Published Paper Details:

The exponential growth of textual data in cloud environments, particularly in applications like Vision-and-Language Navigation, poses critical challenges for storage efficiency and data confidentiality. Traditional deduplication techniques reduce redundant data but compromise security by exposing sensitive information. To address this, we propose DEDUCT, a secure and efficient hybrid deduplication system that combines client-side preprocessing and encryption with server-side content-defined chunking and storage optimization. DEDUCT ensures confidentiality through convergent encryption, enabling deduplication over encrypted data while resisting side-channel attacks. Experimental evaluation on benchmark textual datasets demonstrates a compression rate of up to 66% without sacrificing data security. The system supports multiple file types, secure file sharing, and role-based access control, making it highly suitable for resource-constrained devices and modern cloud infrastructure. This approach achieves a balance between security and storage efficiency, paving the way for secure file handling in large-scale cloud systems.