AI Content Authenticity Detector

An advanced, full-stack enterprise platform designed to detect whether digital content (text, image, audio, video) is human-created, AI-generated, manipulated (deepfake), or synthetic.

Project Overview

With the rapid advancement of Generative AI, distinguishing between authentic human content and AI-generated media is becoming critically important. This system provides a multi-modal authenticity scoring engine, leveraging state-of-the-art Deep Learning models to analyze content and produce an Explainable AI (XAI) report.

Features

• Text Analysis: Detects content generated by LLMs (ChatGPT, Claude, Gemini, etc.) using Transformer-based NLP models (RoBERTa, DistilBERT). Analyzes perplexity, burstiness, and stylometry.
• Image Authenticity: Identifies GAN-generated images, Stable Diffusion artifacts, and manual manipulations using CNNs and Vision Transformers (ViT).
• Video Deepfake Detection: Spots facial warping, lip-sync anomalies, and frame inconsistencies.
• Audio Verification: Detects synthetic voices and AI voice cloning using spectrogram analysis.
• Fact-Checking Engine: Cross-references claims with trusted knowledge bases.
• Explainable AI (XAI): Generates human-readable reports and heatmaps explaining why content was flagged.
• Blockchain Notarization: Creates immutable hashes for verified authentic content.
• Real-Time Extension: Browser plugin for instant social media verification.

Architecture

This project is built using a modern, scalable microservices architecture:

• Frontend: Next.js, React, Tailwind CSS, Recharts (for analytics).
• Backend: FastAPI (Python), PostgreSQL, Redis.
• Machine Learning: PyTorch, Hugging Face Transformers, OpenCV.
• Deployment: Docker, Docker Compose.

See docs/ARCHITECTURE.md for detailed diagrams and system flow.

Getting Started

(Instructions will be added once the initial Docker setup is complete)

Use Cases

See docs/USE_CASES.md for real-world applications including academic integrity, journalism, and social media moderation.

Future Scope

• Live streaming detection (WebRTC integration).
• Mobile application for on-the-go verification.
• Federated learning to improve models without centralizing user data.