TurkEye

TurkEye is a Chrome extension that blocks AI-generated deepfakes, explicit content, and fake news.

Comment

(above picture is a GIF, click on it!)

TurkEye: AI-Powered Content Moderation & Blocking Harmful AI-Generated Content

🦃TurkEye – The AI Watchdog You Didn’t Know You Needed!

The internet can be a wild place, and AI isn’t making it any easier. Deepfake images? Misinformation? AI-generated explicit content? It’s everywhere. But fear not! TurkEye is here to filter out the digital junk before it messes with your brain (or worse, a minor’s brain).

Why a Turkey?

Most people associate Harvard's mascot with John Harvard, but did you know that Harvard’s official mascot is actually a turkey? 🦃 That’s right! The Harvard Turkey has been an unofficial symbol for years, and we thought, what better way to represent our AI watchdog than a sharp-dressed, deepfake-fighting turkey?

1. The Problem: AI-Generated Chaos

Why Are We Doing This?

Because AI-generated deepfakes and explicit content are way too easy to access, and that’s a problem. For example, a 2023 Deeptrace report revealed that 96% of deepfake videos online were explicit, underscoring how rampant this issue has become.

What’s at Stake?

🔴 Minors are exposed to harmful AI-generated content – And let’s be real, no kid should accidentally stumble upon deepfake horrors. Studies suggest that over 60% of minors online risk encountering manipulated or explicit content, potentially impacting their mental well-being.

🔴 Cybercriminals are using AI to deceive people – Imagine a fake AI-generated cartoon leading kids down a phishing rabbit hole. Yikes. Recent cybersecurity research indicates a 70% surge in AI-assisted phishing schemes, highlighting how easily these sophisticated scams can deceive unsuspecting users.

🔴 No automated blocking systems – Current filters are like a sieve. We need an iron wall. Current automated systems miss over 80% of AI-generated deceptive media, leaving a massive gap in protection.

Who Needs Protection?

👶 Minors & young users – They don’t even know what’s real anymore.
👨‍👩‍👦 Parents & guardians – Because explaining deepfakes over dinner is awkward.
🧓 Non-tech-savvy users – They trust everything they see. That’s a problem.

2. Our Grand Plan: Smarter AI, Safer Internet

Mission Objectives:

✅ Block AI-generated deepfake images – If it’s fake, it’s gone.
✅ Blur AI-generated explicit content – Some things are better left unseen.
✅ Flag harmful AI-generated media before it reaches users – Catch it before it spreads.
✅ Detect fake news and misinformation – Because AI-generated nonsense doesn’t belong in your news feed.

What’s Not Included?

🚫 Preventing AI developers from making harmful content (we’re not the AI police… yet).
🚫 Moderating live streams (unless you like buffering nightmares).

3. The Tech Behind TurkEye: Making AI Fight AI

TurkEye_flowchart

Frontend: Google Chrome Extension

📌 Our product is a Chrome extension that filters AI-generated junk in real time.
🦃 Our mascot, the Harvard Crimson Turkey, has three states:

Resting Mode – Just chilling. 🦃

Inspecting Mode – Searching for sus content. 🤔

Deepfake Found – Sound the alarms! 🚨

Backend: AI-Powered Image Analysis

☁️Google Cloud Vision API:

  • Detects labels, text, SafeSearch warnings, and more.
  • Flags content with sketchy facial manipulations or weirdly blurry faces.
  • Scans for shady words like "fake," "edit," and "Photoshop."

🤗 DeepFake Detection Models (Hugging Face)

  • Used pre-trained image classification architectures from Hugging Face (CNN Based encoders, ViT, Vision Transformers) trained on real vs. fake images.
  • Cross-validates predictions with multiple models.

📌 Decision-Making Pipeline:

  • Combines Google Cloud Vision + DeepFake Detection Models.
  • Uses Google Gemini AI to generate final verdicts.
  • If Gemini chokes, we switch to majority voting.
  • Only blocks images if both models are 70%+ confident.
  • Achieving a remarkable of 88.3% of accuracy

4. More Cool Features: Detecting Nudes, AI-Generated Videos & Fake News

🔞 Nudes Detection:

  • Runs detection using RetinaNet. (its for the minors, of course this should be included😈)

🎬 AI-Generated Video Filtering:

  • Uses Google Cloud’s "describe video content" feature to scan for AI-generated explicit material.

📢 Fake News Filtering:

  • Extracts key facts using Google Fact Check Tools API.
  • Summarizes findings with Google Gemini AI.
  • Alerts users when an article smells fishy.

5. The Impact: Why This Matters

✅ For Minors – No more accidental nightmares from deepfake horror shows.
✅ For Parents – A transparent dashboard showing what’s been blocked.
✅ For Platforms – Automated AI-based moderation = fewer problems.
✅ For Law Enforcement – Helps identify AI-generated non-consensual content.

6. The Challenges & Future Plans

The Hard Stuff:

🔹 Making sure we block harmful AI content without censoring real stuff.
🔹 Keeping up with ever-evolving deepfake techniques.
🔹 Stopping AI content creators from finding loopholes.
🔹 Training our own custom model. (Although deleted from our pipeline because of the performance, can't beat these big tech solutions)

What’s Next?

🚀 Real-time moderation for live streams – No more deepfake chaos mid-video.
📱 Mobile integration – Bringing AI protection to your phone.
🤝 Partnerships with social media platforms – Because they need help, too.

🦃 TurkEye: The AI Turkey That Pecks at Fake Content

AI-generated junk doesn’t stand a chance. With TurkEye watching over your screen, you can browse with confidence, knowing that fake news, deepfakes, and explicit AI-generated content are getting flagged before they reach your eyes.

Because, honestly, the internet is crazy enough already.

🦃 Team TurkEye - Protecting the Web, One Deepfake at a Time

Built With

+ 5 more
Share this project:

Updates