SECURED DATA STORAGE MANAGEMENT WITH DEDUPLICATION IN CLOUD COMPUTING AND LOCAL GPT INTEGRATION
Pavan Myana, Siddhi Patil, Navrati Saxena, William B. Andreopoulos · 2025
Cloud computing offers scalable storage but raises challenges in confidentiality, redundancy, and secure data sharing. This work presents a framework that integrates deduplication, Proxy Re-Encryption (PRE), and LocalGPT to address these issues. Deduplication is performed before encryption using SHA-256, reducing redundancy while preserving security. PRE enables file sharing by re-encrypting symmetric keys for authorized users without exposing plaintext or keys. To support private, intelligent document interaction, LocalGPT is integrated with local embeddings and vector storage, enabling retrieval-augmented question answering offline. The framework is evaluated for unauthorized access, duplicate detection, PRE workflow, and LocalGPT validation. Results demonstrate effective access control ($401 / 409$errors), efficient deduplication ($\sim 22 ~\text{ms}$end-to-end, 0.44 ms hashing), successful PRE-based sharing, and accurate retrieval with 1-2 s query latency. This shows the feasibility of combining secure storage with local large language models for privacy-preserving cloud data management.