StreamTrace

📌 Project Story: StreamTrace

🔹 Inspiration

The rapid growth of digital sports broadcasting has made high-quality media widely accessible, but it has also increased unauthorized redistribution and piracy. Observing how easily sports content is copied and shared without permission inspired us to build a system that not only detects misuse but also traces its origin. We wanted to create a practical solution that demonstrates how technology can help protect digital assets in real time.

🔹 What We Learned

Through this project, we explored concepts from digital security and computer vision, especially perceptual hashing and watermarking techniques. We learned how small transformations in media (cropping, resizing, compression) affect detection and how robust algorithms can still identify similarities. We also gained experience in building interactive applications and designing systems that balance accuracy with performance.

We used similarity measurement based on Hamming distance:

$$ D = \sum_{i=1}^{n} (h_1^i \neq h_2^i) $$

where ( h_1 ) and ( h_2 ) are hash values of two media files.

🔹 How We Built the Project

We developed StreamTrace, an intelligent platform that combines fingerprinting and watermarking:

• Fingerprinting: Generated perceptual hashes for uploaded media to create unique identifiers.
• Watermarking: Embedded dynamic user-specific text into media for traceability.
• Detection Engine: Compared hashes of suspicious content with stored originals using similarity thresholds.
• Risk Scoring: Calculated a simple risk metric based on similarity:

$$ \text{Risk Score} = 100 - (d \times 15) $$

where ( d ) is the hash difference.

🔹 Challenges We Faced

One of the main challenges was ensuring detection accuracy even when media was modified. Slight changes in images sometimes increased hash differences, requiring careful tuning of thresholds. Another challenge was balancing simplicity with innovation—building a prototype that works reliably within limited time while still showcasing impactful features. Integrating watermarking with detection in a seamless workflow also required multiple iterations. 🔹 Outcome

The final prototype successfully demonstrates how sports media can be protected using a combination of AI-based fingerprinting and dynamic watermarking, offering both detection and traceability in a simple, scalable system.

Built With

• antigravity
• mongodb
• mongoose
• react