Volume 8, Issue 7, July – 2023 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Mi Amor - A Privacy-Enhanced Environment for

Online Matchmaking
1
Prasham Mehta ; 2Keval Shah 3
Rohit Raval ; 4Manan Shah
Information Technology Computer Science (A.I) Information Technology Computer Science (A.I)
Atharva College of Engineering NMIMS Atharva College of Engineering NMIMS

Abstract:- With the growing popularity of online dating I. INTRODUCTION

platforms, concerns regarding user privacy and data
security have become increasingly significant. In this Online dating has gained immense significance in
research paper, we propose a dating web application today's world, enabling people to connect with individuals
that prioritizes user privacy while offering secure data from diverse locations and engage in conversations based on
management. The application incorporates a unique face shared interests and compatibility. The rapid growth of
recognition system, horoscope-based matching, online dating can be attributed to various factors, including
compatibility percentage, and location-based filtering to career-related motivations, where individuals seek like-
help users find potential partners with ease. By minded people or explore potential partnerships, ranging
employing face verification at regular intervals, the from casual to serious relationships[4].
application ensures that users are personally engaged in
conversations, reducing the possibility of third-party However, it is important to acknowledge that online
involvement and increasing transparency[2]. dating apps face acceptance challenges in certain regions
Furthermore, the application employs a comprehensive due to concerns surrounding privacy and sharing constraints.
registration process, including face registration, to The potential misuse of someone's identity or unauthorized
minimize fake accounts and enhance user authenticity. sharing of images raises legitimate concerns. Therefore, our
Users have the flexibility to customize their profiles by dating application places a strong emphasis on privacy and
appending horoscopes, editing bios, and adding security, addressing these issues[5].
images[1]. The application streamlines the matching
process, allowing users to double-tap to express interest Online dating applications involve significant sharing
and swipe left or right to view the next profile. A of personal information with strangers, creating a delicate
bookmarking feature is also provided to facilitate future balance between users' expectations of privacy and the
interactions or changes in user actions. Notably, the application's sharing configurations[2]. Resolving this
application eliminates the common practice of charging conflict is crucial for the successful adoption of any online
users to identify who has liked their profiles, providing dating application. Multiple approaches exist for designing
instant access to interested individuals and fostering an online dating application, as each company prioritizes its
prompt communication. To enhance user experience, the own sharing objectives, users' perceived preferences,
application employs scrolling functionality for profile privacy and security considerations, technological
browsing and empowers users with the ability to capabilities, and monetization goals.
personalize the application's themes to suit their
preferences[1]. Once mutual interest is established, a This article proposes the integration of robust security
real-time chat messaging feature is activated, enabling and privacy features, which have always been critical
users to engage in meaningful conversations and foster concerns in dating applications. Our solution involves
connections. The backend infrastructure leverages Face regular user verification every 60 seconds, without intrusive
Net and other machine learning models to implement the pop-ups or notifications.
proposed functionalities effectively. The process involves
registering the user's face during initial setup, followed To address the issue of profile misuse and fake
by regular face verification at 60-second intervals. To accounts, we utilize two machine learning models, Face Net
optimize storage and processing, a machine learning and Mobile Net, which capture images from the device's
model is employed to extract and store only the essential camera. Facial detection is performed by outlining a border
features from the images, resulting in efficient data around the face. The standard protocol mandates capturing
management and improved processing speed[11]. and verifying the user's image at the initial stage, allowing
subsequent image uploads only from verified users.
Keywords:- Dating Web Application, Privacy-Preserving, Unverified photos are not uploaded, and all stored photos
Face Recognition, Compatibility Matching, user are encrypted to ensure privacy.
Authenticity, Machine Learning, Real-Time Chat
Messaging.

IJISRT23JUL1498 www.ijisrt.com 1950

Volume 8, Issue 7, July – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
In case of violations such as nudity, violence, or before pursuing a relationship can lead to less complications
explicit content, an automated system promptly removes and a greater chance of compatibility.
such photos and bans the associated accounts. Furthermore,
our application introduces a bookmark feature to address a  Career :-
common limitation observed in other dating platforms. Many individuals have a preference for a partner with
Users can mark specific profiles and revisit them at a later the same career. This stems from their passion for their own
time, providing added convenience and flexibility. profession, as they believe that having a partner from the
same field will result in enhanced productivity and the
Among popular dating applications like Tinder, Happn, ability to generate innovative ideas together. For instance, an
and Bumble, privacy and security concerns have led to engineer may specifically choose to marry another engineer
instances of fraud and fake profiles [3]. While these to foster mutual professional growth and shared expertise.
applications typically verify users only during initial setup,
our application offers on going verification every 60 III. TECHNOLOGIES AND PROGRAMMING
seconds, significantly enhancing security and minimizing LANGUAGES USED
the misuse of others' images and profiles.
 Dart –
II. FACTORS USED FOR FINDING Dart, a client-side programming language widely
A PARTNER utilized in Flutter, stands out for its simplicity, minimalistic
nature, and remarkable efficiency. With just a single line of
There are multiple factors involved in finding a code, Dart can achieve multiple functionalities. When
compatible partner, including preferences, horoscope, comparing Dart to Java, we observe that in Dart, a single
hobbies, interests, and career. Each of these factors plays a line of code is adequate for developing both frontend and
vital role in determining the compatibility of an individual. backend functionalities, whereas Java requires separate code
for each.
 Preferences:-
Many individuals have specific expectations and Google developed the Dart language in 2013, and since
desires when it comes to finding a matching partner, and its inception, it has significantly influenced numerous
preferences play a crucial role in such cases. For example, a application developers in various ways. Its key advantage
person may desire a partner who shares their interest in lies in its ease of learning and implementation, making it
activities like clubbing or volunteering at social events. accessible even for beginners to swiftly create native apps.
With a Java-like syntax, developers familiar with OOP
 Horoscope :- programming can seamlessly transition into coding Dart,
Many individuals believe that horoscope signs can allowing for a simplified learning experience while
enhance compatibility with their partners. Horoscopes are leveraging the framework's tools and terminologies. Dart's
rooted in astrology, but it is important to note that not emphasis on "High performance" plays a pivotal role,
everyone universally or scientifically believes in horoscope positioning it as a favoured language among application
compatibility. However, many individuals have a mind-set developers. Dart-powered applications demonstrate faster
that a person belonging to a specific zodiac sign possesses execution compared to other programming languages,
certain qualities that align well with their own, leading them thanks to features like JIT and AOT. JIT enables Hot Reload
to believe in the compatibility of specific zodiac sign functionality, enhancing development efficiency, while
pairings. AOT facilitates quick start-up and improved application
execution.
 Hobbies :-
Individuals seeking compatibility often look for When considering every aspect of app development,
partners who share their hobbies and passions, such as Dart excels in every parameter, solidifying its position as the
sports, fitness, cooking, and more. This alignment of top programming language for building native apps. The
interests enhances their connection and overall inclusion of Flutter is just the tip of the iceberg, as the
compatibility. For example, suppose an individual has a combination of Dart and Flutter provides developers with an
keen interest in sports or other activities and desires a optimal environment to efficiently create apps. With
partner who shares the same hobby or passion, be it sports, continuous improvements and expanding functionalities,
fitness, cooking, or any other. People often seek Dart and Flutter have become a boon in the world of
compatibility based on hobbies, believing it contributes to a application development.
stronger connection between them.
 Flutter –
 Interests :- In today's business landscape, having a mobile
When individuals seek a partner based on shared application is essential whether you run a start-up or an
interests, it encompasses various factors such as common established business. Not only does it contribute to business
passions, educational background, business pursuits, and growth, but it also aids in marketing efforts. Flutter, an
shared events. Interests can also encompass crucial aspects open-source and free software development kit (SDK), is a
from an individual's perspective, including preferences for prime example of a platform that enables the creation of
serious or casual dating. Understanding a person's interests

IJISRT23JUL1498 www.ijisrt.com 1951

Volume 8, Issue 7, July – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
native iOS and Android applications using a single The two main hyper parameters of Mobile Net are the
codebase. "width multiplier" and "resolution multiplier." The width
multiplier reduces the number of channels in each layer,
Flutter's comprehensive toolkit offers all the necessary effectively decreasing the model's size and computational
features for cross-platform application development. Flutter complexity. On the other hand, the resolution multiplier
has made a significant impact on the world of application scales down the input image resolution, further reducing the
development, providing a range of useful features and required computation. What makes these hyper parameters
functionalities while being easy to implement. One standout advantageous is their flexibility in adjusting the inference
feature is "Hot Reload," which eliminates the need for speed, model size, and accuracy based on the constraints of
developers to manually refresh their code implementation. the target devices. Mobile Net’s primary objective is to keep
With Hot Reload, changes are automatically applied, the model lightweight while capturing complex patterns[7].
allowing developers to see progress in real-time. Its architecture comprises depth-wise separable and point-
Additionally, Flutter's extensive widget library proves wise separable convolutional layers. The size of the
immensely valuable. By importing the required library into convolution layer is set to 1x1 to increase or decrease the
the pubspec.yaml file, developers can swiftly incorporate number of channels as needed.
desired functionalities into their apps within minutes.
 AWS(Amazon Web Services)-
In today's era, clients demand efficient and visually It provides users with a secure and optimized solution
appealing user interfaces (UI) for their applications, for storing their databases. From its inception, AWS has
regardless of the level of functionality. Flutter ensures the maintained a strong reputation owing to its extensive range
provision of clean UI right from the initial stages of app of services, encompassing compute, storage, database, and
development. With its extensive range of features and machine learning capabilities[13].
functionalities, Flutter emerges as the top choice for app
development, guaranteeing a seamless user experience and Now that we have an understanding of the services
polished UI design. offered by AWS, let's delve into a detailed overview of each
of them.
 Face Net –
The FaceNet model, developed by Google in 2017, is a  Compute Services:
deep learning model that utilizes a convolutional neural When it comes to running applications, the need for
network (CNN) architecture. Its purpose is to extract facial resizable servers is fulfilled by Amazon Elastic Compute
features and generate embedding’s by mapping facial Cloud (EC2). Additionally, AWS Lambda offers a distinct
characteristics into a high-dimensional space. Face Net's feature where code execution is possible without the hassle
performance has been remarkable, as it has achieved state- of managing servers.
of-the-art results in various benchmark face recognition
datasets, including Labeled Faces in the Wild (LFW) and the  Storage Services:
YouTube Face Database. To meet our storage needs, Amazon Simple Storage
Service (S3) offers a scalable object storage solution.
Face Net performs at par when it comes to Furthermore, AWS provides Amazon Elastic Block Store
performance as it generates a high quality face mapping. (EBS), a user-friendly and scalable block storage service
“ZF-Net” and “Inception” Network are the main machine specifically designed for Amazon Elastic Compute Cloud
learning architecture from which the images are used for (EC2).
mapping. Talking about the Loss function, Triplet loss is
used as a method to train the architecture and this loss  Database Services:
function more efficient than other loss functions. Face Net AWS is renowned for its exceptional storage and
has such a high accuracy and robustness such that every database offerings. Amazon Relational Database Service
other application is using this model for biometric (RDS) is designed for relational databases, while Amazon
authentication, social media applications like snapchat, Dynamo DB is tailored for No SQL databases. Additionally,
dating app also nowadays it is highly implemented in (HCI) for high-performance databases, Amazon Aurora provides
Human computer interaction[6]. an efficient and optimized solution. These databases not
only deliver superior performance but also prioritize
 Mobile Net – security, ensuring data remains well-protected.
Mobile Net is considered a backup model for Face Net
due to multiple factors that arise during the implementation IV. GRAPHICAL USER INTERFACE (GUI)
of Face Net. Face Net, being a heavy model, is not suitable
for small applications or devices running on lower software  Overview
versions. To ensure the stability of the application and Upon opening the application, users will encounter a
prevent crashes, Mobile Net is used as a backup. Mobile Net standard protocol consisting of two options: login and sign
itself is a lightweight model, making it easier to implement up. In the login process, users are required to provide either
on lower-end devices and applications. their registered email ID or mobile phone number. For sign
up, users must input their full name, gender, phone number,
and email address. Verification is carried out through an

IJISRT23JUL1498 www.ijisrt.com 1952

Volume 8, Issue 7, July – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
OTP (One-Time Password) system. All the aforementioned V. FEATURES
fields are mandatory. Additionally, above these fields, users
will encounter a terms and conditions page that must be  Bookmarking
approved before successfully creating an account. To While selecting a partner on a dating application, it is
complete the registration process, users are prompted to noticed that keeping a record of past profiles is not possible.
upload a selfie where their face is clearly visible. Similar to Therefore, this application introduces a feature namely
the aforementioned social media applications, our dating app bookmarking. With this feature, users can save profiles they
will incorporate similar features. Users can vertically scroll like or not sure about, for future reference. This convenient
to discover new profiles. Conversely, scrolling in the option allows users to revisit profiles and keep track of
opposite direction (downward) will display previously potential partners. The bookmarking feature is designed to
viewed or visited profiles. To indicate disinterest, users can be user-friendly, making it easy for users to use. It enables a
horizontally scroll (left or right) to reject a profile. systematic organization of profiles based on individual
preferences, aiding in decision making. Additionally, the
To express mutual interest, both users must like each bookmarking feature proves useful for future reference, such
other's profiles, after which a messaging option becomes as sharing profiles with friends or family or saving them
available. By tapping once on another user's profile, discreetly.
additional photos and descriptions provided by that user are
revealed. The index page consists of three components: the  Theme Customization
navigation bar, the main body, and the sidebar. Theme customization plays a crucial role in any
application as it enhances user interest and improves the
 Navigation Bar :- overall user experience. By allowing users to customize
The navigation bar includes a profile button that allows themes, it becomes possible to improve the visual
users to view their own user profile, including their profile appearance and provide a seamless experience that aligns
picture, description, and the ability to add, modify, or delete with their expectations. Usually, the users have the option to
information. Users can specify their preferred gender and choose between dark and light themes on their phones. Here
answer compatibility questions, which are used to display the user can also set themes based on their preferred colour
suitable profiles. palette. We provide a range of theme options for users to
select from, allowing them to personalize their experience.
 Main Body :- Additionally, this app offers font customization, giving users
The main body serves as a container for displaying a variety of font size choices. This feature enhances user
user profiles. When viewing other users' photos, a double- experience by catering to different font preferences, such as
tap provides a more detailed view of the profile, including small or large size fonts. Users can customize the font size
the user's bio and additional photos. In the bottom right according to their own needs and preferences, resulting in an
corner, there is a bookmark option that allows users to save overall excellent user experience.
profiles for future reference. By scrolling vertically, users
can discover new profiles, and a like button is available at  Like, Dislike and Preview:-
the bottom of each photo, similar to other social media apps. In our dating application, we have incorporated a user-
If two users mutually like each other's profiles, it results in a friendly interface inspired by popular social media platforms
match, unlocking the chat option for them to communicate. such as Facebook, Instagram, and Twitter. Similar to the
In the case where a user wishes to reject a profile, they can swiping feature in Bumble and Tinder, users can explore
swipe left or right to dismiss it. other profiles by scrolling vertically. To express interest,
users can double-tap on a profile to indicate a like. A single
 Sidebar :- tap allows users to view the user's bio, photos, and other
The side bar offers various options for users to explore details. To dislike a profile, users can swipe either to the left
within the dating application. The account section allows or right.
users to access and edit their profile, including photos and
bio. A bookmark option is available to view saved profiles.
Users can also log out of their accounts using the logout
option. If any issues, errors, or glitches arise, users can seek
help and support through the designated option to report
problems and find solutions.

Overall, this design provides a user-friendly interface

with intuitive navigation, a comprehensive profile viewing
experience, and convenient features for managing
preferences and interacting with potential matches.

IJISRT23JUL1498 www.ijisrt.com 1953

Volume 8, Issue 7, July – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
or to create a new account. If the user is new to the
application, they are guided to create an account using their
email address. On the other hand, already registered users
can directly log in. Once a registered user logs in, a
verification code is dispatched to their email address. The
user is then requested to input this code. Successful
verification occurs when the entered code matches the one
provided in the email. Following verification, the user gains
access to the application's data. This verification code
system plays a crucial role in confirming the authenticity of
user email addresses[12].

B. Verification Looping:
Once the user's verification is complete and they gain
access to the data, a continuous loop is initiated. In this loop,
the user's image is captured and stored in the AWS database.
Every 60 seconds, a new image is captured and compared
with the previous image taken 60 seconds ago. If the
comparison reveals that the new image matches the previous
one, the process continues seamlessly without interrupting
the user. However, if another user begins using the
application and their image is captured after 60 seconds but
fails to match the previous image, a toast message
displaying "User's face not detected" will appear on the
application screen[8].The application will proceed smoothly
once the user's face is detected, but if detection fails, the
alert message will persist.

C. Pre-Processing of the Data:

Pre-processing of an image is the process of making
changes in the image to make it more befitting and decent
for feeding it to the corresponding machine learning
algorithm or neural network. It amplifies the image quality,
decreases image size and removes unwanted data or
imperfections from the image. This process occurs after the
photo of the user is captured at the beginning, during the
user registration, and also after each time the image is
captured, i.e., every 60 sec. There are many techniques
included in the pre-processing of an image such as
normalization, noise prevention, increasing the contrast,
cropping and many more and the order in which these
techniques are performed varies from image to image.

 Image Capturing
The first and foremost process is capturing the image
from the front camera. There are two images to be compared
in this application, the first is manually recorded during the
registration process where the user’s photo is captured and
saved in the database, and the second is recorded every 60
sec while using the app. Both the images are pre-processed
Fig 1.Backend Functionality of the app
after they are captured and then optimized further to send it
to the neural network[10].
VI. BACKEND
 Image Normalization
A. Authentication using AWS:
During this process, the pixel values of an image are
As discussed previously, let's now explore the
rescaled to a range between 0 and 1. This technique is
authentication process within AWS. AWS is widely
employed to convert the pixel intensity values of each image
recognized for its strong focus on security. When a user
within this range, allowing for easier and more efficient
accesses an application integrated with AWS, they are
image comparison. As a result, the captured photos in the
required to go through an authentication procedure. During
application are normalized, ensuring that the comparison of
this process, the user is prompted to provide their login
images is conducted on a standardized scale.
credentials. They are presented with two options: to sign in

IJISRT23JUL1498 www.ijisrt.com 1954

Volume 8, Issue 7, July – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
 Noise Removal  Contrast Amplification
Noise removal is a crucial step in image processing, For instance if the picture is taken in a dark
although it is not necessary for every image as some may background or the surrounding lighting is weak, in that case
have minimal or no noise depending on the camera quality. we need the contrast to be enhanced so that the face can be
There are various types of noise models, including uniform, properly visible and identified. Therefore after using
impulse, exponential, Gaussian, Rayleigh, and more. The contrast amplification, we can perform further
common approach to noise removal is filtering, where filters functionalities such as image optimization and feature
are applied to the images. Filters are typically square extraction. It catalyzes the process of face identification to a
matrices, also called masks, smaller than the image size, large extent.
with 3x3 matrices being commonly used. Different types of
filters exist for filtering, consisting of pre-discovered 3x3  Image Optimization
matrices that simplify the noise removal process. Each During the account creation process, a photograph of
matrix has a specific function and can effectively eliminate a the account holder is captured using the front camera.
particular type of noise. The two main categories of filters Subsequently, optimization techniques are applied to reduce
are spatial and statistical filters. Spatial filters include the image size while maintaining its quality as effectively as
arithmetic mean, geometric mean, harmonic mean, and possible. This same process is repeated every 60 seconds on
contraharmonic mean filters. Statistical filters encompass the newly captured image. The purpose is to verify whether
median, max, min, midpoint, and alpha-trimmed mean the same person who created the account is still using it by
filters. These filters provide distinct functionalities to comparing both images. Considering the substantial data
effectively remove noise from images. generated by the captured images, it becomes essential to
optimize each stored image. The optimization process
Spatial filters play a crucial role in image filtering. One involves three key factors: image compression, dimensions
such example is the arithmetic mean filter, which calculates and file format selection. To address these, we have
the average of all the elements within the filter's applied converted each image into the JPEG file format. This format
region in the image. Although this filter helps in reducing strikes a balance between reducing the image size without
noise, it also introduces a slight blurring effect to the image, compromising quality beyond a certain threshold. The result
which limits its frequent usage. On the other hand, the is a significantly reduced file size while maintaining the
geometric mean filter computes the geometric mean instead image's desired level of quality. The image is simply
of the arithmetic mean. Unlike the arithmetic mean filter, the compressed using python code where we have implemented
geometric mean filter preserves more image details, the image.save() function. This function allows the
resulting in less loss of fine details. Additionally, the developer to change the quality of the image using the
harmonic mean filter calculates the harmonic mean. Out of parameter quality. The range of values inside the quality
these three filters, the harmonic mean filter is considered the parameter lies between 0 to 95. The quality of the image can
most effective for noise removal. easily be altered by just inputting the quality value which we
need for the image.
Likewise, the statistical filters also use masks for the
filtering process and these filters help in eliminating noise in There are two types of image compression, lossy and
the same frequency range in the image. The filters used in lossless compression. In lossy compression, the size of the
this category select a single pixel value from the area of image is reduced along with the image quality whereas in
pixels it is occupying on the image and replace it with the lossless compression, the quality of the image remains intact
pixel in the center of the filter. The center pixel value is and the size of the image is reduced.
replaced with: the median value of all the pixels inside the
mask in the median filter, the maximum value of all the The next step is to take care of the dimensions of the
pixels inside the mask in the max filter, the minimum value image. When a photo is captured using the front or rear
of all the pixels inside the mask in the min filter and the camera of the phone, it is of high resolution having
average of the maximum and minimum value of all the dimensions over 1800 pixels which can acquire a lot of
pixels inside the mask in the midpoint filter. Median, max space in the database. Hence, we need to reduce the
and min filters are efficient in the removal of impulse noise dimensions of the image to an appropriate size to fit it in the
and midpoint filters in the removal of gaussian and uniform database. The final step is choosing the right file format. The
noise. three main file formats necessary for an image are PNG,
JPEG and GIF. PNG image format is a file type of
In this application, we have used a combination of uncompressed images. The images in this format are high
various types of filters which include spatial as well as resolution, high quality and higher file size. The JPEG file
statistical filters. Since the image captured from the front format contains compressed images. In this format, the
camera is a very raw form of an image and may or may not image is compressed using lossy as well as lossless
contain high resolution or high quality image, it might compression techniques in which the size of the image is
contain a lot of as well as many different types of noises. reduced to a certain extent and the quality is also not
Hence, we also need different types of filters for compromised below a certain limit. The file format used in
corresponding types of noises. this app for compression is JPEG format.

IJISRT23JUL1498 www.ijisrt.com 1955

Volume 8, Issue 7, July – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
A proper combination of all these three components,  Privacy and Security:
i.e., image quality, dimensions and file format, can reduce Ensuring user privacy and data security is paramount.
the size of the image by a significant quantity. We need to Performance evaluation can focus on the effectiveness of the
check for each image what will be its best combination of face registration and verification process in preventing
quality, dimensions and its file format to reduce its size to impersonation and ensuring transparency. Additionally,
the minimum value possible for that image. robust measures to safeguard user data and prevent
unauthorized access are essential.
To further reduce the size of the image, we can crop
the image to get rid of undesired space around the image, if  Matching Accuracy:
possible. This can also be very beneficial in reducing the The success of the partner matching algorithm is
size of the image as it is a crucial step in image critical. Performance can be assessed by evaluating the
preprocessing[9]. accuracy of matches based on horoscopes, interests,
compatibility percentages, and user satisfaction with
 Feature Extraction suggested matches. Aligning user preferences with actual
The image contains many important segments among matches can indicate the effectiveness of the matching
many trivial segments. We need to extract those parts of the mechanism.
image which are significant and relevant for us. These parts
include important patterns, outlines, and many more  Real-Time Chat:
characteristics which are important for us to extract in order The performance of the real-time messaging feature
to input them into the neural network. There are many can be gauged by its reliability, responsiveness, and user
different feature extraction methods such as edge detection, satisfaction. Factors such as message delivery speed, chat
corner detection, texture analysis, blob detection etc. Edge stability, and user engagement are important indicators of
detection finds edges between the subject and the performance in this area.
surrounding and tries to extract it so that the user’s face can
be detected[14]. Corner detection is used to extract the  Image Processing Efficiency:
corners in the image. This can be useful to those users Evaluating the performance of image processing
having some sharp face parts such as the chin. Blob algorithms, including facenet and machine learning models,
detection detects blobs in the image. A blob is something is crucial. Key aspects to consider are the speed, accuracy,
that is a large object having a bright appearance in a dark and resource efficiency of face verification, image
background. Hence blob detection can be used in this compression, and feature extraction. Performance
application if the photo is taken at night time or a dark assessment ensures efficient and effective image processing.
environment. Texture analysis is used to dissect the image
based on the different textures of the image.  Scalability:
The project's ability to handle increased user volume
 Storing the Features in AWS and data load is significant. Performance evaluation should
Once the feature data is prepared, you can proceed include testing the system's response time, resource
with the upload to AWS. By using the AWS CLI or SDK, utilization, and stability under high user traffic. Scalability is
you establish a connection to your chosen storage service, essential for accommodating growing user bases.
specify the destination bucket or database table, and provide
the file's path or location containing the feature data. The  System Responsiveness:
upload process will transfer the file from your local system The project's responsiveness is a key performance
to the AWS infrastructure. It is essential to have the indicator. Evaluating the system's ability to handle user
necessary permissions and access rights for the upload interactions, such as profile updates, image uploads, and
operation. Monitoring the upload progress and handling any matching requests, with minimal delay or latency can
errors or exceptions that may arise are also important to enhance the user experience.
ensure a successful completion of the upload process. After
the upload is completed, the feature data will be stored in the  Error Handling and Exception Management:
designated location within your chosen AWS storage Robust error handling mechanisms are important for
service, ready for further analysis or retrieval. maintaining smooth operations. Evaluating how the project
handles errors, exceptions, and unexpected scenarios, such
D. Performance as network disruptions or database failures, can determine
its resilience and overall performance.
 User Experience:
The project aims to deliver a seamless and user- By regularly monitoring performance, gathering user
friendly experience. Performance can be evaluated by feedback, and making iterative improvements, the project
measuring the ease of registration, profile customization, can optimize its performance and deliver a superior user
partner matching, and user interaction. Gathering user experience.
feedback and conducting satisfaction surveys can provide
valuable insights into the overall user experience.

IJISRT23JUL1498 www.ijisrt.com 1956

Volume 8, Issue 7, July – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
VII. RESULTS The integration of machine learning models, such as
facenet, has proven effective in the backend for face
registration, verification, and comparison. By capturing and
processing temporary images every 60 seconds, the
application ensures the user's identity is consistently
verified. The utilization of machine learning algorithms for
feature extraction and compression enhances processing
speed and optimizes storage space.

The project has successfully addressed various

challenges, including image preprocessing, noise removal,
and image optimization. The implementation of spatial and
statistical filters has significantly improved image quality by
reducing noise while preserving important details.

The performance evaluation of the application has

Fig 2: Real Time Chat Messaging demonstrated positive results. The user experience is
seamless, with efficient partner matching and real-time chat
functionality. The system has exhibited scalability, handling
increasing user volumes and data loads without
compromising performance. Privacy and security measures
have proven robust, safeguarding user data and preventing
unauthorized access.

Overall, the project's results highlight its contribution

to the field of dating applications by offering enhanced
privacy, user-friendly features, efficient face verification,
and image processing techniques. The successful
implementation and positive performance outcomes validate
the effectiveness of the proposed solution in providing a
secure and enjoyable dating experience for users.

VIII. CONCLUSION

Fig 3: Edit your Bio The development of a dating web application that
prioritizes user privacy and data security is a crucial step
The results of this project indicate the successful towards providing a safe and transparent platform for
development and implementation of a dating web individuals seeking romantic connections. By incorporating
application that prioritizes user privacy and data security. features such as face registration and verification, the
The application offers a range of features, including partner application ensures authenticity and reduces the risk of
matching based on horoscopes, interests, compatibility impersonation. The collection of basic user details, along
percentage, and location. By registering the user's face and with the implementation of machine learning models for
implementing regular face verification, the application facial recognition, helps in minimizing the creation of fake
ensures transparency and minimizes the risk of accounts and enhancing the overall user experience.
impersonation or unauthorized use.
The inclusion of personalized matching criteria, such
The user experience has been enhanced through as horoscopes, interests, compatibility percentages, and
intuitive navigation, profile customization options, and the location, allows users to connect with potential partners
ability to interact with potential matches. The scroll based on shared traits and preferences. Additionally, the
functionality for browsing profiles, double-tap for liking, introduction of convenient features like double-tapping to
and swipe gestures for navigation provide a user-friendly express interest, swiping for profile browsing, and
interface. The feature to bookmark profiles for future bookmarking options enhances user engagement and
reference adds convenience. facilitates efficient exploration of potential matches.

One notable advancement is the elimination of fees for On the backend, the utilization of machine learning
viewing people who liked the user's profile, which is models, including FaceNet, optimizes the processing of user
typically charged by other dating applications. This images while ensuring privacy. The temporary image
streamlined approach enables users to easily identify those capture and comparison process effectively verifies user
interested in them and initiate interactions without wasting identity, allowing the application to run smoothly. The
time. adoption of space-saving techniques for image feature

IJISRT23JUL1498 www.ijisrt.com 1957

Volume 8, Issue 7, July – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
extraction helps minimize resource requirements and [14 ]. Gaurav Kumar,Pradeep Kumar Bhatia, "A Detailed
improve processing speed. Review of Feature Extraction in Image Processing
Systems", pp.260952140
By combining privacy-focused design, personalized
matching, intuitive user interface, and robust backend
implementation, this dating web application strives to
provide a secure and enjoyable experience for users. As
technology continues to evolve, ongoing improvements and
updates will be necessary to address emerging challenges
and meet user expectations.

Overall, this research paper highlights the importance

of user privacy and data security in the context of dating
applications, demonstrating the potential of innovative
solutions to foster meaningful connections while
maintaining a high level of transparency and trustworthiness
in the digital dating landscape.

REFERENCES

[1 ]. Maria Stoicescu, “‘Social impact of online dating

platforms. A case study on tinder,” pp.20381722.
[2 ]. Maria-Valentina Stoicescu, Stefania Matei, Razvan
Rughinis “Sharing and Privacy in Dating Apps,”
pp.18793018
[3 ]. Yangyu Hu, Haoyu Wang, Yajin Zhou"Dating with
Scambots: Understanding the Ecosystem of
Fraudulent Dating Applications," pp.20968444.
[4 ]. Aaron Smith, maeve duggan "Online Dating &
Relationships" Pew Research Center, October 21,
2013
[5 ]. Ángel Castro, Juan Ramón Barrada "Dating Apps
and Their Sociodemographic and Psychosocial
Correlates: A Systematic Review" PMCID:
PMC7557852
[6 ]. Florian Schroff, Dmitry Kalenichenko , James
Philbin "FaceNet: A Unified Embedding for Face
Recognition and Clustering," pp. 7298682
[7 ]. Xinzheng Xu, Meng Du, Huanxiu Guo, Jianying
Chang, Xiaoyang Zhao"Lightweight FaceNet Based
on MobileNet," Vol.11 No.1, January 2021
[8 ]. Fengwei Gu, Jun Lu, Guihua Xia, Zhiguang Feng
"Face Verification Technology Based on FaceNet
Similarity Recognition Network" pp. 20890826
[9 ]. Ayodeji Olalekan Salau, Shruti Jain, "Feature
Extraction: A Survey of the Types, Techniques,
Applications" pp.19228068.
[10 ]. Gang Wang; Bo Li, Yongfei Zhang, Jinhui Yang,
"Background Modeling and Referencing for Moving
Cameras-Captured Surveillance Video Coding in
HEVC," pp.18164975
[11 ]. Jian-Xiang Rong, Lei Zhang, Hua Huang, Fang-Lue
Zhang, "IMU-Assisted Online Video Background
Identification", pp. 21818674
[12 ]. Siem Girmay, Faniel Samsom, Asad Masood
Khattak, "AI based Login System using Facial
Recognition", pp.21363102
[13 ]. Abdullah Alqahtani, Hina Gull, "Cloud Computing
and Security Issues—A Review of Amazon Web
Services", Volume 13, Number 22

IJISRT23JUL1498 www.ijisrt.com 1958