• 

  Light Theme

• 

  Pc Dashboard

• 

  Dark Theme

Story of Inspiration

During COVID and recent elections, misinformation spread faster than verified facts, costing people money, health, and even lives. On the finance side, phishing scams and frauds are at an all-time high. I wanted to build something that combines my love for algorithms with a real-world solution that protects society. That’s how TrustNet was born to fight misinformation and scams with scalable algorithms.

What It Does

🌐 Users paste a URL, text, or financial message.

🤖 The system runs NLP + classification models trained on verified datasets.

📊 Graph algorithms detect propagation patterns (fake news often comes from low-credibility clusters).

🔒 In finance/cybersecurity, anomaly detection flags suspicious payment requests or phishing attempts.

✅ Returns a Trust Score (0–100) with reasons, e.g. “Likely fake (source unreliable, suspicious language).”

🔔 Browser extension / web app integration for instant checks.

How We Built It

Frontend: React + TailwindCSS

Backend: Node.js (Express) or Python (FastAPI)

AI/ML: Python (scikit-learn, HuggingFace Transformers for NLP)

Database: MongoDB / PostgreSQL for storing sources & labeled datasets

Algorithms:

NLP classification (BERT model fine-tuned on fake news dataset)

Graph centrality detection (for misinformation propagation)

Anomaly detection (Isolation Forest) for scams

Cloud: Deployed on Vercel (frontend) + Render/Heroku (backend)

CI/CD & Hosting: GitHub Actions + GitHub Pages demo

API Integration: Optionally fact-checking APIs (e.g. Google Fact Check Tools API)

Challenges We Ran Into

Getting high-quality labeled fake news datasets.

Combining NLP accuracy with graph-based verification.

Avoiding false positives you can’t just call everything fake.

Real-time performance while scanning URLs.

Accomplishments We’re Proud Of

Built a working prototype in weeks.

Achieved ~90% accuracy on a test dataset of news articles.

Integrated a browser extension demo so users can try it instantly.

Designed a Trust Score metric that is simple but powerful.

What We Learned

Power of algorithms in real-world social good applications.

How cybersecurity, finance, and social media problems overlap.

Importance of UI/UX users need simple yes/no answers, not raw AI complexity.

What’s Next In Trust Net : AI - Powered Fake News & scam Detection Platform

Scale the system to analyze real-time Twitter/X feeds.

Partner with fact-checking organizations.

Expand into banking APIs for real-time phishing/fraud detection.

Release as a Chrome/Firefox extension for mass adoption.

Built With

• fastapi-database:-mongodb/postgres-cloud-services:-vercel-(frontend)
• javascript-frameworks:-react
• languages:-python
• render/heroku-(backend)-ml/ai:-huggingface-transformers
• scikit-learn-others:-github-actions
• tailwind