Building “Passwle” System Based on Siamese Neural Network

Abstract

Security is one of the important aspect in our life. Almost every gadgets and websites in this world use some security measure like password and passcode. with every security measure there are some security fault and many hacking method to attack said security measure. To answer those problem “Passwle” system which is a Graphical Password with recall-based technique is created. “Passwle” system created by using Python version 3.7.10 as the programming language with Siamese Neural Network for training and prediction process and using Omniglot Dataset as the dataset. For the frontend, Node.js is used to deliver the User Interface and communicate with the backend, backend side use Flask Framework to provide the API and predicting the image that used as password in Google Colaboratory and Google Drive is used for the database. After some testing, “Passwle” system currently can not be used as an alternative to the current security measure, because the process time for sign in and sign up is significantly longer than conventional security measure which take average 4.35 seconds for sign up and 18.88 seconds for sign in process and there are some major flaw in the user-side security which have 92% success rate for Social engineering attack and system-side security flaw.