Behind The Scenes

Overview

This is an AI based web application which leverages Machine Learning and Deep Learning to indentify potential fake profiles/accounts in Instagram. This is solely intended for educational purposes and shall be used for assisting users in making informed choices and promoting digital well-being.

Flow of Work

First things first, we get our hands dirty with web scraping. Using a library called InstaLoader, we automate the "stalking" process, grabbing data like follower / following counts, posts, profile pictures, bio, and links straight from Instagram. Don’t worry; it’s all above board (or as ethical as scraping can be). This eliminates the need for users to input data manually because, let’s face it, who has the patience for that today?
Raw data? No thanks!
It’s time to clean, organize, and transform that messy heap of numbers and binary information into something the machine learning model can actually digest. We engineer some fancy features like username-to-fullname similarity, the follower-following ratio and percentage of weird characters in the names. These quirks are what make fake profiles stand out like a sore thumb.
Now comes the fun part, teaching the machine how to spot fakes. We feed it all that juicy, processed data and train it with algorithms like Random Forest or SVM and even use Artificial Neural Networks to immitate life-like thinking processes of these homo-sapiens. With a whopping 92% accuracy, this model is basically the Sherlock Holmes of Instagram. And yes, it even comes with a confidence score because machines can brag too.
Who wants to stare at boring CUI screen ? That’s where web development saves the day. The sleek, user-friendly interface does all the heavy lifting—accepting profile links, crunching numbers in the backend, and spitting out results with visual flair. Think of it as the tuxedo-clad butler presenting your data on a silver platter.
Finally, it’s showtime! The entire setup is deployed on the cloud (because local hosting is completely outdated now). Now, users across the globe can use this tool, and the server won’t break a sweat. Scalable, reliable, and always ready to impress.

Materails & Inspirations

  1. Systematic Literature Review: Instagram Fake Account Detection Based on Machine Learning

    February 2022Engineering MAthematics and Computer Science (EMACS) Journal 4(1):25-31
    DOI:10.21512/emacsjournal.v4i1.8076
    LicenseCC BY-SA 4.0
    Authors:

    • Juandreas Ezarfelix
    • Nathannael Jeffrey
    • Novita Sari

  2. Fake Instagram Profile Identification and Classification using Machine Learning

    Publisher: IEEE

  1. Datasets :-
    • The Twitter Username Alias Dataset

      This page contains the username/real-name dataset from our publication on "Aligning Entity Names with Online Aliases on Twitter". The dataset contains 110k twitter usernames with their profile names, as well as 110k false username/profile pairs. All usernames were taken from public twitter pages, and aligned with the names entered on their profiles. The dataset is ideal for research in aligning how people create aliases or nicknames for themselves.

  1. Higher Accuracy
    • The model achieves a 92% accuracy, demonstrating exceptional reliability in identifying fake Instagram profiles.
  2. Automated Data Extraction
    • Unlike solutions that rely on manual user input for features, this project utilizes web scraping to automatically extract data directly from Instagram profiles.
    • This reduces the burden on users, eliminates human error in data entry, and enhances the precision of the input features.
  3. Comprehensive Feature Set
    • Your model considers both standard metrics (like followers, following, and posts) and additional derived features (e.g., ratio of numeric characters, username similarity, bio structure) to ensure thorough analysis.
    • Combining automated extraction and feature diversity enables a more holistic approach.
  4. User-Friendly Website Interface
    • Your project provides a professional and visually appealing website interface, making it accessible and intuitive for users with varying technical expertise.
    • The website includes features like step-by-step instructions, confidence scores, disclaimers, and additional pages for terms and conditions, further enhancing usability and professionalism.
  5. Time Efficiency
    • Automated web scraping significantly reduces the time needed for analysis, as users do not have to manually input data, making your tool faster and more efficient.
  6. Practical Application
    • Your project aligns with real-world usage by simulating how people interact with Instagram profiles without requiring them to gather data manually, thus reflecting a more realistic scenario.
  7. Broader Appeal
    • By integrating modern web development techniques, your project offers a seamless experience that caters to both tech-savvy users and those unfamiliar with technical processes.
  8. Educational and Research-Oriented
    • The tool is designed not just for functionality but also for ethical and educational use, with clear disclaimers, terms, and usage guidelines to prevent misuse.

Privcy Policy

At ValidGram.AI, we take user privacy very seriously. This project has been developed purely for educational purposes to showcase the capabilities of machine learning and web development in detecting fake profiles on Instagram. We want to assure you that no user data is collected, stored, or shared at any point during the use of this project.

Key Points of the Privacy Policy
No Data Storage The project does not have a backend database or any storage mechanism to save user-provided information. Any inputs provided by the user for testing purposes are processed in real-time and immediately discarded once the results are displayed.
No Third-Party Integration The project does not use any third-party tools, APIs, or services that collect or track user data. Your interactions remain entirely within the project interface.
Anonymity Guaranteed Users are not required to log in, register, or provide any personal or sensitive information to access or use the project. All usage is anonymous.
Purpose of the Project This project is strictly for demonstration and educational purposes, designed to provide insight into how machine learning algorithms can detect patterns associated with fake profiles. It is not intended for any commercial use, and it does not interact with Instagram's servers or violate any of their terms of service.
Transparency As developers, we value transparency. The project is open-source (if applicable), and you can review the code to verify that no data is collected or stored. This ensures that users can trust the platform and its operations.
No Tracking Mechanisms The project does not use cookies, trackers, or analytics tools to monitor user behavior. You can freely explore and test the functionality without concerns about being tracked or monitored.

Hiii !

I'm Soham,the creator behind this web application.

Hi, I’m Soham Dutta, a passionate developer with a knack for turning innovative ideas into reality. From diving deep into machine learning algorithms to crafting seamless web experiences, I thrive at the intersection of technology and creativity.
This project is a reflection of my commitment to leveraging AI for impactful, real-world solutions.When I’m not coding, you’ll probably find me exploring new tech trends or brainstorming the next big idea or probably screaming and strumming my guitar ( p.s. : my neighbours don't like me that much).
Let’s connect and build something extraordinary!

Card image cap