Encrypted Similarity Matching Framework for Privacy-Safe Video Deduplication in Cloud Storage

Abstract

The rapid growth of digital media sharing platforms has increased concerns regarding copyright protection, data privacy, and unauthorized duplication of multimedia content. The background of this project focuses on developing a secure video copy detection system that ensures content originality while maintaining data confidentiality through advanced encryption techniques. As online video sharing becomes more widespread, detecting duplicate content without exposing sensitive media data has become a critical challenge for cloud-based platforms. The primary problem addressed in this work is the difficulty of identifying duplicate videos securely while preserving user privacy. Traditional video management systems typically rely on direct file comparison or metadata analysis, which may compromise privacy, require extensive computational resources, or fail to detect duplicates effectively. Additionally, many systems cannot perform meaningful operations on encrypted data, forcing decryption before analysis and increasing security risks. To overcome these challenges, the proposed system introduces a Django-based web application integrated with Fully Homomorphic Encryption (FHE) using the Paillier cryptosystem. Video files are partially encrypted into numerical representations, enabling duplicate detection through mathematical operations directly on encrypted data without exposing original content. The system supports secure user authentication, encrypted video upload, duplicate detection, database storage, and controlled downloading of unique or reference videos. The significance of the proposed system lies in enhancing privacy-preserving multimedia management by combining encryption-based computation with automated duplicate detection. This approach improves data security, protects intellectual property, reduces storage redundancy, and provides a scalable framework for secure video sharing platforms.