CyberFury | AI Forensic Lab

CyberFury is a high-performance digital forensics tool designed to distinguish between authentic photography and AI-generated imagery. By combining deep learning pixel analysis with exhaustive metadata inspection, it provides a comprehensive verdict on the legitimacy of digital evidence.

---

Core Features

1. Neural Pixel Analysis

The engine utilizes the Organika/sdxl-detector transformer model to scan for "synthetic fingerprints"—microscopic patterns left behind by diffusion models like Stable Diffusion XL. It provides a real-time confidence score for every scan.

2. EXIF Forensic Suite

The tool deep-dives into the image header data to uncover:

• AI Software Signatures: Detects traces of Midjourney, DALL-E, ComfyUI, and more.
• Camera Integrity: Identifies missing "Make" and "Model" tags common in generated images.
• Format Discrepancies: Flags suspicious WebP or PNG files that lack standard capture metadata.

3. Integrated Fact-Checking Bridge

To ensure a 360-degree investigation, CyberFury provides direct portals to:

• Filename Source Search: Automatically queries Google for specific file naming conventions used by AI generators.
• Google Fact-Check Database: Cross-references the image against known debunked visual claims.
• Lens Integration: A streamlined workflow for reverse-image searching to find the original context of a file.

---

How It Works

1. Upload: Provide a JPG, PNG, or WEBP file for analysis.
2. Execute: The CyberFuryEngine processes the image through a GPU/CPU-accelerated neural network.
3. Metadata Scrub: The system extracts and parses EXIF tags, looking for "Suspicious Indicators."
4. Verdict: The UI generates a high-contrast report:
   • REAL: High confidence that the image was captured by a physical camera.
   • AI: High probability of synthetic generation.

---

Technical Architecture

• Frontend: Streamlit (Cyberpunk-themed UI)
• ML Engine: Hugging Face Transformers (AutoModelForImageClassification)
• Image Processing: PIL (Pillow)
• Inference: PyTorch (Support for CUDA acceleration)

---

By: Chirayu Mahajan

⚖ Disclaimer

Forensic analysis is probabilistic. While CyberFury uses state-of-the-art detection models, it should be used as one part of a broader investigative process. Highly edited real photos or heavily compressed AI images may occasionally yield false positives/negatives.