Signature-Matching-With-Tensorflow

Signature verification through handwriting analysis is one of the most common tasks in forensic document analysis. It is very important to compare questioned writing/signatures with the genuine one, when it comes to data protection and identity management. We are utilizing machine learning techniques to perform handwriting analysis by matching images of a questioned handwriting with the genuine ones to authenticate writing/signatures. Machine Learning algorithms will adapt and improve their performance as the number of handwriting samples increase for learning and do more accurate analysis.

Tech Stack:

• Python
• Flask
• Tensorflow
• MongoDB
• NodeJs (for testing)

To run the pipeline do the following:

docker-compose up

This will put the backend up. To use the front end just open the file

SigAuthUI/src/main/templates/signaturePage.html

in your browser!.

Installing Project on EC2

# First we need to chmod 400 out public key
$path = ".\MLSigBackend.pem"
# Reset to remove explict permissions
icacls.exe $path /reset
# Give current user explicit read-permission
icacls.exe $path /GRANT:R "$($env:USERNAME):(R)"
# Disable inheritance and remove inherited permissions
icacls.exe $path /inheritance:r

# Now we can ssh into the EC2
ssh -i "MLSigBackend.pem" ubuntu@ec2-35-173-252-77.compute-1.amazonaws.com

#Grant super user
sudo su

#Update apt-get
apt-get update

#install git
apt-get install git-core

##########Docker Installation##########

sudo apt-get update

sudo apt-get install \
    apt-transport-https \#!/Login/post_login
    ca-certificates \
    curl \
    gnupg-agent \
    software-properties-common

curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -

sudo add-apt-repository \
   "deb [arch=amd64] https://download.docker.com/linux/ubuntu \
   $(lsb_release -cs) \
   stable"

sudo apt-get update

sudo apt-get install docker-ce docker-ce-cli containerd.io

#######################################

#For running the docker container persistantly
docker run -d --restart=always -p 80:80 -t my_flask_app

########Messing around with API ports###########

docker build -t api .
docker run --rm -it -p 80:5000 api

##Closing containers
docker-compose down
--volume (Full teardown)

Testing the Backend Endpoints

To test the backend endpoints simply go to localhost

Here you will see 4 different endpoints to test

• /Login/create_tenant
• /Login/login_tenant
• /Signature/check_status
• /Signature/verify_signature

You can click on them to expand and then there will be fields to enter input thru. It will make calls to the API and return results.

End to End and Performance Testing

Navigate to the testing folder. cd testing

To install the needed dependencies you can run

npm i

In th there are folders with images for testing and a script to act as a client that sends all the pictures in a folder with 1 image per request asynchronously.

Each directory has a different number of imgs (125, 250, 500, 750, 1000). You pass the name of the directory as an argument like so:

node upload.js img2

The results will print the body returned fromeach requests as well as the number of total requests sent and any failed deliveries.

If you are interested in timing, I used the simple time utility from my Unix shell. So my commands looked like this:

time node upload.js img2

Front-End Tests

We have unit tests for the logic of the front end js files. These can be founf in SigAuthUI/src/test