Скачать или смотреть Built an AI Forensic Lab to Catch Deepfakes & Scams (TrustShield Demo)

Scams aren’t obvious anymore. It’s no longer broken English and misspelled emails — now we’re facing deepfaked voices, perfectly edited invoices, and phishing sites that look completely real. Traditional antivirus tools simply can’t keep up.

That’s why we built TrustShield.

TrustShield isn’t “just another GPT wrapper.” It’s a forensic-grade AI engine that inspects what the human eye usually misses: pixel compression, metadata trails, UI inconsistencies, and hidden web behavior.

What makes TrustShield different

🔍 Time-Traveler Detection
Our Payment Analyzer knows the history of app UI design. If a screenshot claims to be from today but uses an outdated Google Pay layout from 2019, it’s instantly flagged as fake.

🩻 X-Ray Vision for Documents
We use Error Level Analysis (ELA) to generate heat maps of images. If someone edited a number or name in Photoshop, the altered pixels literally light up.

🧠 Smart Chat and Scam Conversation Analysis
Instead of keyword spam filters, TrustShield uses a waterfall AI system. A fast BERT model catches obvious junk while a specialized Gemini agent detects romance scams and social-engineering intent.

🌐 Sandboxed Safe Browser
When you scan a link, we don’t just check a blacklist. A headless browser safely opens the site, follows redirects, and uses computer vision to see whether it visually imitates real banking portals.

Tech Stack

⚙️ Backend: Python (FastAPI)
🕵️ Forensics: OpenCV and Playwright
🤖 AI: Google Gemini 3 and BERT
🎨 Frontend: React

Let me know what you think of the project in the comments! If you want to check out the code, the GitHub is below. 👇
github.com/aryxnsdfs/trust-shield-ai