BERT model for SQL Injection

Doing a simple api with flask and BERT model for the SQL Injection Dataset https://www.kaggle.com/datasets/sajid576/sql-injection-dataset.

Model code is based on https://www.kaggle.com/code/bobaaayoung/bert-word2vec-lstm-cnn-text-classification-ipynb.

This model is part of a kong api gateway plugin solution for detecting offensive requests: https://github.com/mateuslima90/kong-plugin-ai-attack-detection

How to use

Download the model weights and save it in bert-sql-api/artifact/. The steps are

# clone this repo
cd bert-sql-api
mkdir artifact
wget -P artifact 'https://drive.google.com/file/d/1m8HPQLToDfvE9tIE_cisuG8qp9b-lntf'

Trained model weights available here..

Local

Install all the requirements from requirements.txt in local machine

pip install --no-cache-dir -r requirements.txt

and run the application python app.py and the grpc service python predict_server.py

Docker

Alternatively run the dockerfile

docker run -p 5000:5000 -p 50051:50051 gkdogifjhif/bert-sql-api

Call Example

An example of using is

curl -X POST http://localhost:5000/predict \
-H "Content-Type: application/json" \
-d '{"input1": "\' DELETE 1=1"}'

Or passing more than one key

curl -X POST http://localhost:5000/predict \
-H "Content-Type: application/json" \
-d '{"campo1": "DELETE 1=1","campo2": "SELECT * FROM users WHERE id = 1","input": "1 UNION SELECT 1, version() limit 1,1"}'