Inspiration
The increasing misuse of AI to create realistic deepfakes is fueling misinformation, identity theft, and digital fraud. We aimed to build a system that helps detect manipulated media and strengthen trust in digital content.
What it does
This project detects deepfake images or video frames using deep learning. Users upload media, and the system analyzes visual artifacts and facial inconsistencies to predict authenticity with a confidence score.
How we built it
We trained a CNN-based model on deepfake datasets. The system preprocesses media by extracting and analyzing faces, then runs predictions through a backend model with a simple user interface for uploads and results.
Challenges we ran into
Major challenges included dataset imbalance, maintaining model accuracy across varied media conditions, and optimizing performance for faster detection.
Accomplishments that we're proud of
We built a functional prototype capable of detecting manipulated media and demonstrating practical AI-based cybersecurity defense.
What we learned
We gained experience in computer vision, deep learning optimization, dataset handling, and building real-world AI security applications.