TensorFlow Malicious Requests Detection Demo

This project demonstrates how to apply TensorFlow and neural networks for detecting malicious requests and visualize results on TensorBoard

Prerequisites

• Python3
• TensorFlow
• git

Clone Repository

Clone the repository and switch to the target directory

git clone "https://github.com/Dorokhov/tensorflow-keras-malicious-requests-demo"
cd <path_to_source_code_directory>

Demo

Step 1

Train Your Model

python train.py

Step 2

Visualize Training Results on TensorBoard

Run the following code from command line:

python tensorboard-visualisation.py
tensorboard --logdir="logs"

Copy tensorboard URL and open it in your famous browser:

Open "Projector" tab

On the left "Data" panel change "Color by" selector to "Class" value:

Ensure "Label Mode" is enabled ("A" icon is checked):

Now, you should be able to see visualisation of log request classification:

Step 3

Evaluate Custom Requests

Evaluate regular request:

python eval.py '{"username"="vova" "password"="0"}'

the system displays accuracy:

Evaluate SQL Injection request:

python eval.py '{"username"=";drop database" "password"="0"}'

the system displays accuracy: