The Deep Learning Revolution in Facial Recognition for Secure Login Systems

Introduction

In today’s online world, having strong security measures is more important than ever. Traditional ways of logging in, like using a username and password, are not always secure enough. This has led to more cases of hacking and identity theft. As a result, there’s a big push to find better ways to check who’s trying to access an account or system. Facial recognition for secure login system is one of the new methods. It’s becoming a popular choice because it adds an extra layer of security. Instead of just typing in a password, the system can recognize your face to make sure it’s really you. This technology is starting to be used in many different places to keep things safer.

In this blog, we’re going to talk about how deep learning, a type of artificial intelligence, is being used to make online security much better. We’re focusing on how combining two advanced technologies, called Multi-task Cascaded Convolutional Networks (MTCNN) and FaceNet, can create a really strong two-step way to check who’s trying to log in or access a system.

Learning Objectives

• Explore a complete project flow encompassing user registration, login, image capture, and verification for a holistic application perspective.
• Implement pretrained models to calculate cosine similarity, pivotal in verifying user images for authentication, ensuring accuracy and reliability.
• Integrate MongoDB or similar databases for efficient storage and retrieval of user data and image embeddings, ensuring streamlined access and management.
• Develop an authentication process using image-based verification, demonstrating user identity validation through captured images for secure system access.

This article was published as a part of the Data Science Blogathon.

Identifying the Problem

Problems like stolen passwords, people’s identities being used without their permission, and hackers getting into places they shouldn’t be are all reasons why we need a new kind of security system. This new system should not only be super secure but also dependable.

The Quest for Enhanced Security

Yet, the need for a robust and reliable authentication method, especially in scenarios where security is paramount, continues to grow. Traditional systems often fall short, paving the way for a more sophisticated approach—one that doesn’t just ensure high security but also guarantees reliability, scalability, and user-friendliness.

Leveraging Deep Learning for Enhanced Authentication

We’re working on a new way to check who’s trying to log in to a system, which is really important for keeping things secure. We’re using a mix of two smart technologies. One is great at spotting faces (MTCNN), and the other is really good at telling faces apart (FaceNet). By putting these two together, we’re making a security system that knows you by your face, which is perfect for places that need to be very careful about who they let in.

System Architecture Overview

This architectural design showcases the flow of processes within the facial recognition-based login authentication system. It involves multiple components working together seamlessly to ensure secure user authentication.

Components

1. User Interface: Represents the interface for user interaction, allowing users to input credentials and undergo facial recognition.
2. Authentication Module: This module handles the initial username-password authentication step.
3. MTCNN (Multi-task Cascaded Convolutional Neural Network): Responsible for precise face detection in images or video frames.
4. FaceNet: Manages the extraction of facial features and embeddings.
5. Facial Feature Database(MongoDB): Stores and manages the extracted facial embeddings securely.
6. Matching Module: Compares the facial embeddings during the authentication process.
7. Decision Module: Determines the authenticity of the user based on the matching results.
8. Access Control: Grants access to authorized users upon successful authentication.

Project Structure

We’re building a well-organized system that uses facial recognition to make sure the right people are getting access to secure areas. The system is set up in a way that each part has its own space, like folders for different tasks. This makes it easier to work on, improve, and make bigger over time.

controller/
│
├── app_controller/
├── auth_controller/
└── __init__.py

docs/
│
└── setup.md

venv/

face_auth/
│
├── src/
│   ├── business_val/
│   │   ├── user_embedding_val.py
│   │   └── user_val.py
│   │
│   ├── config/
│   │   └── database/
│   │       └── (config files)
│   │
│   ├── constant/
│   │   ├── auth_constant.py
│   │   ├── database_constants.py
│   │   └── embedding_constants.py
│   │
│   ├── data_access/
│   │   ├── user_data.py
│   │   └── user_embedding_data.py
│   │
│   ├── entity/
│   │   ├── user.py
│   │   └── user_embedding.py
│   │
│   ├── exception/
│   │   └── (exception handling)
│   │
│   ├── logger/
│   │   └── (logging configuration)
│   │
│   ├── utils/
│   │   └── util.py
│   │
│   └── __init__.py
│
└── logs/
    └── (log files)

static/
│
└── application/
    ├── assets/
    ├── css/
    ├── images/
    ├── js/
    └── vendor/

templates/
│
├── addd-app.html
├── error.html
├── index.html
├── login_embedding.html
├── login.html
├── register_embedding.html
├── register.html
├── unauthorized.html
└── view-app.html

.dockerignore
.gitignore
app.py
requirments.txt
setup.py

Here’s how we’ve set things up:

• controller: This is like the brain of the operation, where all the main instructions are.
• docs: Here we keep all the documents that explain how everything works.
• face_auth: This folder is dedicated to the actual facial recognition part.
• logs: Any records of what the system does go here, so we can look back if we need to.
• static: This is for files that don’t change often, like images or styles for the website.
• templates: These are the blueprints for how web pages will look.

We also have parts like:

• business_val:This is where we make sure that what we’re building is valuable for the business.
• data_access: This is all about how we get to and manage the data we need.
• entity: we define the types of data we’re dealing with.
• configuration: This is where we set up the settings for how the system should work.

Understand Core Fundamental Concepts in Deep Learning

Why Pretrained Models?

Transfer Learning:

• Explanation: Pretrained models, especially in the field of computer vision, have been trained on large datasets to solve complex tasks like image classification.
• Benefits: Utilizing these pretrained models allows us to leverage the knowledge gained during their training. This is known as transfer learning, where a model trained on one task is adapted for a new but related task.

+---------------------+
| Pretrained Model    |
| (e.g., FaceNet)     |
+---------------------+
         |
+---------------------+
|    Transfer Learning|
|     to New Task      |
+---------------------+

Resource Efficiency

• Explanation: Training a deep neural network from scratch requires substantial computational resources and time. Pretrained models save these resources.
• Benefits: By using pretrained models, we benefit from the feature extraction capabilities learned during their training, reducing the need for extensive training on new datasets.

Understanding Embeddings

Concept

• Explanation: Embeddings are dense vector representations of data in a lower-dimensional space. In the context of face recognition, facial embeddings are numerical vectors that capture essential features of a face.
• Benefits: Embeddings encode meaningful information about faces in a way that is conducive to similarity comparison.

Utilization in Face Recognition

• Explanation: In face recognition, embeddings are used to measure the similarity between different faces. A distance metric, often cosine similarity, is employed to quantify the likeness between face embeddings.
• Benefits: This allows for efficient comparison and matching of faces, forming the basis for authentication.

+---------------------+
|    Facial Image     |
|                     |
+---------------------+
           |
+---------------------+
|  Facial Embedding   |
|   (Dense Vector)    |
+---------------------+
           |
+---------------------+
|  Face Recognition   |
|  (Cosine Similarity)|
+---------------------+

Prerequisites for Deep Authenticator

1. Programming Proficiency 🖥️: A foundational understanding of Python programming is essential. Familiarity with concepts like functions, variables, and data structures is highly recommended.
2. Web Development Basics 🌐: A grasp of fundamental web development concepts is beneficial. This includes HTML, CSS, and basic JavaScript for understanding the frontend components.
3. API Knowledge 🔄: Familiarity with API concepts is necessary, as Deep Authenticator utilizes FastAPI to create APIs. Understanding HTTP methods (GET, POST) and how to handle requests and responses is crucial.
4. Database Understanding 🗃️: Basic knowledge of databases, particularly MongoDB, is recommended. Understanding concepts like collections, documents, and basic CRUD operations will be advantageous.
5. Deep Learning Basics 🤖: While not mandatory, a basic understanding of deep learning concepts will enhance comprehension, especially regarding face detection (MTCNN) and face embeddings (FaceNet).
6. Security Concepts 🔐: An awareness of security concepts, especially in the context of authentication, will be beneficial. This includes an understanding of token-based authentication and encryption.
7. Version Control 🔄: Proficiency with version control systems like Git is optional but recommended for collaborative development.

Step-by-Step Project Setup

Environment Preparation

• Python Installation: Ensure Python is installed on your machine. You can download and install Python from the official
• Virtual Environment Creation: Create a virtual environment using venv:

conda create -p ./venv python=3.8 -y

Virtual Environment Activation: Activate the virtual environment:

conda activate ./env

Dependency Installation

Install Required Packages: Use pip to install the necessary dependencies listed in the requirements.txt file:

pip install -r requirements.txt

edit .env file Configuration

SECRET_KEY=KlgH6AzYDeZeGwD288to79I3vTHT8wp7
ALGORITHM=HS256
DATABASE_NAME='UserDatabase'
USER_COLLECTION_NAME='User'
EMBEDDING_COLLECTION_NAME='Embedding'
MONGODB_URL_KEY= '[Put Your MongoDB url]'

Running the Application

• Use the provided app.py file to launch the application:

python app.py
or 
uvicorn app:app

Project Flow Chart

Project FlowChart

User and Embedding Classes Created and Saved

The Python classes for User and Embedding have been coded and stored in the designated ‘entity‘ folder, ready for use in the authentication system. These classes will facilitate the management of user information and their corresponding facial embeddings.

user.py

import uuid


class User:
    def __init__(
        self,
        Name: str,
        username: str,
        email_id: str,
        ph_no: str,
        password1: str,
        password2: str,
        uuid_: str = None,
    ):
        self.Name = Name
        self.username = username
        self.email_id = email_id
        self.ph_no = ph_no
        self.password1 = password1
        self.password2 = password2
        self.uuid_ = uuid_
        if not self.uuid_:
            self.uuid_ = str(uuid.uuid4()) + str(uuid.uuid4())[0:4]

    def to_dict(self) -> dict:
        return self.__dict__

    def __str__(self) -> str:
        return str(self.to_dict())

user_embedding.py

class Embedding:
    """Entity class for user embedding"""

    def __init__(self, UUID: str = None, user_embed=None) -> None:
        self.UUID = UUID
        self.user_embed = user_embed

    def to_dict(self) -> dict:
        return self.__dict__

    def __str__(self) -> str:
        return str(self.to_dict())

Authentication Validation Classes Explained

LoginValidation Class

• Objective: Validates user login details, including email and password.
• Working: Checks if email and password are provided, validates email format, and authenticates user login credentials against stored data in the database.

RegisterValidation Class

• Objective: Validates user registration details like name, email, password, and ensures their uniqueness.
• Working: Verifies that all required fields are provided, checks the format of the email, matches passwords, and confirms uniqueness of username and email before saving user data.

These classes are essential for ensuring the security and accuracy of user authentication and registration processes within the system.

user_val.py

import re
import sys
from typing import Optional

from passlib.context import CryptContext

from face_auth.data_access.user_data import UserData
from face_auth.entity.user import User
from face_auth.exception import AppException
from face_auth.logger import logging

bcrypt_context = CryptContext(schemes=["bcrypt"], deprecated="auto")


class LoginValidation:
    """_summary_
    """

    def __init__(self, email_id: str, password: str):
        """_summary_

        Args:
            email_id (str): _description_
            password (str): _description_
        """
        self.email_id = email_id
        self.password = password
        self.regex = re.compile(
            r"([A-Za-z0-9]+[.-_])*[A-Za-z0-9]+@[A-Za-z0-9-]+(\.[A-Z|a-z]{2,})+"
        )

    def validate(self) -> bool:
        """validate: This validates the user input

        Args:
            email_id (str): email_id of the user
            password (str): password of the user
        """
        try:
            msg = ""
            if not self.email_id:
                msg += "Email Id is required"
            if not self.password:
                msg += "Password is required"
            if not self.isEmailValid():
                msg += "Invalid Email Id"
            return msg
        except Exception as e:
            raise e

    def isEmailValid(self) -> bool:
        if re.fullmatch(self.regex, self.email_id):
            return True
        else:
            return False

    def verifyPassword(self, plain_password: str, hashed_password: str) -> bool:
        """_summary_

        Args:
            plain_password (str): _description_
            hashed_password (str): _description_

        Returns:
            bool: _description_
        """
        return bcrypt_context.verify(plain_password, hashed_password)

    def validateLogin(self) -> dict:

        """This checks all the validation conditions for the user registration
        """
        print(self.validate())
        if len(self.validate()) != 0:
            return {"status": False, "msg": self.validate()}
        return {"status": True}

    def authenticateUserLogin(self) -> Optional[str]:
        """_summary_: This authenticates the user and returns the token
        if the user is authenticated

        Args:
            email_id (str): _description_
            password (str): _description_
        """
        try:
            print(self.validateLogin())
            logging.info("Authenticating the user details.....")
            if self.validateLogin()["status"]:
                userdata = UserData()
                logging.info("Fetching the user details from the database.....")
                user_login_val = userdata.get_user({"email_id": self.email_id})
                if not user_login_val:
                    logging.info("User not found while Login")
                    return False
                if not self.verifyPassword(self.password, user_login_val["password"]):
                    logging.info("Password is incorrect")
                    return False
                logging.info("User authenticated successfully....")
                return user_login_val
            return False
        except Exception as e:
            raise AppException(e, sys) from e


class RegisterValidation:

    """_summary_: This authenticates the user and returns the status
    """

    def __init__(self, user: User) -> None:
        try:
            self.user = user
            self.regex = re.compile(
                r"([A-Za-z0-9]+[.-_])*[A-Za-z0-9]+@[A-Za-z0-9-]+(\.[A-Z|a-z]{2,})+"
            )
            self.uuid = self.user.uuid_
            self.userdata = UserData()
            self.bcrypt_context = CryptContext(schemes=["bcrypt"], deprecated="auto")
        except Exception as e:
            raise e

    def validate(self) -> bool:

        """This checks all the validation conditions for the user registration

        Returns:
            _type_: string
        """
        try:
            msg = ""
            if self.user.Name == None:
                msg += "Name is required"

            if self.user.username == None:
                msg += "Username is required"

            if self.user.email_id == None:
                msg += "Email is required"

            if self.user.ph_no == None:
                msg += "Phone Number is required"

            if self.user.password1 == None:
                msg += "Password is required"

            if self.user.password2 == None:
                msg += "Confirm Password is required"

            if not self.isEmailValid():
                msg += "Email is not valid"

            if not self.isPasswordValid():
                msg += "Length of the pass`word should be between 8 and 16"

            if not self.isPasswordMatch():
                msg += "Password does not match"

            if not self.isDetailsExists():
                msg += "User already exists"

            return msg
        except Exception as e:
            raise e

    def isEmailValid(self) -> bool:
        """_summary_: This validates the email id

        Returns:
            bool: True if the email id is valid else False
        """
        if re.fullmatch(self.regex, self.user.email_id):
            return True
        else:
            return False

    def isPasswordValid(self) -> bool:
        if len(self.user.password1) >= 8 and len(self.user.password2) <= 16:
            return True
        else:
            return False

    def isPasswordMatch(self) -> bool:
        if self.user.password1 == self.user.password2:
            return True
        else:
            return False

    def isDetailsExists(self) -> bool:
        username_val = self.userdata.get_user({"username": self.user.username})
        emailid_val = self.userdata.get_user({"email_id": self.user.email_id})
        uuid_val = self.userdata.get_user({"UUID": self.uuid})
        if username_val == None and emailid_val == None and uuid_val == None:
            return True
        return False

    @staticmethod
    def getPasswordHash(password: str) -> str:
        return bcrypt_context.hash(password)

    def validateRegistration(self) -> bool:

        """This checks all the validation conditions for the user registration
        """
        if len(self.validate()) != 0:
            return {"status": False, "msg": self.validate()}
        return {"status": True}

    def saveUser(self) -> bool:
        """_summary_: This saves the user details in the database
        only after validating the user details

        Returns:
            bool: _description_
        """
        try:
            logging.info("Validating the user details while Registration.....")
            if self.validateRegistration()["status"]:
                logging.info("Generating the password hash.....")
                hashed_password: str = self.getPasswordHash(self.user.password1)
                user_data_dict: dict = {
                    "Name": self.user.Name,
                    "username": self.user.username,
                    "password": hashed_password,
                    "email_id": self.user.email_id,
                    "ph_no": self.user.ph_no,
                    "UUID": self.uuid,
                }
                logging.info("Saving the user details in the database.....")
                self.userdata.save_user(user_data_dict)
                logging.info("Saving the user details in the database completed.....")
                return {"status": True, "msg": "User registered successfully"}
            logging.info("Validation failed while Registration.....")
            return {"status": False, "msg": self.validate()}
        except Exception as e:
            raise e

Application Overview – Web-App Initialization

User Login and Registration Embedding Validation

UserLoginEmbeddingValidation Class

• Objective: Validates the embedding of the user during login.
• Working: Detects the face from the provided image, generates embeddings, compares them with the stored embeddings in the database, and validates the user’s identity based on cosine similarity.

UserRegisterEmbeddingValidation Class

• Objective: Saves the embedding of the user during registration.
• Working: Generates embeddings from the provided images during registration, calculates the average embedding, and saves this data in the database for subsequent authentication.

These classes are pivotal for ensuring the accuracy of user authentication by comparing embeddings and facilitating secure registration by storing image embeddings for future verification.

import io
import sys
from ast import Bytes
from typing import List

import numpy as np
from deepface import DeepFace
from deepface.commons.functions import detect_face
from PIL import Image

from face_auth.constant.embedding_constants import (
    DETECTOR_BACKEND,
    EMBEDDING_MODEL_NAME,
    ENFORCE_DETECTION,
    SIMILARITY_THRESHOLD,
)
from face_auth.data_access.user_embedding_data import UserEmbeddingData
from face_auth.exception import AppException
from face_auth.logger import logging


class UserLoginEmbeddingValidation:
    def __init__(self, uuid_: str) -> None:
        self.uuid_ = uuid_
        self.user_embedding_data = UserEmbeddingData()
        self.user = self.user_embedding_data.get_user_embedding(uuid_)

    def validate(self) -> bool:
        try:
            if self.user["UUID"] == None:
                return False
            if self.user["user_embed"] == None:
                return False
            return True
        except Exception as e:
            raise e

    @staticmethod
    def generateEmbedding(img_array: np.ndarray) -> np.ndarray:
        """
        Generate embedding from image array"""
        try:
            faces = detect_face(
                img_array,
                detector_backend=DETECTOR_BACKEND,
                enforce_detection=ENFORCE_DETECTION,
            )
            # Generate embedding from face
            embed = DeepFace.represent(
                img_path=faces[0],
                model_name=EMBEDDING_MODEL_NAME,
                enforce_detection=False,
            )
            return embed
        except Exception as e:
            raise AppException(e, sys) from e

    @staticmethod
    def generateEmbeddingList(files: List[Bytes]) -> List[np.ndarray]:
        """
        Generate embedding list from image array
        """
        embedding_list = []
        for contents in files:
            img = Image.open(io.BytesIO(contents))
            # read image array
            img_array = np.array(img)
            # Detect faces
            embed = UserLoginEmbeddingValidation.generateEmbedding(img_array)
            embedding_list.append(embed)
        return embedding_list

    @staticmethod
    def averageEmbedding(embedding_list: List[np.ndarray]) -> List:
        """Function to calculate the average embedding of the list of embeddings

        Args:
            embedding_list (List[np.ndarray]): _description_

        Returns:
            List: _description_
        
        """
        avg_embed = np.mean(embedding_list, axis=0)
        return avg_embed.tolist()

    @staticmethod
    def cosine_simmilarity(db_embedding, current_embedding) -> bool:
        """Function to calculate cosine similarity between two embeddings

        Args:
            db_embedding (list): This embedding is extracted from the database
            current_embedding (list): This embedding is extracted from the current images

        Returns:
            int: simmilarity value
        """
        try:
            return np.dot(db_embedding, current_embedding) / (
                np.linalg.norm(db_embedding) * np.linalg.norm(current_embedding)
            )
        except Exception as e:
            raise AppException(e, sys) from e

    def compareEmbedding(self, files: bytes) -> bool:
        """Function to compare the embedding of the current image with the embedding of the database

        Args:
            files (list): Bytes of images

        Returns:
            bool: Returns True if the similarity is greater than the threshold
        """
        try:

            if self.user:
                logging.info("Validating User Embedding ......")
                # Validate user embedding
                if self.validate() == False:
                    return False

                logging.info("Embedding Validation Successfull.......")
                # Generate embedding list

                logging.info("Generating Embedding List .......")
                embedding_list = UserLoginEmbeddingValidation.generateEmbeddingList(
                    files
                )

                logging.info("Embedding List generated.......")
                # Calculate average embedding

                logging.info("Calculating Average Embedding .......")
                avg_embedding_list = UserLoginEmbeddingValidation.averageEmbedding(
                    embedding_list
                )
                logging.info("Average Embedding calculated.......")

                # Get embedding from database
                db_embedding = self.user["user_embed"]

                logging.info("Calculating Cosine Similarity .......")
                # Calculate cosine similarity
                simmilarity = UserLoginEmbeddingValidation.cosine_simmilarity(
                    db_embedding, avg_embedding_list
                )
                logging.info("Cosine Similarity calculated.......")

                if simmilarity >= SIMILARITY_THRESHOLD:
                    logging.info("User Authenticated Successfully.......")
                    return True
                else:
                    logging.info("User Authentication Failed.......")
                    return False
            logging.info("User Authentication Failed.......")

            return False
        except Exception as e:
            raise AppException(e, sys) from e

    # def get_user_embeeding_object(self, uuid_:str) -> Embedding:
    #     """_summary_

    #     Args:
    #         user_embedding (dict): _description_

    #     Returns:
    #         Embedding: _description_
    #     """
    #     try:

    #         user_embedding = self.user_embedding_data.get_user_embedding(uuid_)
    #         return user_embedding
    #     except Exception as e:
    #         raise AppException(e, sys) from e


class UserRegisterEmbeddingValidation:
    def __init__(self, uuid_: str) -> None:
        self.uuid_ = uuid_
        self.user_embedding_data = UserEmbeddingData()

    def saveEmbedding(self, files: bytes):
        """This function will generate embedding list and save it to database
        Args:
            files (dict): Bytes of images

        Returns:
            Embedding: saves the image to database
        """
        try:
            embedding_list = UserLoginEmbeddingValidation.generateEmbeddingList(files)
            avg_embedding_list = UserLoginEmbeddingValidation.averageEmbedding(
                embedding_list
            )
            self.user_embedding_data.save_user_embedding(self.uuid_, avg_embedding_list)
        except Exception as e:
            raise AppException(e, sys) from e

Camera Activation for Registration Embedding

Generating Embedding in the Backend

MongoDB Client Initialization for User Data

• Objective: Handles the MongoDB connection and database access for user-related data.
• Working: Initializes the MongoDB client using the specified URL. It establishes a connection to the MongoDB database, ensuring a singleton pattern to maintain a single client instance throughout the application. This class serves as an interface for accessing and interacting with the MongoDB database storing user information.

import pymongo

from face_auth.constant.database_constants import (
    DATABASE_NAME,
    MONGODB_URL_KEY
)


class MongodbClient:
    client = None

    def __init__(self, database_name=DATABASE_NAME) -> None:
        if MongodbClient.client is None:
            mongo_db_url = MONGODB_URL_KEY
            if "localhost" in mongo_db_url:
                MongodbClient.client = pymongo.MongoClient(mongo_db_url) 
            else:
                MongodbClient.client=pymongo.MongoClient(mongo_db_url)
        self.client = MongodbClient.client
        self.database = self.client[database_name]
        self.database_name = database_name

Application Controller for User Authentication Through Facial Recognition

This FastAPI controller module facilitates user authentication via facial recognition. It consists of multiple endpoints:

1. GET /application/: Renders the login page with embedded facial recognition.
2. POST /application/: Validates the user’s facial embeddings upon login, authenticating them if successful.
3. GET /application/register_embedding: Renders the registration page with facial embedding functionality.
4. POST /application/register_embedding: Saves the facial embeddings during user registration.

Working Overview

• Login Flow: The loginEmbedding function compares the facial embeddings of uploaded images with those stored in the database. If authenticated, it renders the login page.
• Registration Flow: The registerEmbedding function saves the facial embeddings into the database during user registration. Upon successful embedding storage, it redirects to the login page.

# [Imports]

# Router initialization
router = APIRouter(
    prefix="/application",
    tags=["application"],
    responses={"401": {"description": "Not Authorized!!!"}},
)

# Templates initialization
templates = Jinja2Templates(directory=os.path.join(os.getcwd(), "templates"))

# Environment variable setup
os.environ["CUDA_VISIBLE_DEVICES"] = "-1"

# ImageForm class for handling image data
class ImageForm:
    # Constructor to initialize image attributes
    def __init__(self, request: Request):
        # Initialize image attributes
        self.request: Request = request
        self.image_list = [f"image{i}" for i in range(1, 9)]
        self.images = [None] * 8

    # Method to handle form data and decode base64 images
    async def create_oauth_form(self):
        form = await self.request.form()
        for idx, image_name in enumerate(self.image_list):
            image_data = form.get(image_name)
            if image_data:
                self.images[idx] = io.BytesIO(base64.b64decode(image_data.split(",")[1])).getvalue()

# Route for the main application page
@router.get("/", response_class=HTMLResponse)
async def application(request: Request):
    try:
        user = await get_current_user(request)
        if user is None:
            return RedirectResponse(url="/auth", status_code=status.HTTP_302_FOUND)
        
        return templates.TemplateResponse("login_embedding.html", context={
            "request": request,
            "status_code": status.HTTP_200_OK,
            "msg": "Logged in Successfully",
            "user": user['username']
        })

    except Exception as e:
        # Error handling for exceptions
        msg = "Error in Login Embedding in Database"
        return templates.TemplateResponse("error.html",
            status_code=status.HTTP_404_NOT_FOUND,
            context={"request": request, "status": False, "msg": msg},
        )

# [Other Routes - loginEmbedding, registerEmbedding, etc.]

This controller handles user authentication and embedding storage through FastAPI endpoints, incorporating facial recognition functionality for login and registration.

Authentication Module for User Management

This FastAPI module manages user authentication and session handling. It comprises endpoints for user login, registration, token creation, and logout.

Features

1. Token Generation: Creates access tokens upon successful login for user authentication and session management.
2. Login and Registration: Provides endpoints for user login and registration, validating user credentials and handling user sessions.
3. Session Management: Manages user sessions and cookies to maintain user authentication throughout the application.
4. Error Handling: Handles exceptions and errors during authentication processes, rendering appropriate error responses.

import os
import io
import base64
from typing import Optional
from fastapi import APIRouter, File, Request, status
from starlette.responses import RedirectResponse, HTMLResponse
from fastapi.templating import Jinja2Templates
from controller.auth_controller.authentication import get_current_user
from face_auth.business_val.user_embedding_val 
import UserLoginEmbeddingValidation, UserRegisterEmbeddingValidation

router = APIRouter(prefix="/application", tags=["application"], 
responses={"401": {"description": "Not Authorized!!!"}})
templates = Jinja2Templates(directory=os.path.join(os.getcwd(), "templates"))
os.environ["CUDA_VISIBLE_DEVICES"] = "-1"

class ImageForm:
    def __init__(self, request: Request):
        self.request: Request = request
        self.images: List[Optional[str]] = [None] * 8

    async def create_oauth_form(self):
        form = await self.request.form()
        for i in range(8):
            self.images[i] = form.get(f"image{i + 1}")

@router.get("/", response_class=HTMLResponse)
async def application(request: Request):
    try:
        user = await get_current_user(request)
        if user is None:
            return RedirectResponse(url="/auth", 
            status_code=status.HTTP_302_FOUND)
        return templates.TemplateResponse("login_embedding.html", 
        context={"request": request, "status_code": status.HTTP_200_OK, 
        "msg": "Logged in Successfully", "user": user['username']})
    except Exception as e:
        msg = "Error in Login Embedding in Database"
        response = templates.TemplateResponse("error.html",
         status_code=status.HTTP_404_NOT_FOUND, context={"request": request, "
         status": False, "msg": msg})
        return response 

# Define other routes (loginEmbedding, register_embedding, etc.) similarly...

This module encapsulates user authentication functionalities, enabling login, registration, and session handling within the FastAPI framework.

For Exception and Logger scripting files follow on GitHub

Run the Project

Initial Access

• Access the application via the browser (localhost).
• The user initiates the registration process.

Application Overview – Web-App Initialization

Registration

• User proceeds with the registration, providing necessary details.
• Camera access is granted for capturing 8 images.
• The captured images are stored as part of the user’s registration in MongoDB.

Camera Activation – Registration Embedding

Collecting Images after Camera Activation

View MongoDB Embedding Information

Generating Embedding in the Backend

Login

• Once registered, the user logs into the system using their username and password.

Application Overview – Web-App Initialization

Setting Up the Login Camera

User Authentication Successful

Authentication

• To authenticate, the system accesses the camera for capturing a new image.
• The system computes the similarity between the captured image and the stored images in the database (using cosine similarity or another metric).
• Based on the similarity score, access is either granted or denied to the user.

Figure 14: Formula of Cosine Similarity

This flow illustrates the sequential steps a user takes within the application, from registration to authentication, leveraging the camera to verify identity through image comparison.

Limitation

• Limited Dataset Impact: Performance relies heavily on the diversity of the dataset.
• Resource Demand: High computational requirements may affect system performance.
• Environmental Influence: Varying conditions like lighting and angles can impact accuracy.
• Security and Privacy: Potential concerns exist regarding data privacy and security.
• Bias and Ethics: Algorithms might display biases or ethical concerns.
• Regulatory Compliance: Adhering to regulations and ethical standards is challenging.

Conclusion

In facial recognition, two primary steps involve capturing images and comparing similarities. Initially, the system collects images either during registration or login through camera activation. The backend processes these images to generate unique embeddings stored in a database, such as MongoDB. During login, the camera captures a new image for comparison, and the backend computes the cosine similarity against stored embeddings, authenticating the user if a match is found.” This two-step process ensures secure access using facial recognition technology.

Key Takeaways

• Learn the basics of facial recognition and its real-world applications in security, access control, and marketing.
• Be mindful of the ethical considerations related to facial recognition, including privacy concerns, potential biases, and the responsible use of personal data.
• Grasp the functioning of facial recognition systems, from capturing facial images to analyzing unique features and matching them against stored templates.
• Recognize the limitations and challenges of facial recognition, including accuracy issues and vulnerability to hacking. Consider the future implications and the importance of ethical development and regulation in this technology.

Frequently Asked Questions

The media shown in this article is not owned by Analytics Vidhya and is used at the Author’s discretion.

suyodhanj6

As a Data Scientist, I leverage my expertise in statistical analysis, machine learning, and data visualization to derive insights and make informed decisions. I have experience working with various programming languages, databases, and machine learning frameworks, enabling me to tackle complex data problems and deliver actionable results. I am a collaborative problem-solver who can work with stakeholders to deliver scalable and secure data solutions.