Secure Information Retrieval System for a Cloud–based Platform

Abstract

With the increasing reliance on cloud storage, ensuring data confidentiality and controlled access during information retrieval has become a major concern. This paper presents a secure cloudbased information retrieval system that enables privacy-preserving search over encrypted data stored in the cloud. The system is designed to protect sensitive documents during storage and retrieval, ensuring that only authorized users can perform keyword-based searches and access the content. Key components of the system include AES-256 encryption, Bloom filter-based secure indexing, trapdoor generation for obfuscated query processing, and role-based access control (RBAC) for enforcing access permissions. The methodology employed involves a multi-layered architecture that integrates natural language processing (NLP) for query interpretation, a preprocessing pipeline for cleaning and normalizing input text, and the BM25 relevance ranking algorithm for efficient and accurate document retrieval. The system was developed and tested using the Cranfield dataset, which was structured and indexed for encrypted querying. The results demonstrate significant improvements in both security and search performance. The system achieved a precision and recall rate exceeding 97%, with a ROC-AUC score of 1.00, indicating highly reliable classification of relevant documents. The system offers better protection of sensitive data, improved retrieval accuracy, and scalable access control. These results affirm the system's effectiveness for secure document handling in real-world cloud environments.