERT based SQL Injection Detection Model**

**Model description**

This model, based on DistilBERT, is specifically tailored for the detection of SQL injection attacks. 
Through fine-tuning using the Hugging Face's Trainer API, the model has been trained to identify potentially malicious SQL queries with high accuracy.

- __Architecture:__ DistilBERT
- __Fine-tuning Method:__ Trainer API
- __Performance Metrics:__
  - __F1-score:__ 99.86%
  -  __Accuracy:__ 99.99%
- __Training Epochs:__ 6

**Dataset description**

The model was fine-tuned on the SQL Injectiom dataset, curated and made available by SAJID576 on Kaggle. 
This dataset comprises of 30,920 rows of SQL queries, including both benign and malicious examples, providing a comprehensive training corpus for robust model development.

- __Dataset Source:__ https://www.kaggle.com/datasets/sajid576/sql-injection-dataset/data
- __Size:__ 30,920 rows