• 

  Trust What You See

🛡️ VerifyAI — Project Story

Inspiration

We noticed that fake profiles, manipulated IDs, and AI-generated images are becoming more common online. This creates problems for platforms like social media, job portals, and communities. We wanted to build a tool that could instantly verify if a profile or ID is real, making online interactions safer and faster.

What it does

VerifyAI allows users to:

• Upload an ID or provide a profile link.
  
• Detect fake or manipulated images using AI.
  
• Extract and analyze metadata to find inconsistencies.
  
• Get a Trust Score that summarizes the authenticity of the profile or document.
  

All this happens within seconds, giving platforms a fast way to filter real users from suspicious ones.

How we built it

We built VerifyAI using a layered approach:

1. Image Analysis
   AI models detect if images or IDs are fake or edited.

2. Metadata Extraction
   Reads metadata like device info, timestamps, and editing history.

3. Profile Link Checking
   Checks basic info and activity patterns for suspicious behavior.

4. Trust Score System
   Combines all signals into a final score:
   $$ \text{Trust Score} = w_1(\text{Authenticity}) + w_2(\text{Metadata}) + w_3(\text{Behavior}) $$
   where (w_1, w_2, w_3) are weights assigned to each factor.

5. Frontend & Backend

   • Frontend: React, TypeScript, Tailwind CSS
     
   • Backend: Base44 (authentication, database, integrations)
     
   • Libraries: framer-motion, lucide-react, @tanstack/react-query, recharts

6. Deployment
   The app is hosted live at verifyai-21.base44.app.

Challenges we ran into

• Detecting AI-generated images because many look very realistic.
  
• Handling different image formats (JPG, PNG, HEIC) and missing metadata.
  
• Running multiple checks quickly while maintaining speed.
  
• Connecting frontend and backend and fixing small deployment bugs.
  
• Completing everything within the hackathon timeline.

Accomplishments that we're proud of

• Built a working AI model that detects fake or edited images.
  
• Created a full system that generates a Trust Score.
  
• Made a clean, user-friendly website for verification.
  
• Successfully deployed the project online at verifyai-21.base44.app.
  
• Learned to combine AI, metadata, and web development in one project.

What we learned

• How AI can detect manipulated or fake images.
  
• How to extract and analyze metadata for verification.
  
• How to combine multiple signals into a single Trust Score.
  
• Basics of building frontend, backend, and deploying online.
  
• Teamwork, debugging, and problem-solving under time pressure.
  
• The importance of creating a user-friendly interface.

What's next for VerifyAI

• Implement video/live-photo verification to detect deepfakes.
  
• Build an API or SDK for easy integration into other platforms.
  
• Add an admin dashboard to monitor verification stats.
  
• Collect feedback and real-world test data to improve AI accuracy.
  
• Build Chrome and Firefox extensions to allow verification directly in the browser once Base44 supports extensions.

Built With

• backend
• base44
• database
• framer
• lucide
• management
• motion
• query
• react
• tanstack
• typescript