Anti-Cheat

This is the page where camera permissions are given and then we can access and give tests.

🚀 About the Project 💡 Inspiration

As education and hiring rapidly moved online, one question kept coming up: how can we truly trust remote exams?

Most proctoring tools felt either too expensive, invasive, or dependent on streaming video to external servers. That not only increased costs but also raised serious privacy concerns for students.

We wanted a better balance. Something as secure as a physical invigilator, yet as private as your own laptop.

So we built an AI Proctor that runs entirely inside the browser, where every frame is processed locally and no video ever leaves the device.

🤖 What it does

Our Enhanced Anti Cheat Platform transforms a regular webcam into an intelligent exam supervisor that monitors behavior in real time.

It can:

✔ Detect prohibited objects such as phones, books, laptops, and tablets ✔ Track head pose and gaze direction to identify suspicious movements ✔ Count the number of people in the frame to prevent impersonation ✔ Log violations locally with timestamps without storing or transmitting video

Everything happens instantly, privately, and fully on device.

🛠️ How we built it

We designed the system with performance and simplicity at the core.

✔ Next.js and React for a fast, responsive interface ✔ MediaPipe Face Mesh for high precision face landmarks and gaze tracking ✔ TensorFlow.js with COCO SSD for real time object detection inside the browser ✔ Redux Toolkit for managing alerts, logs, and application state ✔ NextAuth for secure authentication

All AI inference runs on the client side, eliminating server delays and protecting user data.

🧩 Challenges

Balancing accuracy, speed, and privacy was our biggest challenge.

Running multiple AI models together initially caused lag. Object detection sometimes confused hands with phones. Frequent state updates triggered unnecessary React re renders. And keeping everything local without blocking the UI required careful optimization.

We solved these with smarter detection intervals, confidence threshold tuning, web worker processing, and optimized state management. The result is a smooth and reliable experience even on everyday devices.

🏆 Accomplishments we are proud of

✔ Instant alerts with zero server latency ✔ 100 percent privacy first architecture with no video uploads ✔ Accurate phone and book detection ✔ Clean and intuitive dashboard for administrators ✔ Fully browser based with no installation required

🧠 What we learned

Modern browsers are far more powerful than we imagined. Running neural networks directly on the edge is not just possible, it is production ready.

We also discovered that simply showing visible monitoring feedback often discourages cheating more effectively than strict penalties.

🔮 What’s next

We are continuing to improve the platform with:

✔ Audio analysis to detect whispering or speech ✔ Advanced eye tracking for higher precision monitoring ✔ Screen sharing detection ✔ Integrations with LMS platforms like Canvas, Moodle, and Blackboard ✔ Offline support through Progressive Web App capabilities

Our mission is simple. Make secure online assessments accessible to everyone without sacrificing privacy.